Réunion du Groupe de travail sur les microréseaux Tuesday, November 15, 216 Operating Reserve Quantification Considering System Uncertainties and Day-aead Optimal Dispatcing of a Microgrid wit Active PV Generators Xingyu YAN (xingyu.yan@ec-lille.fr) Supervisor: Bruno FRANCOIS & Daker ABBES L2EP, Ecole Centrale Lille, France
Contents 2/2 Part I: Background: Problem, Objective and Metods Part II: Operating Reserve (OR) Quantification to Cover Uncertainty 1. Power Forecasting wit Artificial Neural Networks (ANN) 2. Net Demand Uncertainty Analysis 3. Power Reserve Quantification Part III: Operating Power Reserve Dispatcing Strategies Part IV: Day-aead Unit Commitment Problem wit Dynamic Programming 1. Non-linear Constraints 2. Unit Commitment Problem wit Dynamic Programming 3. Optimization Strategies 4. Application: Case Study and Simulation Results Part V: A User-friendly EMS and Operational Planning Supervisor Part VI: Conclusions
I. Background: Problem, Objective and Metods I.1 Problem: Uncertainty and Power Reserve 3/2 Uncertainty: Intermittent renewable energy sources (RES) power is difficult to predict. To cover te risk: Additional Operating Reserve (OR) is needed. Massive RES increases te uncertainty in power system and OR is mandatory to maintain te system security level. Today te consumption/production balancing and OR provision are performed by conventional generators. Constant Reserve "Uncertain Variable Reserve" Pref Pref Pref is given by te system operators (power dispatc) Problem: ow to precisely quantify te OR and locate it into te generators, witout losing te system security level.
I. Background: Problem, Objective and Metods I.2 General Organization of Energy Management System (EMS) Microgrid supervision can be analyzed and classified in different timing scales and functions. Day Hour Second Millisecond 4/2 Electricity Market Load Management Frequency and Voltage Control Fast Dynamic Storage Availability Load Prediction RES Power Prediction Operational Planning Long-term Energy Management Energy Storage Availability Energy Adjustment Medium-term RES Power Capability Sort-term Power Management
I. Background: Problem, Objective and Metods I.3 Objectives and Metods 5/2 Predictive Analysis for Uncertainty: PV power and load forecasting Operating Reserve Quantification: Loss of load probability (LOLP) OR Dispatcing Strategies on Generators Day-aead Optimization Planning: Unit commitment problem wit dynamic programming A User-friendly EMS and Operational Planning Supervisor Microgrid wit PV Active Generator (AG) PV Active Generator (PV AG): PV panels combined wit a storage system to provide ancillary services.
II. Operating Reserve (OR) Quantification to Cover Uncertainty II.1 Data Predictive Analysis and Forecasting Data management: Collect, Mining, and Predictive Analysis PV Power and Load Forecasting wit ANN 6/2 PV power forecasting Wit Artificial Neural Networks (ANN) Load forecasting wit ANN
II. Operating Reserve (OR) Quantification to Cover Uncertainty II.2 Net Demand () Uncertainty Analysis Te real is composed of te forecasted and an error: ~ = N D 7/2 First Metod: Day-aead Net Demand Errors Forecast; forecasting errors Second Metod: Calculation from te PV Power and te Load Forecast Errors Estimation. D+1 D Forecasted Load Database Sensed Load Database D+1 D D+1 D Forecasted PV Database Sensed PV Database D+1 D ~ L 24 1 ~,..., L + _ L L 24,..., 1 L Prediction Error Forecast ANN, L 1 24..., Last 24 prediction errors ~,..., ~ L L 24 Mean average and Standard deviation ~ P v ~ ~,..., Pv 24 1, _ + PV PV 24,..., 1 Prediction Error Forecast ANN ~ PV..., PV 24 Pv Mean average and Standard deviation, Pv 1 24..., L L, ( ) 2 PV, ( PV ) 2 : Mean value : Standard deviation ( ) 2 L L ( ) PV 2 ( PV ) 2 Equations (8) and (9), ( ) 2
Net Demand Uncertainty [kw] Load Demand Uncertainty [kw] PV uncertainty [kw] II. Operating Reserve (OR) Quantification to Cover Uncertainty II.3 Uncertainty Assessment and Power Reserve Quantification 8/2 uncertainty assessment for eac ½ our Net Demand Forecast ~ N D x: Desired probability ~,..., 1 24 x F( B, Upper bound (B up ) ) Normal Probability Density Function (F(B µ,σ )) Normal Inverse Cumulative Distribution Function (F -1 ) 1 + + 2 B _ + e Lower bound (B lower ) ( 2( 2 ) 2 ) ~ N ~ N B F -1 ( x, ) B : F( B, ) d pdf D _ max D _ min x 2 15 1 5 9% probability 8% probability 7% probability 6% probability Forecasted PV 4 8 12 16 2 24 Time [ours] 12 1 8 6 9% probability 4 8% probability 7% probability 2 6% probability Forecasted Load 1 4 8 12 Time [ours] 16 2 24 8 6 9% probability 4 8% probability 7% probability 2 6% probability Net Forecasted 4 8 12 16 2 24 Time [ours]
Power Reserve [kw] PR Probability Forecast Load Forecast PV II. Operating Reserve (OR) Quantification to Cover Uncertainty II.4 Operating Power Reserve Quantification Power reserve quantification LOLP represents te probability tat load exceeds PV power. LOLP.1.8.6 = prob( L P ) PR pdf ( )d LOLP= x % 6 3 9/2 Day-aead forecasted PV, load and power reserve 8 1 13 16 19 22 1 4 7 12.4.2 Power surplus Accepted risk x % PR Power deficit 6 8 2 1 13 16 19 22 1 4 7 Wit 1 % of LOLP 1 15 1 5 25 2 15 1 Time [Hour] 5 1 2 3 4 5 6 7 8 9 1 x % of LOLP 8 1 13 16 19 22 1 4 7 Time (alf our) Targets Day-aead dynamic power reserve quantification.
OR PR OR OR (kw PR OR OR More details can be found ere: X. Yan, D. Abbes, B. Francois, and Hassan Bevrani Day-aead Optimal Operational and Reserve Power Dispatcing in a PV-based Urban Microgrid, EPE 216, ECCE Europe, Karlsrue/ Germany. III. Operating Power 1Reserve Dispatcing III. Strategies: OR Provision by MGT and PV AG OR in MGT3 1/2 8 1 13 16 19 22 1 4 7 2 Strategy 1: OR on tree MGTs only 15 2 1 (a) (b) H1 H2 (a) H1 OR in MGT1 OR in MGT2 OR OR in in MGT3 OR in MGT3 MGT1 OR in PV AG OR in MGT2 5 1 OR in MGT3 8 1 13 16 19 22 1 4 7 2 Time (alf our) (b) H2 OR in MGT1 158 1 13 16 19 22 1 OR 4 in MGT2 7 2 OR in MGT3 1 (b) H2 OR in MGT1 15 OR in PV AG OR in MGT2 5 OR in MGT3 1 OR in PV AG 5 8 1 13 16 19 22 1 4 7 Time (alf our) 8 1 13 16 19 22 1 4 7 Eac PV AGs equally Time (alf our) contributed te OR Strategy 2: OR on tree MGTs and tirteen PV AGs Tree MGTs
24 our aead forecasting and power reserve qualify IV. Day-aead Unit Commitment Problem wit Dynamic Programming IV.1 Sceme of Day-aead Optimal Power Reserve Planning (t) P ~ v L ~ OR Constraints Power balance MGT efficiency Maximization of RE Battery energy Ф(t) (t) Optimization Dynamic programming x(t) u(t) Day-aead optimal operation planning Electrical system modeling PV based AG Loads MGT Operational and Reserve Power Dispatcing Tecnological features CO 2 emissions Economic cost 11/2 J Criteria Objective function Operational planning of MGT Focus on te design of te MCEMS under particular constraints. Uint commitment problem (UCP) wit dynamic programming (DP) is developed in order to reduce te economic cost and CO 2 equivalent emissions.
CO2 equivalent (kg) Fuel consumption IV. Day-aead Unit Commitment Problem wit Dynamic Programming IV.2 Non-linear Constraints 12/2 Security: Reserve power assessment wit x % of LOLP; Power balancing: N ( t) = PL ( t) Pres( t) PAG _ n( t) ( i( t) PMGT _ M n1 i1 i ( t)) Maximization of renewable energy usage: considering te battery capacity limitation (more PV power, larger battery storage!) MGT corresponding inequality constraint: % P ( t) P ( t) 1% P ( t) 5 M _ max_ i M _ i M _ max_ i Power output (%) Partial load ratio (%)
P AG P MGT3 P MGT2 P MGT1 IV. Day-aead Unit Commitment Problem wit Dynamic Programming IV.3 Unit Commitment Problem (UCP) wit Dynamic Programming 13/2 UCP: Optimal Operational of a cluster of MGTs (since te PV power is prior source) x(t) = P MGT _ 1 ( t), P MGT _ 2 u(t) = 1( t), 2( t),..., i( t) Optimization Objectives: ( t),..., P MGT _ i ( t) 1. Economic criteria: minimize total fuel cost; 2. Environmental criteria: minimize CO2 emission; 3. Best compromise criteria: make a compromise. 3 15 8 1 13 16 19 22 1 4 7 3 15 8 1 13 16 19 22 1 4 7 6 3 8 1 13 16 19 22 1 4 7 5 25 8 1 13 16 19 22 1 4 7 Time (alf our) Dynamic Programming (DP) Systematically evaluates a large number of possible decision in a multi-step problem considering te "transition costs". Start t=t Calculate F(T, x(t), u(t)) wit x(t) and u(t) to minimize J(T). t=t-1 Optimization: calculate te minimum J(t) wit vectors x(t) and u(t), using recursive equation F(t, x(t), u(t)). F 1 (t) F 2 (t) F 3 (t) F 4 (t). F 1 (T) F 2 (T) F 3 (T) F 4 (T). F 1 (T) F 2 (T) F 3 (T) F 4 (T). No t=1 Stop Yes F 1 (1) F 1 (2) F 1 (T-1) F 1 (T) F 2 (1) F 2 (2) F 2 (T-1) F 2 (T) F 3 (1) F 3 (2) F 3 (T-1) F 3 (T) F 4 (1) F 4 (2) F 4 (T-1) F 4 (T)....
IV. Day-aead Unit Commitment Problem wit Dynamic Programming IV.4 Case Study and Simulation Results (1) 14/2 In tis case: rated load (11 kw), rated PV power (55 kw) and te OR (wit 1 % of LOLP) coming from te net demand uncertainty assessment. Scenario Optimized criteria Cost ( ) Pollution (kg) OR on AG (%) E battery _ Max Witout PV Power Strategy 1: Strategy 2: Witout PV power None 219 1392 Environmental 212 1196 Economic 21 1263 None 183 1156 8.2 Environmental 181 167 8.2 Economic 178 112 8.2 None 182 198 4 54.1 Environmental 179 991 4 54.1 Economic 177 161 4 54.1 Power reserve dispatcing, one day aead Strategy 1 4% of OR is on PV AG! Strategy 2 39% 14% 47% MGT1 MGT2 MGT3 AG= 1% 19% 71% MGT1 MGT2 MGT3 AG= 4% 18% 3% 12% MGT1 MGT2 MGT3 AG
P AG P MGT3 MGT Power Ratio P MGT2 P MGT1 LOLP (%) P AG LOLP (%) P MGT3 P MGT2 P MGT1 MGT Power Ratio IV. Day-aead Unit Commitment Problem wit Dynamic Programming IV.4 Results (2): MGTs Load Ratio and System Security Strategy 1: OR on MGTs 3 2 1 3 8 1 13 16 19 22 1 4 7 2 1 8 5 1 13 16 19 22 1 4 7 5 8 1 13 16 19 22 1 4 7.5 1.5 1,8,6,4,2 Strategy 1 None Environmental Economic MGT1 MGT2 MGT3 MGT Power Ratio wit Strategy 1 15/2 8 1 13 16 19 22 1 4 7 T ime (alf our) Strategy 2: OR on MGTs and PV AG 3 2 1 8 3 2 1 13 16 19 22 1 4 7 1 8 1 13 16 19 22 1 4 7 5 5 8 1 13 16 19 22 1 4 7 8 1 13 16 19 22 1 4 7 T ime (alf our).8 7 2 1 13 16 19 22 1 4 7 Strategy 2 1 7 1 13 16 19 22 1 4 7 Time (alf our) 1,8,6,4,2 None Environmental Economic MGT1 MGT2 MGT3 MGT Power Ratio wit Strategy 2
P AG E Battery P MGT3 P MGT2 P Battery P MGT1 P AG P AG E Battery P MGT3 P MGT2 P Battery P MGT1 P AG IV. Day-aead Unit Commitment Problem wit Dynamic Programming IV.4 Results (3): Battery State of Carge Strategy 1: OR on MGTs 3 2 1 3 8 1 13 16 19 22 1 4 7 2 1 8 5 1 13 16 19 22 1 4 7 5 8 1 13 16 19 22 1 4 7 5 16/2 8 1 13 16 19 22 1 4 7 2-2 8 1 13 16 19 22 1 4 7 1 5 8 1 13 16 19 22 1 4 7 T ime (alf our) Strategy 2: OR on MGTs and PV AG 3 2 1 8 3 2 1 13 16 19 22 1 4 7 1 8 1 13 16 19 22 1 4 7 5 5 8 1 13 16 19 22 1 4 7 8 1 13 16 19 22 1 4 7 T ime (alf our) 8 1 13 16 19 22 1 4 7 Time (alf our) 5 25 8 1 13 16 19 22 1 4 7 4-4 8 1 13 16 19 22 1 4 7 6 4 2 8.2 kw 54.1 kw 8 1 13 16 19 22 1 4 7 Time (alf our)
V. A User-friendly EMS and Operational Planning Supervisor V.1 General Framework 17/2 Objective: to conceptualize te overall system operation and to provide a complete set of user-friendly GUI to properly model and study te details of PV AG, load demand, and MGTs. Data Collection and System Uncertainty Analysis Big Data Collecting Predictive Analysis Forecast wit ANN Mining Training. Validation Test Adopted Metods for te Urban Microgrid Simulator Uncertainties Assessment for OR Quantification PV Uncertainty Load Uncertaitny Uncertainty Assessment LOLP OR Quantify Constrains OR Dispatcing for UC Problem wit DP Unit Commitment Dynamic Programming Dispatcing Strategies Environmental, Economic, Best Compromise Optimization LOLP Ceriteria MGT & PV AG Scenarios H&L
V. A User-friendly EMS and Operational Planning Supervisor V.2 Microgrid Simulator Frame Design 18/2 Main Interfaces Data Collect and System Uncertainty Analysis Individual Modules Historical Data Collect for ANN Training Day-aead Data Download: Weater Information, Load, and PV Power Data PV Power and Load Demand Forecast by Using Well Trained ANN Historical Data Collect Predictive Analysis: to identify correlated patterns and parameters Data Collect and System Uncertainty Analysis Forecast Model Building: Tree-layer BP ANN Real Time Data Obtain: Weater Data Download Load Data Download PV Power Data Download System Uncertainties Assessment and OR Power Quantification Operational and OR Dispatcing PV Power Uncertainty Load Demand Uncertainty Net Demand Uncertainty OR Quantification Dispatcing Strategies PV AGs and MGTs Power References ANN Training: PV Power Forecast ANN Load Forecast ANN Training Results Display Area Forecasting Wit ANN: PV Power Forecasting Load Forecasting Forecasting Results Display Area Operational and OR Dispatcing System Uncertainties Assessment for OR Quantification Initialization Forecasted PV Forecasted Load OR PV AGs MGTs OR Dispatcing Strategies: Hig PV Power Scenario Low PV Power Scenario OR in MGTs Only OR in MGTs and PV AGs DP for UCP Optimization Criterias: None Economic Environmental Best Compromise Dispatcing Results Display Area Optimization Results Display Area Uncertainty Assessment PV Power Uncertainty Load Uncertainty Uncertainty Assessment: First Metod Second Metod Uncertainties Display Area: PV Power, Load, and (First and second metod) Probabilistic OR Assessment: LOLP, EENS Risk-constrained OR Calculation OR Computer: x % of LOLP Hourly Quantified OR Display
V. A User-friendly EMS and Operational Planning Supervisor V.3 Microgrid Simulator Interface Design wit Matlab GUI 19/2 Demonstration
VI. Conclusion 2/2 Conclusions PV power variability and load demand variability are analyzed. Te ANN algorisms are developed for te PV power and te load forecast. A probabilistic metod for te OR calculation based on two different kind of forecasted uncertainty assessment metods is proposed. Te dynamic joint operational and OR dispatcing strategies are developed. Day-aead optimal operational and OR planning wit DP is proposed by considering different constraints and different optimization strategies. A User-friendly EMS and Operational Planning Supervisor is developed. Prospects Big data for distributed RES uncertainty analysis and a better forecasting results Optimization metod to improve te battery efficiency Build a global EMS to incorporate te predicted uncertainty ranges into te sceduling, load following, and into te regulation processes.
Related Publications 1. X. Yan, B. Francois, and D. Abbes, Solar radiation forecasting using artificial neural network for local power reserve, in Electrical Sciences and Tecnologies in Magreb (CISTEM), 214 International Conference, pp. 1-6. 2. X. Yan, B. Francois, and D. Abbes, Operating power reserve quantification troug PV generation uncertainty analysis of a microgrid, in PowerTec, 215 IEEE Eindoven, 215, pp. 1-6. 3. X. Yan, D. Abbes, B. Francois, and Hassan Bevrani Day-aead Optimal Operational and Reserve Power Dispatcing in a PV-based Urban Microgrid, EPE 216, ECCE Europe, Karlsrue/ Germany. 4. X. Yan, B. Francois, and D. Abbes, Uncertainty Analysis for Power Reserve Quantification in an Urban Microgrid Including PV Generators, Elsevier, Renewable Energy, under review. 5. X. Yan, B. Francois, and D. Abbes, Operating Reserve Quantification and Day-aead Optimal Dispatcing of a Microgrid wit Active PV Generators, IET, under review. Tank you for your attention!