Keeping medium-voltage grid operation within secure limits

Transcription

1 Chaire académique ORES «Smart Grids Smart Metering» Journée d études, Fac. Polytech. UMons, 20 nov Keeping medium-voltage grid operation within secure limits Thierry Van Cutsem Hamid Soleimani Bidgoli

2 Context Distribution networks are expected to host larger amounts of dispersed renewable generation voltage and congestion (thermal overload) problems are expected to occur more often but, hopefully, over limited periods of time Test system with: 75 MV buses 22 DG units (doubly fed induction & small synchronous generators) 2

3 Context Reinforcing the network ( fit-and -forget ) to deal with such temporary problems would be too expensive there is a good opportunity to use Distributed Generation (DG) units as control means to remove the security limit violations this is a service for which DG unit operators/owners can be financially compensated market and interactions between actors under discussion Loads with new profiles e.g. electric vehicles, heat pumps, etc. flexible loads are expected to also provide control means through remote control complementing smart meters this presentation, however, focuses on DG units only. 3

4 Context Active distribution networks will be called to assist transmission system in normal, and even more in stressed, operating conditions e.g. power factor improvement at the point of connection Automatic control schemes are needed to assist the Distribution System Operator in correcting voltage or congestion emergencies keeping the MV grids within desired operating limits coordinating their actions with transmission system operator 4

5 Task 1. Interaction models Task 2. Long-term planning Task 6. Validation & generalisation GREDOR Gestion des Réseaux Electriques de Distribution Ouverts aux Renouvelables Task 3. Operational planning Task 5. Data collection Task 4. Real-time control 5

6 Desired features ( Centralized + system-wide model ) or ( Distributed + simple logic )? offers more advanced control capabilities requires a communication infrastructure but cost expected to be much lower than cost of reinforcing the network exploit less expensive controls first e.g. reactive power modulation preferred to active power curtailment act in a non discriminatory and transparent manner optimize a system-wide objective with efforts shared by all relevant DG units drive the system from the current (unacceptable) to the desired (secure) operating point should not rely on models which may not be available / accurate especially for loads (sensitivity to voltage not well known!) rely on a simplified model (e.g. infrequently updated) be robust with respect to inaccuracies of this simplified model 6

7 Centralized controller : inputs and outputs (voltage setpoint of LTC) controller V P, Q, V measurements set-points (refreshed (updated every ~ 10 s) s) P, Q, V u = P g Q g V tap active powers generated by DG units reactive powers generated by DG units P, Q, V voltage set-point of transformer Load Tap Changer (LTC) (optional) 7

8 At time k, the controller : Principle of Model Predictive Control uses the last measurements and a model to predict the system response at N p future times computes an optimal sequence of N c future controls (N c N p ) applies the first component only. At time k + 1, the whole procedure is repeated. predicted output measurement k k + 1 k + N p discrete time control u k k + 1 k + N c discrete time 8

9 Predicted system evolution At time k : voltages : V k + 1 = V k + S V u k u k 1 for i = 1,, N p measurements just received V k + 2 = V k S V u k + 1 u k V k + i = V k + i 1 + S V u k + i 1 u k + i 2 currents : I k + 1 = I k + S I u k u k 1 for i = 1,, N p measurements just received I k + i = I k + i 1 + S I u k + i 1 u k + i 2 9

10 Constraints on controls : for i = 1,, N c : u min u k + i u max on rate of change of controls : for i = 1,, N c : u min u k + i u k + i 1 u max on voltages : for i = 1,, N p : on currents : for i = 1,, N p : V min k + i V k + i V max k + i I k + i I max k + i progressive tightening of bounds 10

11 Multistep optimization problem (solved at time k) min u N c 1 P g k + i P ref (k + i) WP 2 + N c 1 Q g k + i Q ref (k + i) WQ 2 i=0 i=0 subject to : for i = 1,, N p V k + i = V k + i 1 + S V u k + i 1 u k + i 2 I k + i = I k + i 1 + S I u k + i 1 u k + i 2 V min k + i V k + i V max k + i I k + i I max k + i for i = 1,, N c u min u k + i u max u = P g Q g V tap u min u k + i u k + i 1 u max 11

12 Mode 1 Network data static data measurements set points State estimation DSO Controller DSO : Distribution System Operator Real-time measurements P Q Local controller P meas Q meas V meas Non Dispatchable DG units P, Q MPPT MPPT : Maximum Power Point Tracking 12

13 Test system 22 DG units controlled controls adjusted every 10 s N c = 3 N p = 3 (larger if LTC actions anticipated) 13

14 Case 1 Mode 1. Wind increase (t = s, all 22 wind generators). Congestion corrected by controller. 14

15 Corrective reports Mode 2 Decision by non-dso actor static data measurements set points Network data State estimation DSO Controller DSO : Distribution System Operator Real-time measurements P Q P meas Q meas V meas Dispatchable DG units P, Q 15

16 Case 2 Mode 2. Active power increased (t= s) on 13 dispatchable generators. Overvoltage corrected by controller. 16

17 Reference evolution in Modes 1 and 2 min N c 1 P g k + i P ref (k + i) WP 2 + N c 1 Q g k + i Q ref (k + i) WQ 2 i=0 i=0 Power changes of the DG units are not known by the controller by default, the last measured powers are used as reference Limit violation : entering corrective mode P ref (k) P ref (k + 1) P ref (k + N c ) P ref (k) P ref (k + 1) P ref (k + N c ) P ref (k) P ref (k + 1) P ref (k + N c ) (similarly for reactive power) discrete time possible improvement : use a short-term prediction of power evolution 17

18 Corrective reports Mode 3.a Network data Decision by non-dso actor P, Q static data measurements set points information State estimation Real-time measurements DSO Controller P Q 0 near-future schedule P meas Q meas V meas DG units P, Q 18

19 Corrective reports Mode 3.b DSO Operational planning static data measurements set points information Network data P, Q State estimation Controller 0 near-future schedule Real-time measurements P Q P, Q P meas Q meas V meas DG units 19

20 Reference evolution in Mode 3 min N c 1 P g k + i P ref (k + i) WP 2 + N c 1 Q g k + i Q ref (k + i) WQ 2 i=0 i=0 The power schedules of the DG units are known by the controller and are used as reference P ref (k + N c ) P ref (k + N c ) P ref (k + 1) P ref (k) P ref (k + 1) P ref (k) P ref (k) P ref (k + 1) P ref (k + N c ) (similarly for reactive power) discrete time the controller has better anticipation capability 20

21 Case 3 Mode 1 : 9 generators - wind increase (t= s) Mode 3 : 12 other generators - power changed (t= s) overvoltages corrected by controller generation schedule 21

22 Case 4 Improvement of power factor at transformer level Initial situation : 13 MW Q tfo = 5 Mvar unity power factor (Q g =0) Additional constraint in optimization : Q tfo 2 Mvar 22

23 Case 4 Improvement of power factor at transformer level. Load tap changer response anticipated, allowing temporary overvoltage 23

24 Handling of transformer Load Tap Changer (LTC) Separate operation of LTC the LTC maintains its distribution voltage within a dead-band it is an external control device, not adjusted by the controller but its actions are anticipated and taken into account by the controller V k + i = V k + i 1 + S V u k + i 1 u k + i 2 prediction horizon extended accordingly + V V tap V d γ k + i Integrated operation of LTC (k) = 1 if tap change predicted at time k = 0 otherwise the LTC voltage set-point V tap is used as control variable the controller handles V tap as a continuous variable, ignoring dead-band this approximation is corrected by the MPC scheme LTC actions triggered by the set-point change are also anticipated 24

25 Centralized controller collecting measurements and adjusting set-points of DG units to satisfy operating constraints: currents below limits voltages inside bounds power factor at connection point with transmission system relies on concept of Model Predictive Control moving the operating point progressively from current to desired state compensating for modelling inaccuracies (as a closed-loop control) anticipating the effect of known changes (Mode 3) uses a simple, infrequently updated sensitivity model takes into account the load tap changer operation as a separate controller Summary or by controlling its voltage set-point constrained optimization problem compatible with real-time operation 25

26 Perspectives Approach improved and extended in the context of GREDOR project Extensions of formulation : treat discrete controls as such in optimization reset DG units at maximum / scheduled power after emergency situation has been corrected and operating conditions improve treat flexible loads and storage devices as additional control variables mitigate high voltage problems in LV grid due to photo-voltaic installations etc. Implementation aspects : practical telecommunication needs tests on models of GREDOR DSO systems coordination with operational planning (Task 3 of GREDOR) etc. 26

27 References G. Valverde and T. Van Cutsem, Model predictive control of voltages in active distribution networks, IEEE Transactions on Smart Grids, T. Van Cutsem and G. Valverde, Coordinated voltage control of distribution networks hosting dispersed generation, 22 nd International Conference on Electricity Distribution (CIRED), Stockholm (Sweden), June 2013 G. Valverde and T. Van Cutsem, Control of Dispersed Generation to Regulate Distribution and Support Transmission Voltages, Proc IEEE PowerTech conference, Grenoble (France), June 2013 H. Soleimani Bidgoli, Model predictive control of congestion and voltage problems in active distribution networks, submitted for presentation at the 23 rd International Conference on Electricity Distribution (CIRED), Rome (Italy), June Thank you for your attention! 27

28 En fait il n a pas répondu à la question «will we be smart by twenty-twenty?» De la tarte aux myrtilles!!?? Non merci! Il paraît que cet hiver il vont couper le courant C est ça le «smart»? 28