Grid-connected Advanced Power Electronic Systems Real-time Volt/Var Optimization Scheme for Distribution Systems with PV Integration 02-15-2017 Presenter Name: Yan Chen (On behalf of Dr. Benigni)
Outline Impacts of PV Integration on Distribution Grids Solution: PV Inverter Control to Sustain High Quality of Service A Top-level Day-ahead Control that Optimizes Voltage Deviations and Power Losses A Fast on-line Control that Compensates for PV Generation and Load Variability Communication Network Aware Distributed Voltage Control Algorithms Conclusion 2 2
PV Impact On Distribution Grids Change in feeder voltage profiles, including voltage rise and unbalance Deteriorated power quality: PV-DG intermittency may lead to rapid fluctuations in bus voltage magnitudes Frequent operation of voltage-control and regulation devices, such as on load tap changers (OLTCs), line voltage regulators (VRs), and shunt capacitor banks (SCBs) Change in electric losses, where relatively large reverse power flow may increase power losses 1.12 1.1 Optimal scheduling for Tap1 1.08 1.06 1.04 1.02 1 0.98 0.96 0.94 0:00 8:00 16:00 24:00 Time 3 3
Day-ahead Coordinated Optimal Control Objectives: Determine how to optimally control the related electric elements to minimize the voltage fluctuation and power losses with constraints on the OLTC and SC operations. PV inverter On-load tap changer Shunt capacitor bank PV Inverter VAR control: When the PV generation is not at the maximum level, the unused converter capability can be used for reactive compensation. ( ) = ( ) ( ) ( ) 4 4
Optimal Control Problem Decision variables: Reactive power of PV inverter (continuous variables) OLTC tap position (discrete variables) SC switch state (Boolean variables) Objective function: Total voltage deviation Total power losses Constraints: Reactive power limit of PV inverter: Limit of node voltage magnitudes (ANSI C 84.1): Limit of tap positions of OLTC: t t t x [ Q, Tap, SC ], t 1, 2,... T pv T Nnode Nbr t i t 1 i 1 j 1 t min F ( w VD (1 w) PL ) S ( P ) Q S ( P ) Tap Tap Tap 2 t 2 t 2 t 2 pv pv pv pv pv L t U V V V L t U i Limit of the tap operations of OLTC within a day: TTC TTC Limit of the switch operations of shunt capacitor within a day: TSC j max TSC max 5 5
Overall Process Inputs: Forecasted PV Generation Forecasted Load Demand Distribution Network Information Optimization Process: Pattern Search Algorithm Genetic Algorithm Treated as a black-box model Outputs: Reactive power of PV inverters Tap position of OLTCs Switch State of Shunt Capacitors 6 6
Case Study IEEE 34 Node Test Feeder Controlled Devices Location Decision variables PV inverter Node 34 On-load tap changer Node7-8, ±10 taps with 1% voltage regulation per tap. Shunt capacitor Node 27, could be 0 (disconnected) or 1 (connected) 7 7
Results and Discussions Constraint function TTC=23 TSC=16 TTC=16 TSC=16 TTC=12 TSC=12 TTC=8 TSC=8 TTC=4 TSC=4 Objection function 49.32 52.70 56.51 75.69 85.71 8 8
Discussion The performance of the day-ahead control method is affected by the forecast errors. Solar PV output: errors caused by actual irradiance Cloud cover Aerosols and other atmospheric constituents Temperature Load demand: Temperature Random (stochastic) customer behavior Feeder outages 9 9
Real-time Optimization We propose an online optimal reactive power control strategy to keep the total voltage deviations and power losses to a minimum regardless of unpredicted changes. In order to reduce the additional wear and tear on the physical voltage control devices, the tap position of the OLTC and the switch state of the SC are controlled according to the day-ahead optimal control scheme. The reactive power of the PV is decided by the real-time system status. Real-time system status Real-time control of Qpv Day D Day (D+1) t Day-ahead scheduling for OLTC, SC, and Qpv, PV output and load demand forecast 10 10
Control Structure Measurements (Pinj, Qinj, V, I) Measurements (Pinj, Qinj, V, I) Control Center 11 11
Control Structure Measurements (Pinj, Qinj, V, I) Measurements (Pinj, Qinj, V, I) State Estimation Optimization Algorithm 12 12
Control Structure Controller Controller Controller Controller Controller Controller Control Signal Control Signal (Qpv) 13 13
Control Structure Measurements (Pinj, Qinj, V, I) Measurements (Pinj, Qinj, V, I) Control Center 14 14
Distributed Control Lab Instructure 15 15
Controller Board Position ODROID-U3+ Key Features Upper layer Low-cost, powerful computer Ease of programming Network capable ARM Quad-core 1.7 GHz CPU and 2GB RAM. Xubuntu 13.10 Operation System Position U3 I/O Shield Key Features Middle layer 36 IO ports of GPIO/PWM/ADC Position OPAL-U3-Shield Key Features Bottom layer Contains level shift, amplification, and filter circuitry for different signal requirements between OPAL (-10V-10V) and U3 I/O Shield. Allows access to all IO ports on the U3 I/O Shield 16 16
Controller Board Position ODROID-U3+ Key Features Upper layer Low-cost, powerful computer Ease of programming Network capable ARM Quad-core 1.7 GHz CPU and 2GB RAM. Xubuntu 13.10 Operation System Position U3 I/O Shield Key Features Middle layer 36 IO ports of GPIO/PWM/ADC Position OPAL-U3-Shield Key Features Bottom layer Contains level shift, amplification, and filter circuitry for different signal requirements between OPAL (-10V-10V) and U3 I/O Shield. Allows access to all IO ports on the U3 I/O Shield 17 17
BA14 crio-9035 Embedded Controller Xilinx FPGA for rapid signal processing 1.33 GHz Dual-Core allows wide range of computations Digital and analog I/O modules Analog I/O: 12-bit resolution bidirectional at 20 ks/s Digital I/O: 8 bidirectional channels at 10 MHz GPS module enables synchronous signal measurement 18 18
Slide 18 BA14 add a picture that show the full rack and add some detail on the IO modules BENIGNI, ANDREA, 2/10/2017
Network Emulator: Netropy N91 Test the effect of WAN: Bandwidth Latency and jitter Loss Other impairment Congestion Corruption Queuing and Prioritization Applications: Throughput Responsiveness Quality 19 19
Real-time Simulation of Distribution Grids IEEE 34 Node Test Feeder 4 SSN nodes, 5 subsystems Ts = 50us IEEE 123 Node Test Feeder 7 SSN nodes, 8 subsystems Ts = 50us 20 20
Model Components RT-LAB overview ARTEMis State-Space Nodal (SSN) The SSN algorithm creates virtual state-space partitions of the network that are simultaneously solved using a nodal method at the partition points of connection. The partitions can be solved in parallel on different cores of a PC without delays. 21 21
Test Setup HIL structure 22 22
Case Study 23 23
Case Study Measurements (Pinj, Qinj, V, I) Measurements (Pinj, Qinj, V, I) Qpv State Estimation Optimization Algorithm 24 24
Case Study To evaluate the ANN approach, 17,520 samplings are generated from one year s historical record, and 15%, 15%, 10%, and 5% white noise is added to the domestic load, commercial load, industrial load and street light load, respectively. The comparison of the forecasted results (in red curves) and the real measurements (in blue curves) of the PV output and the load demand. 25 25
Real-time Qpv Control Results The real-time reactive power control is applied to correct the forecast errors of PV output and load demand. The overall objective function value is decreased from 102.30 to 89.82. The voltage magnitudes are more centralized to 1 PU. The power quality improvement is apparent when there is dramatic uncertainty of the PV output. 26 26
Distributed Control Algorithm A modified IEEE-123 power distribution system is simulated on OPAL-RT in real time 10 CompactRIO embedded controllers connected at various points in the grid measure voltage phasors to determine reactive power flow By communicating with one another, these controllers attempt to minimize transmission losses by injecting and absorbing reactive power The control algorithm adapts to the network conditions generated by the network emulator; using either gossip-like or F-DORPF algorithm 27 27
Distributed Control Algorithm 28 28
Conclusions The increasing penetration of distributed and renewable energy resources introduces challenges to the distribution systems operation and control Real-time simulation (of power and communication networks) and Hardware In the Loop simulation are fundamental tools for the design and testing of innovative control solutions 29 29
Thanks for your attention Questions? Dr. Andrea Benigni Department of Electrical Engineering benignia@cec.sc.edu Yan Chen, Ph.D. student Department of Electrical Engineering yc2@email.sc.edu 30 30