ENER/FP7/296009/InSun InSun Industrial Process Heat by Solar Collectors DELIVERABLE 5.3 Report on Predictive control system development and performance tests 1 Work Package 5 Simulation based system design optimisation and performance observation 1 Update: Sept 2015 This document has been produced in the context of the InSun Project. The research leading to these results has received funding from the European Community's Seventh Framework Programme ([FP7/2007-2011]) under grant agreement n ENER/FP7/296009/InSun. All information in this document is provided "as is" and no guarantee or warranty is given that the information is fit for any particular purpose. The user thereof uses the information at its sole risk and liability. For the avoidance of all doubts, the European Commission has no liability in respect of this document, which is merely representing the authors view. Page 1 of 14
Content 1 Abbreviations... 3 2 Introduction... 5 3 Generation of radiation forecast data... 6 3.1 Logging of cloud cover data... 6 3.2 Calculation of hourly irradiance... 7 3.3 Comparison of predicted and measured data... 9 4 Determination of heating up energy demand and predictive control... 11 5 Summary... 14 6 References... 14 Table of figures Figure 1 Soltigua system scheme... 5 Figure 2 Principle of simulation based predictive control... 5 Figure 3 Free available hourly weather forecast... 6 Figure 4 Cloud cover scale according to DWD... 6 Figure 5 Predictive control: Comparison of measured and predicted yield... 9 Figure 6 Predictive control: Comparison of commercially predicted and calculated yield... 10 Figure 7 Daily measured data... 11 Figure 8 System start up data HTF (2014)... 12 Figure 9 System start up data HTF (2015)... 12 Figure 10 System start-up data DSG (2015)... 12 Figure 11 Measuring points HTF/DSG integration circuit... 13 Page 2 of 14
1 Abbreviations DWD Deutscher Wetter Dienst (German Weather Service) φ local Longitude of the location [ ] φ St Standard latitude of the local time zone [ ] Φ Latitude of the location [ ] n Day of the year [-] β Altitude angle [ ] ψ s Azimuth angle [ ] δ Declination [ ] ω Hour angle [ ] ω s Hour angle of sunrise [ ] θ Z Angle of incidence on collector plane [ ] θ K Angle of incidence [ ] ψ K Azimuth of the surface to the south direction [ ] γ Inclination angle [ ] T sun H 0 E 0 E 0 E extr,h E glob,h E glob,h,n E dir,h E difu,h E glob,k E dir,k E difu,k E R Sun Time [h] Daily extra-terrestrial solar radiation [kwh/m²] Solar constant = 1.367 [kw/m²] Solar constant under consideration of the varying distance sun earth [kw/m²] Extra-terrestrial solar radiation on a horizontal area [kw/m²] Terrestrial global radiation on a horizontal area [kw/m²] Cloudiness corrected terrestrial global radiation on a horizontal area [kw/m²] Terrestrial direct radiation on a horizontal area [kw/m²] Terrestrial diffuse radiation on a horizontal area [kw/m²] Terrestrial global radiation in collector plane [kw/m²] Terrestrial direct radiation in collector plane [kw/m²] Terrestrial diffuse radiation in collector plane [kw/m²] Ground reflected radiation [kw/m²] m Airmass [-] T L Linke turbidity factor [-] N Cloudiness factor according to DWD [-] k T Clearness factor according to Orgill and Hollands [-] ρ e ground specific reflection factor ( albedo ) [-] DNI Direct normal irradiation [kw/m²] Q Energy [kwh] C Heat capacity [kwh/k] T startup Start-up duration of HTF / DSG system [min] T start Start temperature of HTF / DSG system [ C] Page 3 of 14
T end P average Operational Temperature HTF / DSG system [ C] Average power output at system integration point HTF / DSG [kw] Page 4 of 14
2 Introduction A steam boiler has been integrated in the steam network of the solar system at Soltigua supplying steam to the brick drying process for time periods with insufficient solar radiation (see Figure 1). Boiler Figure 1 Soltigua system scheme This steam boiler could also be used to heat up the connected very long steam network of the solar collectors what could help to reduce fluctuations in their steam production during system start up. This would also reduce the start-up phase and thereby would lead to a higher system utilization rate of the collector fields. However, in any case it is important, that the expected daily yield of the collector is greater than the energy used by the steam boiler to heat up the steam network. A predictive control system which relies on radiation forecast data and simplified simulation models of the collector fields can be used very effectively (see Figure 2). Figure 2 Principle of simulation based predictive control Page 5 of 14
3 Generation of radiation forecast data Hourly radiation forecast data are sold commercially and would cost for the desired application between 1.000 and 2.000 per year. As the goal is, to develop a cost-efficient and simple system for predictive control, the aim of the work done within InSun was to find an alternative solution. The idea was to rely on freely available weather forecast data and to determine reasonably accurate hourly solar radiation levels from such a system using specific algorithms. 3.1 Logging of cloud cover data Most weather data providers publish hourly weather forecast for at least 24 hours in advance. However, in general they only provide ambient temperature, humidity, wind-speed and cloud coverage on an hourly basis but no solar radiation data (e.g. Accuweather.com, see Figure 3). It is possible with an automated script (Ruby) to access the source code of a website and to log the cloud cover data in freely selectable intervals. This cloud cover data can now be converted into a standard scale according to the German Weather Service DWD (Wesselak & Schabbach, 2009), see Figure 4. Figure 3 Free available hourly weather forecast Weather N E glob,h,n/e glob,h E difu,h,n/e glob,h,n Cloudless, sunny 0 100 % 10 20 % Slightly cloudy 1-3 95 100 % 20 50 % Cloudy 4-6 60 95 % 50 75 % Overcast 7 40 60 % 75 95 % Heavily overcast 8 10 40 % 100 % Figure 4 Cloud cover scale according to DWD From this scale the expected hourly direct and diffuse radiation for the specific location can be calculated taking into account the surface tilt. Page 6 of 14
3.2 Calculation of hourly irradiance To calculate the solar radiation on a (tilted) surface at first the position of the sun has to be determined. Therefore, the local standard time has to be converted to sun time (1) utilizing the time equation (2) (Drück, 2012). T sun = T standard + 4 (φ local φ St ) + E (1) E = 9,81 sin(2 B) 7.53 cosb 1.5 sinb (2) With φ local = Longitude of the location [ ] φ St = Standard latitude of the local time zone [ ] B = (n-81)*0.989 n = day of the year The position of the sun can be described by two angles: a) Altitude angle β: sinβ = cosϕ cosδ cosω + sinϕ sinδ (3) b) Azimuth angle ψ s ( 180 ψ s 180 ): sinψ s = cosδ sinω/cosβ (4) With Φ = Latitude of the location [ ] δ = Declination = 23.45*sin[(284+n)*0,9863] ω = hour angle = 0.25*(T sun - 720); (T sun in minutes) The daily extra-terrestrial solar radiation is calculated by: H 0 = 24 π E 0 (ω s sinδ sinϕ + cosδ cosϕ sinω s ) (5) With ω s = hour angle of sunrise = -tanϕ*tan δ E 0 = Solar constant under consideration of the varying distance sun E 0 = E 0 (1 + 0,033*cos 2 π n 365 ) (6) E 0 = 1.367 kw/m² The average extra-terrestrial solar radiation on a horizontal area for a specific time interval between ω 1 and ω 2 (ω 2 > ω 1 ; [ ]) is calculated by: E extr,h = 12 E π 0 (cosϕ cosδ (sinω 2 sinω 1 ) + 2 π (ω 2 ω 2 ) sinϕ sinδ) (7) 360 Page 7 of 14
The path length of the radiation through the atmosphere is described by airmass. Thereby m = 0 is the intensity of radiation outside the atmosphere and m = 1 would imply a vertical passage through the atmosphere. m = 1 cosθ Z + 0.50572 (6.07995 + 90 θ Z ) 1.6364 (8) With θ Z = Angle of incidence on collector plane [ ] The hourly terrestrial solar radiation is determined under consideration of the predicted ambient temperature and humidity by the following equations. E glob,h = E extr,h exp( T L Ra S 0 m) (9) Ra S 0 m = m 0.9 m+9.4 The Linke turbidity factor T L is determined in coincidense with the ambient temperature and the humidity: very clear, cold air T L = 2; clear warm air T L = 3; warm air with high humidity T L = 4 6; polluted air T L > 6 The influence of the cloudiness on the radiation intensity can be determined with equation (11). The weather conditions are transferred into a rating according to the DWD (German Weather Service, see Figure 4). E glob,h,n = E glob,h (1 a ( N 8 )b ) (11) (10) With a = 0.72 and b = 3.2 The hourly diffuse radiation is determined according to Orgill and Hollands who have examined the following correlation from hourly measured data (Orgill & Hollands, 1977): 1.0 0.249 k T for k T < 0.35 E difu,h = { 1.557 1.84 k E T for 0.35 < k T < 0.75} (12) glob,h,n 0.177 for k T < 0.75 k T = E glob,h,n E extr,h (13) The orientation of the surface, to which the radiation has to be converted, is defined by the inclination angle γ and the azimuth of the surface to the south direction ψ K. To convert the directed radiation to the tilted surface, the angle of incidence θ K must be calculated. cosθ K = cosβ cos(ψ S ψ K ) sinγ + sinβ cosγ (14) θ Z = 90 β (15) Page 8 of 14
E dir,k = E dir,h cosθ K cosθ Z (16) E dir,h = E glob,h,n E difu,h (17) The ground reflected radiation is calculated by E R = E glob,h,n ρ e 1 2 (1 cosθ K) (18) With ρ e as the ground specific reflection factor ( albedo ) Thereby the optimal tilt angle for the 2d tracking system of the Fresnel collector is determined by γ = atan ( cos (ψ K -ψ S ) tanθ Z ) (19) 3.3 Comparison of predicted and measured data The developed methodology was tested to predict the DNI for Gambettola for a certain period in autumn 2015. Afterwards, the predicted solar radiation was compared with the real measured DNI at the installation. For this period weather forecast data of (www.accuweather.com) were used. The comparison of predicted and measured DNI is given in Figure 5. Figure 5 Predictive control: Comparison of measured and predicted yield In this period of 16 days 39.895 kwh of direct normal solar radiation were measured and 37.095 kwh were predicted. This is an average error of only 7%. On an hourly basis the errors are of course much bigger, but the general trend is met sufficiently precise, which makes this cost effective method very interesting for the application in the predictive control system. Page 9 of 14
Yield [kwh/m²] ENER/FP7/296009/InSun Figure 6 shows a comparison between commercially available radiation prediction (meteoblue) and calculated results by the previously described algorithms. Thereby a sufficient match can be obtained. 800 700 600 Gtot predicted (accuweather) [kwh/m²] Gtot predicted (meteoblue) [kwh/m²] 500 400 300 200 100 0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 hour Figure 6 Predictive control: Comparison of commercially predicted and calculated yield Page 10 of 14
Pressure [bar] Temperature [ C] ENER/FP7/296009/InSun 4 Determination of heating up energy demand and predictive control For the calculation of the amount of energy required for the heating up phase two options were considered. The first option is the calculation of the heat capacity of all crucial system components of the steam network. However, this requires very detailed data of all the components installed and would be very time consuming which again would be too costly for a broad application. The second option is the calculation of the thermal capacity from monitoring data. Within this method, the difference between the collector yield and the energy transferred to the process during the time period between system start-up and reaching process temperature is determined (see Figure 7). This allows the calculation of the specific heat capacity [kwh/k] needed to start up the system for different operational conditions. The expected yield of the collector for the actual day can be calculated from a simplified collector model based on efficiency and loss factors and on the predicted DIN distribution over the day. Then the expected collector yield can be compared with the effort to be spent for heating up the steam network by the steam boiler. If the collector yield is not a certain factor larger than the consumed for heating up the steam network the steam network is not pre heated and the collector system is not set in operation for this day. How large this certain factor needs to be selected depends on the electricity consumption of the pumps, controls and motors of the collector field. 14 12 10 8 6 4 2 p_sf_inlet_barg p_steam_drum_barg p_int_2 T_DSG_REC Start-up Start-up 180 160 140 120 100 80 60 40 20 0 Time8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 0 Figure 7 Daily measured data Page 11 of 14
Figure 8, Figure 9 (HTF) and Figure 10 (DSG) show a part of the evaluated data. Within this, this entire start-up s and partially start-up s during operation due to changing irradiation were considered. HTF Date (2014) 01.08. 11.08. 12.08. 13.08. 15.08. 15.08. 11.08. 12.08. Start-up phase [min] 75 84 31 120 63 37 29 33 T start 30 100 171 40 118 32 31 38 T end 227 221 220 194 200 115 100 114 T delta 197 121 49 154 82 83 69 76 P average [kw] 135 128 164 64 92 ERdirect dir,k [kw] average [W/m²] 907 642 747 603 833 768 486 546 Q [kwh] 311 201 79 383 274 201 99 127 C [kwh/k] 1,6 1,7 1,6 2,5 3,3 2,4 1,4 1,7 Figure 8 System start up data HTF (2014) HTF Date (2015) 22.06. 23.06. 25.06. 26.06. 28.06. 29.06. Start-up phase [min] 88 113 102 96 113 118 T start 97 40 29 117 26 40 T end 222 212 218 224 221 216 T delta 125 172 189 107 195 176 P average [kw] 154 119 113 160 124 118 ERdirect dir,k [kw] average [W/m²] 781 597 844 811 760 602 Q [kwh] 259 252 415 293 373 269 C [kwh/k] 2,1 1,5 2,2 2,7 1,9 1,5 Figure 9 System start up data HTF (2015) DSG Date (2015) 01.06. 12.06. 22.06. 25.06. 26.06. 27.06. 28.06. Start-up phase [min] 70 75 66 63 75 75 75 T start 74 90 73 77 83 81 81 T end 175 171 174 174 174 174 174 T delta 101 81 101 97 91 93 93 P average [kw] 168 209 160 210 211 187 194 ERdirect dir,k [kw] average [W/m²] 767 682 780 855 802 737 780 Q [kwh] 335 245 333 312 331 313 336 C [kwh/k] 3,32 3,02 3,30 3,22 3,64 3,36 3,61 Figure 10 System start-up data DSG (2015) Thereby, the amount of Energy Q is determined by Q = (E dir,k eta loop A col P 1000 average) T startup 60 3600 And the heat capacity C is determined by C = Q T end T start (20) (21) Page 12 of 14
T end, T start and E dir,k are available as measured data. Eta loop was determined in steady conditions separately for the DSG and the HTF part. In August 2014, eta loop,htf was 37.5% and eta loop,dsg was 35.6%. In August 2015, eta loop,htf was 40.2% and eta loop,dsg was 36.9%. Regarding the DSG system, P average is the thermal power output of the DSG field towards the integration steam circuit (see sensors with subscript INT2 in the layout, Figure 11). The thermal power generated by the steam drum (integration loop side) is calculated from mass flow rate and temperature measurements (saturated steam is assumed at the output of the steam drum, latent heat is properly accounted). Regarding the HTF system, P average was determined as the thermal power output of the HTF field towards the integration steam circuit (see sensors with subscript INT1 in the layout, Figure 11) The thermal power generated by the oil/steam heat exchanger (steam side) is also calculated from mass flow rate and temperature measurements (saturated steam is assumed at the output of the exchanger). Figure 11 Measuring points HTF/DSG integration circuit The average specific amount of energy consumed during start-up is 3.4 kwh/k for DSG and 2.1 kwh/k for HTF. Based on this values and on the measured system temperature, the amount of energy, that is needed to the start up the system, can be determined for each condition. The expected yield in the predicted period of time must now be greater than this value under consideration of the system efficiency. The predicted time span is thereby limited to a specific time span because otherwise the downtime losses outweigh. Page 13 of 14
5 Summary In this deliverable the possibilities for an easy predictive control method with no follow-up costs have been demonstrated. Thereby a reasonably accurate yield prediction could be achieved. Higher prediction accuracy can be achieved through further adjustments when more operational data is available and the impact of the prediction model can be assessed. The developed system is now ready for its implementation in the control system. Due to the simplicity of the tool, no specific software licences are required. However, internet access is needed to get access to the weather forecast data. Parts of this prediction model are suitable for other applications like PV-Yield predictions within demand side management applications. 6 References Drück, H. (2012). Solarthermie (Teil 1). Stuttgart: Institut für Thermodynamik und Wärmetechnik (ITW). Orgill, J., & Hollands, K. G. (1977). Correlation equation for hourly diffuse radiation on a horizontal surface. Solar Energy, Vol. 19, 357-359. Wesselak, V., & Schabbach, T. (2009). Regenerative Energietechnik. Berlin: Springer Science+Business Media S.A. Page 14 of 14