Basic Concepts in Data Reconciliation. Chapter 6: Steady-State Data Reconciliation with Model Uncertainties

Size: px
Start display at page:

Download "Basic Concepts in Data Reconciliation. Chapter 6: Steady-State Data Reconciliation with Model Uncertainties"

Transcription

1 Chapter 6: Steady-State Data with Model Uncertainties

2 CHAPTER 6 Steady-State Data with Model Uncertainties 6.1 Models with Uncertainties In the previous chapters, the models employed in the DR were considered exact. That is to say, the DR algorithms force the reconciled data to satisfy the models, fyz ( ˆ, ˆ)= 0, exactly. However, perfect models rarely exist and the reconciliation of measured values will more likely be a compromise between inaccuracies in both measurements and process models. In such cases, the reconciled values of the process variables should be obtained by: T 1 minimizing J( yz ˆ, ˆ) = ( y yˆ) V ( y yˆ) (6.1) subject to f ( yˆ, zˆ)= δˆ g ( yˆ, zˆ) 0 where ˆδ is a C 1 vector of model residuals that are the estimates of random model errors f ( y, z)= δ.

3 CHAPTER 6 Steady-State Data with Model Uncertainties 6.1 Models with Uncertainties The process models employed in the DR formulated by (6.1) have uncertainties that account for inaccuracies generated by approximations and thus reflect the modeling error. The variance and covariance of the model errors, Cov( δ), indicate the accuracy of the models. For linear models, the variance and covariance of the model errors can be obtained analytically. But for nonlinear models, they are calculated either by a Monte Carlo simulation, or, neglecting correlations between the model errors, the variances can be approximated using a first order Taylor s expansion around the measurement values to give: V where σ j is the variance of the measurement y j. (6.2) 2 2 = M f i i σ j j= 1 y j y

4 CHAPTER 6 Steady-State Data with Model Uncertainties 6.1 Models with Uncertainties In the Monte Carlo simulation, a series of normally distributed noise for each variable, y j, having the same variance as the measurement, are generated by a computer program. Then the functions, f(y, z), are calculated, along with their variance and covariance. Generated noise vectors y1 y 2... ym M M M M δ = f(y, z) Calculated model errors δ δ... δ 1 2 C M M M M Calculate V(δ 1 ), V(δ 2 ), Cov(δ 1, δ 1 ),

5 CHAPTER 6 Steady-State Data with Model Uncertainties 6.2 Data Algorithm with Uncertain Models When there are uncertain models, the reconciled data should be obtained such that it simultaneously minimizes both the measurement and model errors. Therefore, the data reconciliation problem (6.1) becomes: T T 1 minimizing J( ˆ, ˆ) ( ˆ) 1 yz = y y V ( y yˆ) + f( yz ˆ, ˆ) Ω f( yz ˆ, ˆ) (6.3) subject to g ( yˆ, zˆ) 0 where Ω is the covariance matrix of the model errors. The data reconciliation algorithm formulated by (6.3) considers the reconciled data as being a compromise between the measurements and models. The variance of the models can also be treated as tuning parameters of the algorithm. In other words, if we have a lot of confidence about the models used,

6 CHAPTER 6 Steady-State Data with Model Uncertainties 6.2 Data Algorithm with Uncertain Models then we can set the variance at a small value, otherwise, we set the variance at a large value. There are two limiting cases: (i) The variance of the models is very large compared to the variance of the measurements, then the reconciled data found by (6.3) will be the actual measurements; (ii) The variance of the models is very small compared to the variance of the measurements, then the reconciled data found by (6.3) will behave like the models are exact, which is like using the conventional data reconciliation algorithm formulated by Equation 1.2.

7 CHAPTER 6 Steady-State Data with Model Uncertainties 6.3 Solutions to Data with Uncertain Models Note that the functions for the uncertain models in Equation 6.3 can be linear or nonlinear. For the simplest case, all the variables are measured and the model functions are linear; thus the data reconciliation problem is formulated as: T T 1 minimizing J( ˆ) ( ˆ) 1 y = y y V ( y yˆ) + ( Ayˆ) Ω ( Ayˆ) (6.4) J The conditions for the minimum of (6.4) are = 0. yˆ Rewriting the objective function gives: T T 1 ˆ ˆ 1 ˆ ˆ ˆ J( y) = ( y y) V ( y y) + y ( AΩ A) y Taking the first derivative, we have: J = -1 T 1 2 ( ˆ) + 2 Ω ˆ = yˆ V y y A Ay 0 (6.5)

8 CHAPTER 6 Steady-State Data with Model Uncertainties 6.3 Solutions to Data with Uncertain Models Rearranging Equation 6.5 gives: ˆ = ( Ω + ) T y A A V V y (6.6) which is the solution to the DR with linear uncertain models. For the cooling water network illustrated in Figure 1.3, suppose the true values of the flows satisfy mass imbalances for each node because of fluctuations in the plant. The variance of the imbalances are obtained as: Ω=

9 CHAPTER 6 Steady-State Data with Model Uncertainties 6.3 Solutions to Data with Uncertain Models Using Equation 6.6, the reconciled data constrained by the mass imbalances can be obtained by the following MATLAB code: ********************************************** y=[110.5;60.8;35.0;68.9;38.6;101.4]; V=[ ; ; ; ; ; ]; A=[ ; ; ; ]; O=[ ; ; ; ]; yhat=inv(a'*inv(o)*a+inv(v))*inv(v)*y *********************************************

10 CHAPTER 6 Steady-State Data with Model Uncertainties 6.3 Solutions to Data with Uncertain Models The calculation results are listed here in Table 6.1. Table 6.1: Results of data reconciliation with imbalanced models for a cooling water network Stream No. Raw (kt/h) Reconciled (kt/h) Adjustment (kt/h) CWS FI FI FI FI FI FI 1 2 CWR Figure

11 CHAPTER 6 Steady-State Data with Model Uncertainties 6.3 Solutions to Data with Uncertain Models It is worth noting that the reconciled data for the flows in Table 6.1 don t satisfy the mass balances for each plant. However, if we want to force the reconciled data to exactly satisfy the mass balances, we can artificially use very small values for the variance of the imbalances, and then using Equation 6.6, the reconciled flows will exactly match mass balances for each plant. On the other hand, if we do not have any confidence in the mass balances, we can use very large values for the variance of the imbalances, and then the reconciled data will be equal to the raw measurements.

12 CHAPTER 6 Steady-State Data with Model Uncertainties 6.3 Solutions to Data with Uncertain Models The variances of the model errors are treated as tuning parameters of the DR algorithm. The change in the reconciled data with the changing of the tuning parameters is illustrated in Figure Variance of models 0 Reconciled data Raw measurements Reconciled by exact models Figure 6.1: Change in reconciled data with the change of the model variance

13 CHAPTER 6 Steady-State Data with Model Uncertainties 6.3 Solutions to Data with Uncertain Models If there exist unmeasured variables in the constraints or if the uncertain models used in the DR are nonlinear, the problem can be formulated by Equation 6.3 and solved by such as the quasi-newton method. As mentioned in Chapter 1, data reconciliation techniques can also be used to estimate model parameters. The unknowns in the vector, ẑ, in (6.3) can be unmeasured variables or model parameters. It is worth noting that the number of model equations used in (6.3) must be larger than the number of unmeasured variables and model parameters, so that the DR algorithm has sufficient redundancies in the estimation.

14 CHAPTER 6 Steady-State Data with Model Uncertainties 6.4 Quiz Question 1: The reconciled data for DR constrained by uncertain models (a) will satisfy exact models. (b) will be equal to the raw measurements. (c) is a compromise between the measurements and models. (d) is all of the above. Question 2: In DR which has infinite variance for its model uncertainties, the reconciled data then (a) will satisfy exact models. (b) will be equal to the raw measurements. (c) is a compromise between the measurements and models. (d) is all of the above.

15 CHAPTER 6 Steady-State Data with Model Uncertainties 6.4 Quiz Question 3: For the model, y - x = δ, the variance for y and x are 2.0 and 2.2 respectively. The variance of the model errors δ is (a) = 0.2. (b) = 0.2. (c) 2.2. (d) = 4.2. Question 3: For the model, y 2 -x = δ, the variance for y and x are 2.0 and 2.2 respectively. The variance of δ is (a) = 6.2. (b) = 1.8. (c) calculated by Monte Carlo simulation. (d) approximated by linearizing the function y 2 -x.

16 CHAPTER 6 Steady-State Data with Model Uncertainties 6.5 Suggested Readings Kao, E.P.C. (1997). An Introduction to Stochastic Processes. Duxbury Press, Belmont, Calif., USA. Maquin, D.; Adrot, O. and Ragot, J. (2000). Data reconciliation with uncertain models, ISA Transactions, 39, Mandel, D. ; Abdollahzadeh, A.; Maquin, D. and Ragot, J. (1998). Data reconciliation by inequality balance equilibration: a LMI approach. International Journal of Mineral Processing, 53,

17 Chapter 7:

18 7.1 Formulation of In the previous chapters, the data reconciliation problem was discussed for processes at steady state. However, many chemical processes are intrinsically dynamic and disturbances frequently occur. Consequently, it is desirable to develop dynamic data reconciliation (DDR) strategies for dynamic processes, so that at every sampling instant more accurate and reliable process data is available for real-time control and optimization. In steady-state data reconciliation, only one set of data is used at the current sampling time. Steady-state data reconciliation uses the spatial redundancy of the measurements. There is no previous information used to estimate the current state of the process.

19 7.1 Formulation of Along with the spatial redundancy of measurements, when a process is continuously sampled at discrete points in time, we also have temporal redundancy. The temporal redundancy of measurements is usually used for three types of state estimation as shown in Figure 7.1. Data used for estimation Data used for estimation Data used for estimation t Filtering Prediction Smoothing Figure 7.1: Types of estimation using temporal redundancy of measurements.

20 7.1 Formulation of When the measurements prior to time t, including the measurement at time t, are used to estimate the current state of the process, the estimation is called filtering. When the measurements prior to time t are used to predict the process variables at time t, t+1, t+2, etc., the estimation is called prediction. When the measurements prior to time t, and after time t, are used to estimate the process variables at time t, the estimation is called smoothing.

21 7.1 Formulation of Filtering and prediction can be applied on-line, whereas smoothing can only be used off-line. Dynamic data reconciliation uses both spatial and temporal redundancy to estimate the state of a process, so that at every sampling instant, more accurate and reliable process data are available. For discretely sampled measurements, the DDR problem is typically formulated as: t p T 1 minimizing J( yˆ, zˆ ) = [( y yˆ ) V ( y yˆ )] (7.1) t t t t t t t=0

22 7.1 Formulation of subject to d yˆ (t) f [, y ˆ(t), z ˆ(t)] = 0 dt gy [ ˆ(t), zˆ(t)]= 0 hy [ ˆ(t), zˆ(t)] 0 where y t is an M 1 vector of observed values for measured process variables at time t ( t = 0, 1, 2,, t p ), ŷ t is the vector of reconciled values for the measured process variables at time t, ẑ t is an N 1 vector of estimates for unmeasured variables or model parameters at time t, V is an M M covariance matrix for the measurements, f is a C 1 vector of differential model equality constraints, g is an S 1 vector of algebraic model equality constraints, and h is an E 1 vector of inequality model constraints including simple upper and lower bounds.

23 7.1 Formulation of The dynamic data reconciliation formulated by Equation 7.1 makes the reconciled data satisfy exact models. It applies a moving window approach where t p represents the window width. At each sampling time, only the measurements within the window width, t = 0, 1, 2,, t p, are reconciled, and only the reconciled values for the current time measurements are used for on-line monitoring and control. Other reconciled values of measurements prior to the current time are discarded. Then the moving window moves forward for the next time step. Moving window 0 t p Time Window width=10

24 7.2 Linear Often, process dynamics can be described or approximated by linear differential equations. For a discretely sampled system, it is convenient to use differential equations. The dynamics of a linear time invariant (LTI) process can be written as: xt = Axt-1 + But-1 + w (7.2) where x t is an M 1 vector of the true values of state variables at time t, u t-1 is a P 1 vector of manipulated inputs or disturbances to the process at time t-1, and w is an M 1 vector of random variables that represent model errors (accounting for model mismatch and unknown disturbances corrupting the process). A and B are matrices of appropriate dimensions whose coefficients are known at all times.

25 7.2 Linear In dynamic DR, at each sampling time, both model predicted values found by Equation 7.2 and measurements are available. Under the assumption that measurement and model errors are normally distributed with zero mean and known variance, the estimates (reconciled or adjusted values) of the measured process variables can be obtained by solving an optimization problem that simultaneously minimizes the weighted sum of square measurements and model errors: minimizing subject to J( yˆ ) = ( y yˆ ) V ( y yˆ ) + ( x yˆ ) Ω ( x yˆ ) T 1 T 1 t t t t t m,t t m,t t x = Ayˆ + Bu m,t t-1 t-1 y yˆ y l t u (7.3)

26 7.2 Linear where x m,t denotes the M 1 vector of model predicted values. The manipulated inputs or disturbances u t-1 in (7.2) are assumed to always be exactly known. In Equation 7.3, Ω is an M M variance-covariance matrix of the model predictions. The reconciled values at each sampling time, t, can be obtained analytically by Equation 7.3. Taking partial derivatives of the objective function with respect to ŷ t and setting it equal to zero results in: J yˆ t = 2 V ( y yˆ ) 2 Ω ( x yˆ ) = t t m,t t

27 7.2 Linear Rearranging the above equation gives: yˆ = ( V +Ω ) ( V y +Ω x ) (7.4) t t m,t In Equation 7.4, the variance and covariance of the measurements in V can be obtained from the raw measurements. However, it is sometimes difficult to obtain the variance and covariance of the model predictions in Ω. This is not only because of model inadequacies, but also because of various kinds of unknown disturbances corrupting the process. As a result, we can treat the ratio of the variance of the model predictions, σ m2, to that of the process measurements, σ 2, as tuning parameters in the DDR algorithm.

28 7.2 Linear When σ m2 / σ 2 is used as a tuning parameter for the DDR algorithm, the locus of the reconciled values typically follows an asymptotic curve as a function of σ m2 / σ 2. Figure 7.2, for example, illustrates the locus of a reconciled value with a change in σ m2 / σ 2. The measured value for the liquid level of a tank is 60.0 cm, and the model predicted value of the liquid level is 60.5 cm, at time t. The reconciled value is a point on the curve whose location depends on the values of σ m2 / σ 2, and is bounded by the measured and predicted values. There are two limiting cases: i) σ m2 / σ 2 = 0 which corresponds to an exact model, making the reconciled value equal to the model predicted value; and ii) σ m2 / σ 2 = indicating that the level of confidence on model predictions is poor, and the reconciled values approach the measured values.

29 7.2 Linear Model predicted Liquid level, cm Reconciled Measured σ 2 m/σ 2 Figure 7.2: Locus of reconciled data as a function of the ratio of model to measurement uncertainties in the DDR algorithm

30 7.2 Linear In process control, the presence of measurement noise will often deteriorate controller performance. It is therefore necessary to attenuate the measurement noise before it reaches the controllers. The linear DDR strategy introduced can be applied inside a control loop as a filter to reduce the propagation of measurement noise. The DDR algorithm as an integral part of conventional PID control loops is illustrated in Figure 7.3. The raw measurements of controlled variables are first reconciled by the DDR algorithm, and then used by the controller to calculate the control moves.

31 7.2 Linear Digital Controllers Reconciled data Control Valve Manipulated variables Process Controlled variables Raw measurements Noise Figure 7.3: Application of a DDR algorithm inside a PID control loop

32 7.2 Linear To illustrate dynamic DR strategies inside a control loop, consider a liquid storage tank process with a PI controller represented in Figure 7.3. FI Feed LC Outlet Figure 7.3: Schematic diagram of a storage tank process

33 7.2 Linear The diameter and height of the tank are 1.0 m and 1.2 m, respectively. A PI controller is used to regulate the liquid level of the tank by manipulating the outlet flow. The feed flow to the tank is measured but not controlled. The sampling interval is 1 minute. The measurement errors are assumed to be Gaussian white noise, and the standard deviations for the measurements of the feed flow rate and the liquid level are m 3 /h and m, respectively. The PI controller parameters are K c = -1.87m 2 /h and τ I = 30.0 min. The dynamics of the storage tank process corresponding to a 20% step change in feed flow rate was simulated, and the simulation results are presented in Figure 7.4.

34 7.2 Linear Tank level, h, cm PI controller KC = m 2.h -1 τ Ι = 30.0 min Time, min Figure 7.4: Dynamics of a storage tank for a 20% step change in feed flow at time t = 90 min Outlet flow, Fo, m3/h

35 7.2 Linear As shown in Figure 7.4, the measurements of the liquid level have noise. The manipulated variable, the outlet flow, exhibits considerable variations because the noisy measurements affect the calculated control moves. To attenuate the impact of the measurement noise, a DDR algorithm was embedded into the control loop to reconcile the liquid level before calculating the control moves. The reconciled liquid level at each sampling time is obtained by: minimizing J(h ˆ (7.5) t )= (F 2 i,t-1-f ˆ i,t-1) + (h 2 t-h ˆ ˆ t ) + (h 2 m,t-h t ) σf σh σm Δt subject to h m,t=h ˆ ˆ t-1+ (Fi,t-1-F o,t-1) A 0<h ˆ <h t max

36 7.2 Linear where the model constraint is the mass balance around the tank in discrete form, A is the cross-section of the tank, and Δt is the discrete sampling time. The ratio of the variance of the model predicted values to the variance of the measurements, σ m2 / σ 2, is given as 0.25 in testing the performance of the DDR algorithm. The same feed flow perturbation is used to evaluate the controller performance with the DDR algorithm embedded in the control loop. As shown in Figure 7.5, the reconciled data for the liquid level was less noisy than the raw measurements and closer to their true values. The manipulated variables displayed much smaller variations when the reconciled data was used to calculate the controller action.

37 7.2 Linear Tank level, h, cm Without DDR With DDR True PI controller KC = m 2.h -1 τi = 30.0 min Time, min Figure 7.5: Dynamics of storage tank with embedded DDR inside a PI control loop, for a 20% step change in feed flow at time, t = 90 min Outlet flow, Fo, m3/h

38 7.2 Linear This example demonstrates that the DDR algorithm developed can efficiently attenuate measurement noise, resulting in enhanced controller performance. In the application of DDR, dynamic models have to be developed. For the simple process of a storage tank, a fundamental model, the mass balance, was used. But for complex processes, for example, a distillation column, it is often difficult or impractical to develop fundamental dynamic models. Therefore, empirical input-output models have to be identified. Different techniques are available for developing an empirical dynamic model. These techniques are described under model identification in several textbooks (Ljung, 1987).

39 7.3 with Kalman Filter For convenience, we rewrite the linear time invariant, discrete, dynamic model (7.2) as: x = Ax + Bu + w t t-1 t-1 (7.6) where x t is the M 1 vector of the true values of process variables at time, t, u t-1 is the P 1 vector of process input variables (manipulated and disturbances) that are assumed to be exactly known at time, t-1, A is an M M matrix whose coefficients are known at all times, B is an M P matrix whose coefficients are also known at all times, and w is an M 1 vector of random variables which are assumed to be Gaussian white noise with zero mean and known variance. Equation 7.6 describes the dynamic evolution of the stochastic process variables.

40 7.3 with Kalman Filter The random variable vector, w, in this equation considers the following factors that affect the process: (i) random fluctuations of external disturbances, (ii) random errors in manipulated inputs resulting from electrical noise during control signal transmission and imprecise actuator positioning, and (iii) modeling error during model identification. If all the state variables are measured, the measurement model can be written as: y = x + ε (7.7) t t where ε is the M 1 vector of random measurement error which is also assumed to be Gaussian white noise with zero mean and known variance.

41 7.3 with Kalman Filter The two random vectors, w and ε, are assumed to uncorrelated. The statistical properties of these two random vectors are summarized as: E[ ] E[ ] Cov [ w] = R w = ε = 0 Cov [ ε] = V Cov w,w = Cov εε, = 0 i j i j i j Cov w, ε = 0 i j Given Equations 7.6 and 7.7, the optimal estimation of the process variables can be given by the Kalman Filter.

42 7.3 with Kalman Filter The detailed derivations of the Kalman filter can be found in the textbook Introduction to Optimal Estimation (Kamen and Su, 1999). The Kalman filter has a recursive form, and its equations are given by: xˆ = xˆ + K ( y xˆ ) - - t t t t t (7.8) - where ˆx t is the model predicted values by the deterministic model: - ˆ = ˆ +. t t-1 t-1 x Ax Bu (7.9) K t is called the Kalman gain and is given as: K = P( P + V). t t t (7.10) - - The matrix P t in (7.10) is the covariance matrix of ˆx t. The - matrix is given by: P t

43 7.3 with Kalman Filter - T P = AP A + R t t-1 where is the covariance matrix of ˆx, which is given by: Pt-1 t-1 P = ( I K ) P t-1 t-1 t-1 - (7.11) (7.12) The calculations of the recursive Equations 7.8~12 can be summarized as: Step 1 Filter Initialization: xˆ = E( x ), guess of xˆ P = I x xˆ x xˆ T λ ( λ>0), or guess of E[( )( ) ] Set t=1 Step 2 Time update:

44 7.3 with Kalman Filter - ˆ = ˆ +. t t-1 t-1 x Ax Bu - T P = AP A + R t t-1 Step 3 Measurement update: K = P( P + V). t t t - - xˆ = xˆ + K ( y xˆ ) t t t t t - P = ( I K ) P t t t Step 4 Time increment: Increment t and repeat step 2. It is important to note that the estimates in the LTI case generated by a Kalman filter have been proven to be unbiased and have a minimum mean squared error.

45 7.3 with Kalman Filter - In addition, the matrices P, P, and K t t t are independent of the measurements, y t, but are dependant upon the increment of time, t. They will become constants with the increasing of time, therefore, they can be calculated off-line. Recall the example of the storage tank process. If we neglect the reconciling of the feed flow, and solve the optimization problem (7.5), the reconciled liquid level, at each sampling time, t, can be obtained by the explicit equation: ĥ σ h + σ h 2 2 m t h m,t t = 2 2 σh + σm (7.13) Equation 7.13 can be written in the form: σ ĥ h (h h ) 2 t m = m,t t σh + σm m,t (7.14)

46 7.3 with Kalman Filter Compare the form of Equation 7.14 to Equation 7.8. The term σ /(σ + σ ) in (7.14) is equivalent to the Kalman gain in (7.8). m h m If we model the dynamics of the storage tank as a stochastic difference equation, we obtain: Δt t t-1 i,t-1 o,t-1 h=h + (F -F ) + w A (7.15) where w is Gaussian white noise with zero mean and known variance of R= m 2. The measurement model of the liquid level is: y = h + ε t t (7.16)

47 7.3 with Kalman Filter where ε is the white noise of the liquid level measurements with zero mean and known variance V= m 2. The optimal estimation for the liquid level of the tank, given the two models, (7.15) and (7.16), can be obtained by the Kalman filter: ĥt = hm,t + K(yt h m,t) where is the model predicted value given by: h m,t Δt h m,t=h ˆ t-1+ (Fi,t-1-F o,t-1) A and K is the Kalman gain. The value of the Kalman gain in this case can be calculated recursively through steps 1 to 4 as shown in Table 7.1.

48 7.3 with Kalman Filter Table 7.1: Calculation of Kalman gain for the storage tank Iteration P t P t The Kalman gain converges at K=0.39, so Equation 7.14 can be written as: ĥ = h (h h ). K t t m,t t m,t A guess of an initial value convergence

49 7.4 Nonlinear When nonlinear dynamic models are used as constraints for Equation 7.1 in dynamic data reconciliation, it is generally not possible to analytically obtain a discrete form of the differential constraint equations to represent the process. Liebman et al. (1992) proposed an approach, called nonlinear dynamic data reconciliation (NDDR) to solve this problem. The NDDR method discretizes the nonlinear differential equations by the method of orthogonal collocation on finite elements, such that the differential equations are transformed into algebraic equations at each sampling time. After the discretization, the remaining problem is to minimize the least-squares objective function with the constraints of algebraic equalities and inqualities that can be efficiently solved by the nonlinear programming techniques at each sampling time.

50 7.4 Nonlinear An alternative method to solving the nonlinear dynamic data reconciliation problem is to use the extended Kalman filter (EKF). This extension typically involves replacing the nonlinear model equations with first-order approximations around the current estimates. Detailed descriptions about the EKF method can be found in the textbook Introduction to Optimal Estimation (Kaman and Su, 1999).

51 7.5 Quiz Question 1: If a process is described by a stochastic dynamic model, the term, w, in Equation 7.2 considers (a) modeling error. (b) random external perturbations. (c) errors in input variables. (d) all of the above. Question 2: Measurement noise in a control loop (a) deteriorates the performance of the controller. (b) causes high-frequency oscillations of manipulated variables. (c) masks the true values of controlled variables. (d) all of above.

52 7.5 Quiz Question 3: In the application of Kalman filter, if K I, then (a) the output of the filter approaches model predicted values. (b) the output of the filter approaches raw measurements. (c) it indicates that the variance of model errors are very very large compared to the measurement variance. (d) all of the above. Question 4: The Kalman gain in a Kalman filter (a) is calculated recursively. (b) can be calculated off-line. (c) converges after a certain amount of iterations. (d) all of the above.

53 7.6 Suggested Readings Ljung, L. (1987). System Identification Theory for the User. Prentice-Hall, Englewood Cliff, N.J. Kamen, E.W. and Su, J.K. (1999). Introduction to Optimal Estimation, Springer, London. Bai, S. (2003). Assessment of Controller Performance with Embedded. Master thesis, University of Ottawa, Canada. Liebman, M.J.; Edgar, T.F. and Lasdon, L.S. (1992). Efficient data reconciliation and estimation for dynamic process using nonlinear programming techniques. Comp. Chem. Engng, 16,

SYSTEMTEORI - KALMAN FILTER VS LQ CONTROL

SYSTEMTEORI - KALMAN FILTER VS LQ CONTROL SYSTEMTEORI - KALMAN FILTER VS LQ CONTROL 1. Optimal regulator with noisy measurement Consider the following system: ẋ = Ax + Bu + w, x(0) = x 0 where w(t) is white noise with Ew(t) = 0, and x 0 is a stochastic

More information

1 Kalman Filter Introduction

1 Kalman Filter Introduction 1 Kalman Filter Introduction You should first read Chapter 1 of Stochastic models, estimation, and control: Volume 1 by Peter S. Maybec (available here). 1.1 Explanation of Equations (1-3) and (1-4) Equation

More information

Lessons in Estimation Theory for Signal Processing, Communications, and Control

Lessons in Estimation Theory for Signal Processing, Communications, and Control Lessons in Estimation Theory for Signal Processing, Communications, and Control Jerry M. Mendel Department of Electrical Engineering University of Southern California Los Angeles, California PRENTICE HALL

More information

ROBUST CONSTRAINED ESTIMATION VIA UNSCENTED TRANSFORMATION. Pramod Vachhani a, Shankar Narasimhan b and Raghunathan Rengaswamy a 1

ROBUST CONSTRAINED ESTIMATION VIA UNSCENTED TRANSFORMATION. Pramod Vachhani a, Shankar Narasimhan b and Raghunathan Rengaswamy a 1 ROUST CONSTRINED ESTIMTION VI UNSCENTED TRNSFORMTION Pramod Vachhani a, Shankar Narasimhan b and Raghunathan Rengaswamy a a Department of Chemical Engineering, Clarkson University, Potsdam, NY -3699, US.

More information

Stochastic Models, Estimation and Control Peter S. Maybeck Volumes 1, 2 & 3 Tables of Contents

Stochastic Models, Estimation and Control Peter S. Maybeck Volumes 1, 2 & 3 Tables of Contents Navtech Part #s Volume 1 #1277 Volume 2 #1278 Volume 3 #1279 3 Volume Set #1280 Stochastic Models, Estimation and Control Peter S. Maybeck Volumes 1, 2 & 3 Tables of Contents Volume 1 Preface Contents

More information

Online monitoring of MPC disturbance models using closed-loop data

Online monitoring of MPC disturbance models using closed-loop data Online monitoring of MPC disturbance models using closed-loop data Brian J. Odelson and James B. Rawlings Department of Chemical Engineering University of Wisconsin-Madison Online Optimization Based Identification

More information

DATA ASSIMILATION FOR FLOOD FORECASTING

DATA ASSIMILATION FOR FLOOD FORECASTING DATA ASSIMILATION FOR FLOOD FORECASTING Arnold Heemin Delft University of Technology 09/16/14 1 Data assimilation is the incorporation of measurement into a numerical model to improve the model results

More information

RECURSIVE ESTIMATION AND KALMAN FILTERING

RECURSIVE ESTIMATION AND KALMAN FILTERING Chapter 3 RECURSIVE ESTIMATION AND KALMAN FILTERING 3. The Discrete Time Kalman Filter Consider the following estimation problem. Given the stochastic system with x k+ = Ax k + Gw k (3.) y k = Cx k + Hv

More information

Introduction to System Identification and Adaptive Control

Introduction to System Identification and Adaptive Control Introduction to System Identification and Adaptive Control A. Khaki Sedigh Control Systems Group Faculty of Electrical and Computer Engineering K. N. Toosi University of Technology May 2009 Introduction

More information

Chapter 8. Feedback Controllers. Figure 8.1 Schematic diagram for a stirred-tank blending system.

Chapter 8. Feedback Controllers. Figure 8.1 Schematic diagram for a stirred-tank blending system. Feedback Controllers Figure 8.1 Schematic diagram for a stirred-tank blending system. 1 Basic Control Modes Next we consider the three basic control modes starting with the simplest mode, proportional

More information

Optimal Polynomial Control for Discrete-Time Systems

Optimal Polynomial Control for Discrete-Time Systems 1 Optimal Polynomial Control for Discrete-Time Systems Prof Guy Beale Electrical and Computer Engineering Department George Mason University Fairfax, Virginia Correspondence concerning this paper should

More information

ECONOMETRIC METHODS II: TIME SERIES LECTURE NOTES ON THE KALMAN FILTER. The Kalman Filter. We will be concerned with state space systems of the form

ECONOMETRIC METHODS II: TIME SERIES LECTURE NOTES ON THE KALMAN FILTER. The Kalman Filter. We will be concerned with state space systems of the form ECONOMETRIC METHODS II: TIME SERIES LECTURE NOTES ON THE KALMAN FILTER KRISTOFFER P. NIMARK The Kalman Filter We will be concerned with state space systems of the form X t = A t X t 1 + C t u t 0.1 Z t

More information

EECE Adaptive Control

EECE Adaptive Control EECE 574 - Adaptive Control Basics of System Identification Guy Dumont Department of Electrical and Computer Engineering University of British Columbia January 2010 Guy Dumont (UBC) EECE574 - Basics of

More information

Lecture 9. Introduction to Kalman Filtering. Linear Quadratic Gaussian Control (LQG) G. Hovland 2004

Lecture 9. Introduction to Kalman Filtering. Linear Quadratic Gaussian Control (LQG) G. Hovland 2004 MER42 Advanced Control Lecture 9 Introduction to Kalman Filtering Linear Quadratic Gaussian Control (LQG) G. Hovland 24 Announcement No tutorials on hursday mornings 8-9am I will be present in all practical

More information

Mini-Course 07 Kalman Particle Filters. Henrique Massard da Fonseca Cesar Cunha Pacheco Wellington Bettencurte Julio Dutra

Mini-Course 07 Kalman Particle Filters. Henrique Massard da Fonseca Cesar Cunha Pacheco Wellington Bettencurte Julio Dutra Mini-Course 07 Kalman Particle Filters Henrique Massard da Fonseca Cesar Cunha Pacheco Wellington Bettencurte Julio Dutra Agenda State Estimation Problems & Kalman Filter Henrique Massard Steady State

More information

Model predictive control of industrial processes. Vitali Vansovitš

Model predictive control of industrial processes. Vitali Vansovitš Model predictive control of industrial processes Vitali Vansovitš Contents Industrial process (Iru Power Plant) Neural networ identification Process identification linear model Model predictive controller

More information

DESIGNING A KALMAN FILTER WHEN NO NOISE COVARIANCE INFORMATION IS AVAILABLE. Robert Bos,1 Xavier Bombois Paul M. J. Van den Hof

DESIGNING A KALMAN FILTER WHEN NO NOISE COVARIANCE INFORMATION IS AVAILABLE. Robert Bos,1 Xavier Bombois Paul M. J. Van den Hof DESIGNING A KALMAN FILTER WHEN NO NOISE COVARIANCE INFORMATION IS AVAILABLE Robert Bos,1 Xavier Bombois Paul M. J. Van den Hof Delft Center for Systems and Control, Delft University of Technology, Mekelweg

More information

SELECTION OF VARIABLES FOR REGULATORY CONTROL USING POLE VECTORS. Kjetil Havre 1 Sigurd Skogestad 2

SELECTION OF VARIABLES FOR REGULATORY CONTROL USING POLE VECTORS. Kjetil Havre 1 Sigurd Skogestad 2 SELECTION OF VARIABLES FOR REGULATORY CONTROL USING POLE VECTORS Kjetil Havre 1 Sigurd Skogestad 2 Chemical Engineering, Norwegian University of Science and Technology N-734 Trondheim, Norway. Abstract:

More information

A Study of Covariances within Basic and Extended Kalman Filters

A Study of Covariances within Basic and Extended Kalman Filters A Study of Covariances within Basic and Extended Kalman Filters David Wheeler Kyle Ingersoll December 2, 2013 Abstract This paper explores the role of covariance in the context of Kalman filters. The underlying

More information

4 Derivations of the Discrete-Time Kalman Filter

4 Derivations of the Discrete-Time Kalman Filter Technion Israel Institute of Technology, Department of Electrical Engineering Estimation and Identification in Dynamical Systems (048825) Lecture Notes, Fall 2009, Prof N Shimkin 4 Derivations of the Discrete-Time

More information

Optimization-Based Control

Optimization-Based Control Optimization-Based Control Richard M. Murray Control and Dynamical Systems California Institute of Technology DRAFT v1.7a, 19 February 2008 c California Institute of Technology All rights reserved. This

More information

A New Subspace Identification Method for Open and Closed Loop Data

A New Subspace Identification Method for Open and Closed Loop Data A New Subspace Identification Method for Open and Closed Loop Data Magnus Jansson July 2005 IR S3 SB 0524 IFAC World Congress 2005 ROYAL INSTITUTE OF TECHNOLOGY Department of Signals, Sensors & Systems

More information

RECURSIVE SUBSPACE IDENTIFICATION IN THE LEAST SQUARES FRAMEWORK

RECURSIVE SUBSPACE IDENTIFICATION IN THE LEAST SQUARES FRAMEWORK RECURSIVE SUBSPACE IDENTIFICATION IN THE LEAST SQUARES FRAMEWORK TRNKA PAVEL AND HAVLENA VLADIMÍR Dept of Control Engineering, Czech Technical University, Technická 2, 166 27 Praha, Czech Republic mail:

More information

6.435, System Identification

6.435, System Identification System Identification 6.435 SET 3 Nonparametric Identification Munther A. Dahleh 1 Nonparametric Methods for System ID Time domain methods Impulse response Step response Correlation analysis / time Frequency

More information

5.1 2D example 59 Figure 5.1: Parabolic velocity field in a straight two-dimensional pipe. Figure 5.2: Concentration on the input boundary of the pipe. The vertical axis corresponds to r 2 -coordinate,

More information

Kalman Filter. Predict: Update: x k k 1 = F k x k 1 k 1 + B k u k P k k 1 = F k P k 1 k 1 F T k + Q

Kalman Filter. Predict: Update: x k k 1 = F k x k 1 k 1 + B k u k P k k 1 = F k P k 1 k 1 F T k + Q Kalman Filter Kalman Filter Predict: x k k 1 = F k x k 1 k 1 + B k u k P k k 1 = F k P k 1 k 1 F T k + Q Update: K = P k k 1 Hk T (H k P k k 1 Hk T + R) 1 x k k = x k k 1 + K(z k H k x k k 1 ) P k k =(I

More information

= m(0) + 4e 2 ( 3e 2 ) 2e 2, 1 (2k + k 2 ) dt. m(0) = u + R 1 B T P x 2 R dt. u + R 1 B T P y 2 R dt +

= m(0) + 4e 2 ( 3e 2 ) 2e 2, 1 (2k + k 2 ) dt. m(0) = u + R 1 B T P x 2 R dt. u + R 1 B T P y 2 R dt + ECE 553, Spring 8 Posted: May nd, 8 Problem Set #7 Solution Solutions: 1. The optimal controller is still the one given in the solution to the Problem 6 in Homework #5: u (x, t) = p(t)x k(t), t. The minimum

More information

Control Systems. State Estimation.

Control Systems. State Estimation. State Estimation chibum@seoultech.ac.kr Outline Dominant pole design Symmetric root locus State estimation We are able to place the CLPs arbitrarily by feeding back all the states: u = Kx. But these may

More information

Riccati difference equations to non linear extended Kalman filter constraints

Riccati difference equations to non linear extended Kalman filter constraints International Journal of Scientific & Engineering Research Volume 3, Issue 12, December-2012 1 Riccati difference equations to non linear extended Kalman filter constraints Abstract Elizabeth.S 1 & Jothilakshmi.R

More information

The Kalman Filter ImPr Talk

The Kalman Filter ImPr Talk The Kalman Filter ImPr Talk Ged Ridgway Centre for Medical Image Computing November, 2006 Outline What is the Kalman Filter? State Space Models Kalman Filter Overview Bayesian Updating of Estimates Kalman

More information

Every real system has uncertainties, which include system parametric uncertainties, unmodeled dynamics

Every real system has uncertainties, which include system parametric uncertainties, unmodeled dynamics Sensitivity Analysis of Disturbance Accommodating Control with Kalman Filter Estimation Jemin George and John L. Crassidis University at Buffalo, State University of New York, Amherst, NY, 14-44 The design

More information

Model Based Fault Detection and Diagnosis Using Structured Residual Approach in a Multi-Input Multi-Output System

Model Based Fault Detection and Diagnosis Using Structured Residual Approach in a Multi-Input Multi-Output System SERBIAN JOURNAL OF ELECTRICAL ENGINEERING Vol. 4, No. 2, November 2007, 133-145 Model Based Fault Detection and Diagnosis Using Structured Residual Approach in a Multi-Input Multi-Output System A. Asokan

More information

EECE Adaptive Control

EECE Adaptive Control EECE 574 - Adaptive Control Overview Guy Dumont Department of Electrical and Computer Engineering University of British Columbia Lectures: Thursday 09h00-12h00 Location: PPC 101 Guy Dumont (UBC) EECE 574

More information

On Identification of Cascade Systems 1

On Identification of Cascade Systems 1 On Identification of Cascade Systems 1 Bo Wahlberg Håkan Hjalmarsson Jonas Mårtensson Automatic Control and ACCESS, School of Electrical Engineering, KTH, SE-100 44 Stockholm, Sweden. (bo.wahlberg@ee.kth.se

More information

Linear Discrete-time State Space Realization of a Modified Quadruple Tank System with State Estimation using Kalman Filter

Linear Discrete-time State Space Realization of a Modified Quadruple Tank System with State Estimation using Kalman Filter Journal of Physics: Conference Series PAPER OPEN ACCESS Linear Discrete-time State Space Realization of a Modified Quadruple Tank System with State Estimation using Kalman Filter To cite this article:

More information

DESIGN AND IMPLEMENTATION OF SENSORLESS SPEED CONTROL FOR INDUCTION MOTOR DRIVE USING AN OPTIMIZED EXTENDED KALMAN FILTER

DESIGN AND IMPLEMENTATION OF SENSORLESS SPEED CONTROL FOR INDUCTION MOTOR DRIVE USING AN OPTIMIZED EXTENDED KALMAN FILTER INTERNATIONAL JOURNAL OF ELECTRONICS AND COMMUNICATION ENGINEERING & TECHNOLOGY (IJECET) International Journal of Electronics and Communication Engineering & Technology (IJECET), ISSN 0976 ISSN 0976 6464(Print)

More information

Least squares: introduction to the network adjustment

Least squares: introduction to the network adjustment Least squares: introduction to the network adjustment Experimental evidence and consequences Observations of the same quantity that have been performed at the highest possible accuracy provide different

More information

Gaussian Process Approximations of Stochastic Differential Equations

Gaussian Process Approximations of Stochastic Differential Equations Gaussian Process Approximations of Stochastic Differential Equations Cédric Archambeau Dan Cawford Manfred Opper John Shawe-Taylor May, 2006 1 Introduction Some of the most complex models routinely run

More information

IDENTIFICATION FOR CONTROL

IDENTIFICATION FOR CONTROL IDENTIFICATION FOR CONTROL Raymond A. de Callafon, University of California San Diego, USA Paul M.J. Van den Hof, Delft University of Technology, the Netherlands Keywords: Controller, Closed loop model,

More information

CALIFORNIA INSTITUTE OF TECHNOLOGY Control and Dynamical Systems. CDS 110b

CALIFORNIA INSTITUTE OF TECHNOLOGY Control and Dynamical Systems. CDS 110b CALIFORNIA INSTITUTE OF TECHNOLOGY Control and Dynamical Systems CDS 110b R. M. Murray Kalman Filters 14 January 2007 Reading: This set of lectures provides a brief introduction to Kalman filtering, following

More information

Here represents the impulse (or delta) function. is an diagonal matrix of intensities, and is an diagonal matrix of intensities.

Here represents the impulse (or delta) function. is an diagonal matrix of intensities, and is an diagonal matrix of intensities. 19 KALMAN FILTER 19.1 Introduction In the previous section, we derived the linear quadratic regulator as an optimal solution for the fullstate feedback control problem. The inherent assumption was that

More information

Extended Kalman Filter based State Estimation of Wind Turbine

Extended Kalman Filter based State Estimation of Wind Turbine Extended Kalman Filter based State Estimation of Wind Turbine Kavitha N*, Vijayachitra S# *PG Scholar,.E. Control and Instrumentation Engineering #Professor, Department of Electronics and Instrumentation

More information

MEMS Gyroscope Control Systems for Direct Angle Measurements

MEMS Gyroscope Control Systems for Direct Angle Measurements MEMS Gyroscope Control Systems for Direct Angle Measurements Chien-Yu Chi Mechanical Engineering National Chiao Tung University Hsin-Chu, Taiwan (R.O.C.) 3 Email: chienyu.me93g@nctu.edu.tw Tsung-Lin Chen

More information

Identification of a Chemical Process for Fault Detection Application

Identification of a Chemical Process for Fault Detection Application Identification of a Chemical Process for Fault Detection Application Silvio Simani Abstract The paper presents the application results concerning the fault detection of a dynamic process using linear system

More information

IN neural-network training, the most well-known online

IN neural-network training, the most well-known online IEEE TRANSACTIONS ON NEURAL NETWORKS, VOL. 10, NO. 1, JANUARY 1999 161 On the Kalman Filtering Method in Neural-Network Training and Pruning John Sum, Chi-sing Leung, Gilbert H. Young, and Wing-kay Kan

More information

6.241 Dynamic Systems and Control

6.241 Dynamic Systems and Control 6.241 Dynamic Systems and Control Lecture 24: H2 Synthesis Emilio Frazzoli Aeronautics and Astronautics Massachusetts Institute of Technology May 4, 2011 E. Frazzoli (MIT) Lecture 24: H 2 Synthesis May

More information

arxiv: v1 [cs.sy] 2 Oct 2018

arxiv: v1 [cs.sy] 2 Oct 2018 Non-linear Model Predictive Control of Conically Shaped Liquid Storage Tanks arxiv:1810.01119v1 [cs.sy] 2 Oct 2018 5 10 Abstract Martin Klaučo, L uboš Čirka Slovak University of Technology in Bratislava,

More information

Further Results on Model Structure Validation for Closed Loop System Identification

Further Results on Model Structure Validation for Closed Loop System Identification Advances in Wireless Communications and etworks 7; 3(5: 57-66 http://www.sciencepublishinggroup.com/j/awcn doi:.648/j.awcn.735. Further esults on Model Structure Validation for Closed Loop System Identification

More information

Optimal control and estimation

Optimal control and estimation Automatic Control 2 Optimal control and estimation Prof. Alberto Bemporad University of Trento Academic year 2010-2011 Prof. Alberto Bemporad (University of Trento) Automatic Control 2 Academic year 2010-2011

More information

Adaptive Robust Control

Adaptive Robust Control Adaptive Robust Control Adaptive control: modifies the control law to cope with the fact that the system and environment are uncertain. Robust control: sacrifices performance by guaranteeing stability

More information

Lecture 6: Bayesian Inference in SDE Models

Lecture 6: Bayesian Inference in SDE Models Lecture 6: Bayesian Inference in SDE Models Bayesian Filtering and Smoothing Point of View Simo Särkkä Aalto University Simo Särkkä (Aalto) Lecture 6: Bayesian Inference in SDEs 1 / 45 Contents 1 SDEs

More information

The Kalman Filter. Data Assimilation & Inverse Problems from Weather Forecasting to Neuroscience. Sarah Dance

The Kalman Filter. Data Assimilation & Inverse Problems from Weather Forecasting to Neuroscience. Sarah Dance The Kalman Filter Data Assimilation & Inverse Problems from Weather Forecasting to Neuroscience Sarah Dance School of Mathematical and Physical Sciences, University of Reading s.l.dance@reading.ac.uk July

More information

On Input Design for System Identification

On Input Design for System Identification On Input Design for System Identification Input Design Using Markov Chains CHIARA BRIGHENTI Masters Degree Project Stockholm, Sweden March 2009 XR-EE-RT 2009:002 Abstract When system identification methods

More information

ARIMA Modelling and Forecasting

ARIMA Modelling and Forecasting ARIMA Modelling and Forecasting Economic time series often appear nonstationary, because of trends, seasonal patterns, cycles, etc. However, the differences may appear stationary. Δx t x t x t 1 (first

More information

OPTIMAL CONTROL AND ESTIMATION

OPTIMAL CONTROL AND ESTIMATION OPTIMAL CONTROL AND ESTIMATION Robert F. Stengel Department of Mechanical and Aerospace Engineering Princeton University, Princeton, New Jersey DOVER PUBLICATIONS, INC. New York CONTENTS 1. INTRODUCTION

More information

Nonlinear State Estimation! Particle, Sigma-Points Filters!

Nonlinear State Estimation! Particle, Sigma-Points Filters! Nonlinear State Estimation! Particle, Sigma-Points Filters! Robert Stengel! Optimal Control and Estimation, MAE 546! Princeton University, 2017!! Particle filter!! Sigma-Points Unscented Kalman ) filter!!

More information

CONTROL SYSTEMS, ROBOTICS AND AUTOMATION - Vol. XIII - Nonlinear Observers - A. J. Krener

CONTROL SYSTEMS, ROBOTICS AND AUTOMATION - Vol. XIII - Nonlinear Observers - A. J. Krener NONLINEAR OBSERVERS A. J. Krener University of California, Davis, CA, USA Keywords: nonlinear observer, state estimation, nonlinear filtering, observability, high gain observers, minimum energy estimation,

More information

Steady State Kalman Filter

Steady State Kalman Filter Steady State Kalman Filter Infinite Horizon LQ Control: ẋ = Ax + Bu R positive definite, Q = Q T 2Q 1 2. (A, B) stabilizable, (A, Q 1 2) detectable. Solve for the positive (semi-) definite P in the ARE:

More information

Predictive Control of Gyroscopic-Force Actuators for Mechanical Vibration Damping

Predictive Control of Gyroscopic-Force Actuators for Mechanical Vibration Damping ARC Centre of Excellence for Complex Dynamic Systems and Control, pp 1 15 Predictive Control of Gyroscopic-Force Actuators for Mechanical Vibration Damping Tristan Perez 1, 2 Joris B Termaat 3 1 School

More information

Nonlinear State Estimation! Extended Kalman Filters!

Nonlinear State Estimation! Extended Kalman Filters! Nonlinear State Estimation! Extended Kalman Filters! Robert Stengel! Optimal Control and Estimation, MAE 546! Princeton University, 2017!! Deformation of the probability distribution!! Neighboring-optimal

More information

9 Multi-Model State Estimation

9 Multi-Model State Estimation Technion Israel Institute of Technology, Department of Electrical Engineering Estimation and Identification in Dynamical Systems (048825) Lecture Notes, Fall 2009, Prof. N. Shimkin 9 Multi-Model State

More information

Miscellaneous. Regarding reading materials. Again, ask questions (if you have) and ask them earlier

Miscellaneous. Regarding reading materials. Again, ask questions (if you have) and ask them earlier Miscellaneous Regarding reading materials Reading materials will be provided as needed If no assigned reading, it means I think the material from class is sufficient Should be enough for you to do your

More information

Organization. I MCMC discussion. I project talks. I Lecture.

Organization. I MCMC discussion. I project talks. I Lecture. Organization I MCMC discussion I project talks. I Lecture. Content I Uncertainty Propagation Overview I Forward-Backward with an Ensemble I Model Reduction (Intro) Uncertainty Propagation in Causal Systems

More information

Stochastic and Adaptive Optimal Control

Stochastic and Adaptive Optimal Control Stochastic and Adaptive Optimal Control Robert Stengel Optimal Control and Estimation, MAE 546 Princeton University, 2018! Nonlinear systems with random inputs and perfect measurements! Stochastic neighboring-optimal

More information

SOLVING POWER AND OPTIMAL POWER FLOW PROBLEMS IN THE PRESENCE OF UNCERTAINTY BY AFFINE ARITHMETIC

SOLVING POWER AND OPTIMAL POWER FLOW PROBLEMS IN THE PRESENCE OF UNCERTAINTY BY AFFINE ARITHMETIC SOLVING POWER AND OPTIMAL POWER FLOW PROBLEMS IN THE PRESENCE OF UNCERTAINTY BY AFFINE ARITHMETIC Alfredo Vaccaro RTSI 2015 - September 16-18, 2015, Torino, Italy RESEARCH MOTIVATIONS Power Flow (PF) and

More information

Prof. Dr. Roland Füss Lecture Series in Applied Econometrics Summer Term Introduction to Time Series Analysis

Prof. Dr. Roland Füss Lecture Series in Applied Econometrics Summer Term Introduction to Time Series Analysis Introduction to Time Series Analysis 1 Contents: I. Basics of Time Series Analysis... 4 I.1 Stationarity... 5 I.2 Autocorrelation Function... 9 I.3 Partial Autocorrelation Function (PACF)... 14 I.4 Transformation

More information

Stochastic Collocation Methods for Polynomial Chaos: Analysis and Applications

Stochastic Collocation Methods for Polynomial Chaos: Analysis and Applications Stochastic Collocation Methods for Polynomial Chaos: Analysis and Applications Dongbin Xiu Department of Mathematics, Purdue University Support: AFOSR FA955-8-1-353 (Computational Math) SF CAREER DMS-64535

More information

FIR Filters for Stationary State Space Signal Models

FIR Filters for Stationary State Space Signal Models Proceedings of the 17th World Congress The International Federation of Automatic Control FIR Filters for Stationary State Space Signal Models Jung Hun Park Wook Hyun Kwon School of Electrical Engineering

More information

ROBOTICS 01PEEQW. Basilio Bona DAUIN Politecnico di Torino

ROBOTICS 01PEEQW. Basilio Bona DAUIN Politecnico di Torino ROBOTICS 01PEEQW Basilio Bona DAUIN Politecnico di Torino Probabilistic Fundamentals in Robotics Gaussian Filters Course Outline Basic mathematical framework Probabilistic models of mobile robots Mobile

More information

Infinite Horizon LQ. Given continuous-time state equation. Find the control function u(t) to minimize

Infinite Horizon LQ. Given continuous-time state equation. Find the control function u(t) to minimize Infinite Horizon LQ Given continuous-time state equation x = Ax + Bu Find the control function ut) to minimize J = 1 " # [ x T t)qxt) + u T t)rut)] dt 2 0 Q $ 0, R > 0 and symmetric Solution is obtained

More information

Lecture 6: Multiple Model Filtering, Particle Filtering and Other Approximations

Lecture 6: Multiple Model Filtering, Particle Filtering and Other Approximations Lecture 6: Multiple Model Filtering, Particle Filtering and Other Approximations Department of Biomedical Engineering and Computational Science Aalto University April 28, 2010 Contents 1 Multiple Model

More information

A COMPARISON OF TWO METHODS FOR STOCHASTIC FAULT DETECTION: THE PARITY SPACE APPROACH AND PRINCIPAL COMPONENTS ANALYSIS

A COMPARISON OF TWO METHODS FOR STOCHASTIC FAULT DETECTION: THE PARITY SPACE APPROACH AND PRINCIPAL COMPONENTS ANALYSIS A COMPARISON OF TWO METHODS FOR STOCHASTIC FAULT DETECTION: THE PARITY SPACE APPROACH AND PRINCIPAL COMPONENTS ANALYSIS Anna Hagenblad, Fredrik Gustafsson, Inger Klein Department of Electrical Engineering,

More information

Kalman Filter and Parameter Identification. Florian Herzog

Kalman Filter and Parameter Identification. Florian Herzog Kalman Filter and Parameter Identification Florian Herzog 2013 Continuous-time Kalman Filter In this chapter, we shall use stochastic processes with independent increments w 1 (.) and w 2 (.) at the input

More information

A NONLINEARITY MEASURE FOR ESTIMATION SYSTEMS

A NONLINEARITY MEASURE FOR ESTIMATION SYSTEMS AAS 6-135 A NONLINEARITY MEASURE FOR ESTIMATION SYSTEMS Andrew J. Sinclair,JohnE.Hurtado, and John L. Junkins The concept of nonlinearity measures for dynamical systems is extended to estimation systems,

More information

2.830J / 6.780J / ESD.63J Control of Manufacturing Processes (SMA 6303)

2.830J / 6.780J / ESD.63J Control of Manufacturing Processes (SMA 6303) MIT OpenCourseWare http://ocw.mit.edu 2.830J / 6.780J / ESD.63J Control of Manufacturing Processes (SMA 6303) Spring 2008 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.

More information

The Applicability of Adaptive Control Theory to QoS Design: Limitations and Solutions

The Applicability of Adaptive Control Theory to QoS Design: Limitations and Solutions The Applicability of Adaptive Control Theory to QoS Design: Limitations and Solutions Keqiang Wu David J. Lilja Haowei Bai Electrical and Computer Engineering University of Minnesota Minneapolis, MN 55455,

More information

Kalman Filter Computer Vision (Kris Kitani) Carnegie Mellon University

Kalman Filter Computer Vision (Kris Kitani) Carnegie Mellon University Kalman Filter 16-385 Computer Vision (Kris Kitani) Carnegie Mellon University Examples up to now have been discrete (binary) random variables Kalman filtering can be seen as a special case of a temporal

More information

Factor Analysis and Kalman Filtering (11/2/04)

Factor Analysis and Kalman Filtering (11/2/04) CS281A/Stat241A: Statistical Learning Theory Factor Analysis and Kalman Filtering (11/2/04) Lecturer: Michael I. Jordan Scribes: Byung-Gon Chun and Sunghoon Kim 1 Factor Analysis Factor analysis is used

More information

EL2520 Control Theory and Practice

EL2520 Control Theory and Practice EL2520 Control Theory and Practice Lecture 8: Linear quadratic control Mikael Johansson School of Electrical Engineering KTH, Stockholm, Sweden Linear quadratic control Allows to compute the controller

More information

State Estimation of Linear and Nonlinear Dynamic Systems

State Estimation of Linear and Nonlinear Dynamic Systems State Estimation of Linear and Nonlinear Dynamic Systems Part I: Linear Systems with Gaussian Noise James B. Rawlings and Fernando V. Lima Department of Chemical and Biological Engineering University of

More information

Bayes Filter Reminder. Kalman Filter Localization. Properties of Gaussians. Gaussians. Prediction. Correction. σ 2. Univariate. 1 2πσ e.

Bayes Filter Reminder. Kalman Filter Localization. Properties of Gaussians. Gaussians. Prediction. Correction. σ 2. Univariate. 1 2πσ e. Kalman Filter Localization Bayes Filter Reminder Prediction Correction Gaussians p(x) ~ N(µ,σ 2 ) : Properties of Gaussians Univariate p(x) = 1 1 2πσ e 2 (x µ) 2 σ 2 µ Univariate -σ σ Multivariate µ Multivariate

More information

I. D. Landau, A. Karimi: A Course on Adaptive Control Adaptive Control. Part 9: Adaptive Control with Multiple Models and Switching

I. D. Landau, A. Karimi: A Course on Adaptive Control Adaptive Control. Part 9: Adaptive Control with Multiple Models and Switching I. D. Landau, A. Karimi: A Course on Adaptive Control - 5 1 Adaptive Control Part 9: Adaptive Control with Multiple Models and Switching I. D. Landau, A. Karimi: A Course on Adaptive Control - 5 2 Outline

More information

Contents. PART I METHODS AND CONCEPTS 2. Transfer Function Approach Frequency Domain Representations... 42

Contents. PART I METHODS AND CONCEPTS 2. Transfer Function Approach Frequency Domain Representations... 42 Contents Preface.............................................. xiii 1. Introduction......................................... 1 1.1 Continuous and Discrete Control Systems................. 4 1.2 Open-Loop

More information

Stability of Ensemble Kalman Filters

Stability of Ensemble Kalman Filters Stability of Ensemble Kalman Filters Idrissa S. Amour, Zubeda Mussa, Alexander Bibov, Antti Solonen, John Bardsley, Heikki Haario and Tuomo Kauranne Lappeenranta University of Technology University of

More information

LQR, Kalman Filter, and LQG. Postgraduate Course, M.Sc. Electrical Engineering Department College of Engineering University of Salahaddin

LQR, Kalman Filter, and LQG. Postgraduate Course, M.Sc. Electrical Engineering Department College of Engineering University of Salahaddin LQR, Kalman Filter, and LQG Postgraduate Course, M.Sc. Electrical Engineering Department College of Engineering University of Salahaddin May 2015 Linear Quadratic Regulator (LQR) Consider a linear system

More information

Industrial Model Predictive Control

Industrial Model Predictive Control Industrial Model Predictive Control Emil Schultz Christensen Kongens Lyngby 2013 DTU Compute-M.Sc.-2013-49 Technical University of Denmark DTU Compute Matematiktovet, Building 303B, DK-2800 Kongens Lyngby,

More information

Lie Groups for 2D and 3D Transformations

Lie Groups for 2D and 3D Transformations Lie Groups for 2D and 3D Transformations Ethan Eade Updated May 20, 2017 * 1 Introduction This document derives useful formulae for working with the Lie groups that represent transformations in 2D and

More information

Kalman Filters with Uncompensated Biases

Kalman Filters with Uncompensated Biases Kalman Filters with Uncompensated Biases Renato Zanetti he Charles Stark Draper Laboratory, Houston, exas, 77058 Robert H. Bishop Marquette University, Milwaukee, WI 53201 I. INRODUCION An underlying assumption

More information

NONLINEAR PLANT IDENTIFICATION BY WAVELETS

NONLINEAR PLANT IDENTIFICATION BY WAVELETS NONLINEAR PLANT IDENTIFICATION BY WAVELETS Edison Righeto UNESP Ilha Solteira, Department of Mathematics, Av. Brasil 56, 5385000, Ilha Solteira, SP, Brazil righeto@fqm.feis.unesp.br Luiz Henrique M. Grassi

More information

Solutions for Tutorial 10 Stability Analysis

Solutions for Tutorial 10 Stability Analysis Solutions for Tutorial 1 Stability Analysis 1.1 In this question, you will analyze the series of three isothermal CSTR s show in Figure 1.1. The model for each reactor is the same at presented in Textbook

More information

ON ENTRY-WISE ORGANIZED FILTERING. E.A. Suzdaleva

ON ENTRY-WISE ORGANIZED FILTERING. E.A. Suzdaleva ON ENTRY-WISE ORGANIZED FILTERING E.A. Suzdaleva Department of Adaptive Systems Institute of Information Theory and Automation of the Academy of Sciences of the Czech Republic Pod vodárenskou věží 4, 18208

More information

PROPORTIONAL-Integral-Derivative (PID) controllers

PROPORTIONAL-Integral-Derivative (PID) controllers Multiple Model and Neural based Adaptive Multi-loop PID Controller for a CSTR Process R.Vinodha S. Abraham Lincoln and J. Prakash Abstract Multi-loop (De-centralized) Proportional-Integral- Derivative

More information

Probabilistic Fundamentals in Robotics. DAUIN Politecnico di Torino July 2010

Probabilistic Fundamentals in Robotics. DAUIN Politecnico di Torino July 2010 Probabilistic Fundamentals in Robotics Gaussian Filters Basilio Bona DAUIN Politecnico di Torino July 2010 Course Outline Basic mathematical framework Probabilistic models of mobile robots Mobile robot

More information

If we want to analyze experimental or simulated data we might encounter the following tasks:

If we want to analyze experimental or simulated data we might encounter the following tasks: Chapter 1 Introduction If we want to analyze experimental or simulated data we might encounter the following tasks: Characterization of the source of the signal and diagnosis Studying dependencies Prediction

More information

A Model-Free Control System Based on the Sliding Mode Control Method with Applications to Multi-Input-Multi-Output Systems

A Model-Free Control System Based on the Sliding Mode Control Method with Applications to Multi-Input-Multi-Output Systems Proceedings of the 4 th International Conference of Control, Dynamic Systems, and Robotics (CDSR'17) Toronto, Canada August 21 23, 2017 Paper No. 119 DOI: 10.11159/cdsr17.119 A Model-Free Control System

More information

ANNEX A: ANALYSIS METHODOLOGIES

ANNEX A: ANALYSIS METHODOLOGIES ANNEX A: ANALYSIS METHODOLOGIES A.1 Introduction Before discussing supplemental damping devices, this annex provides a brief review of the seismic analysis methods used in the optimization algorithms considered

More information

ELEC4631 s Lecture 2: Dynamic Control Systems 7 March Overview of dynamic control systems

ELEC4631 s Lecture 2: Dynamic Control Systems 7 March Overview of dynamic control systems ELEC4631 s Lecture 2: Dynamic Control Systems 7 March 2011 Overview of dynamic control systems Goals of Controller design Autonomous dynamic systems Linear Multi-input multi-output (MIMO) systems Bat flight

More information

Nonlinear Observers. Jaime A. Moreno. Eléctrica y Computación Instituto de Ingeniería Universidad Nacional Autónoma de México

Nonlinear Observers. Jaime A. Moreno. Eléctrica y Computación Instituto de Ingeniería Universidad Nacional Autónoma de México Nonlinear Observers Jaime A. Moreno JMorenoP@ii.unam.mx Eléctrica y Computación Instituto de Ingeniería Universidad Nacional Autónoma de México XVI Congreso Latinoamericano de Control Automático October

More information

Cramér-Rao Bounds for Estimation of Linear System Noise Covariances

Cramér-Rao Bounds for Estimation of Linear System Noise Covariances Journal of Mechanical Engineering and Automation (): 6- DOI: 593/jjmea Cramér-Rao Bounds for Estimation of Linear System oise Covariances Peter Matiso * Vladimír Havlena Czech echnical University in Prague

More information

Five Formulations of Extended Kalman Filter: Which is the best for D-RTO?

Five Formulations of Extended Kalman Filter: Which is the best for D-RTO? 17 th European Symposium on Computer Aided Process Engineering ESCAPE17 V. Plesu and P.S. Agachi (Editors) 2007 Elsevier B.V. All rights reserved. 1 Five Formulations of Extended Kalman Filter: Which is

More information