Combining Local and Centralized Voltage
Control Schemes in Active Distribution Networks
Hamid SOLEIMANI BIDGOLI
Thierry VAN CUTSEM
Dept. of Electrical Engineering and Computer Science (Montefiore Institute), University of Li `ege, Belgium
Voltage control architectures
Different control schemes can be contemplated in a distribution system taking into account practical needs, technical limitations of the Distributed Generation Units (DGU), and regulatory policies.
Local Control (LC)
• DGUs may quickly adjust their power outputs based on volt-age (and active power) mea-sured at their terminals
• control embedded in the equipment • no communication infrastruc-ture needed. LC1 LCN DGU1 DGUN Distribution Network Centralized Control (CC)
• periodically gathers measure-ments and sends set-point corrections to the DGUs
• shares the corrective efforts over multiple DGUs
• communication infrastructure
needed. DGU1 DGUN
CC
Distribution Network
Combined Local and Centralized
• fast reactions from LCs after any disturbance
• followed by slower, but coordinated adjustment of DGUs at upper level, to refine the local corrections • enhanced control possibilities
LC1 LCM LCM+1 LCN CC
DGU1 DGUM DGUM+1 DGUN Distribution
Network
Hybrid control
The deployment of the CC and its communication in-frastructure may not be feasible or affordable over all DGUs, for instance on small DGUs or DGUs with older technology.
In the hybrid control scheme, some DGUs are under LC mode only, while others are in CLC mode, accounting for the effect of the former.
LC1 LCM
CC
DGU1 DGUM DGUM+1 DGUN Distribution
Network
LCM+1 LCN
Lower level : local, fast control
The reactive power output of a DGU varies according to the piecewise linear VQ characteristic shown below.
Qg V Qmin Qmax Vmax2loc Vmin2loc Vmax1loc Vmin1loc DGU Pg, V Qg
Vmin2loc , Vmin1loc , Vmax1loc , Vmax2loc : limits defined in LC
As long as the measured terminal voltage lies the dead-band [Vmin1loc , Vmax1loc ], the produced reactive power Qg is kept at zero to minimize the DGU internal losses.
Correction from upper level
The cumulated correction Qcor received from the upper level shifts the VQ characteristic of the DGU under con-trol.
only high voltage part of VQ characteristic shown here, for clarity Qg V Qmin
Vmax1loc +Vcor
Vmaxcnt
Qcor
Vcor Network
characteristic
Upper level : slow, centralized control
A Model Predictive Control scheme is used, which in-volves solving at each discrete time, the multi-step con-strained optimization problem detailed below.
min ∆Qg,ε Nc−1 X i=0 k∆Qg(k + i)k2W + kεk2S where (i = 1, . . . , Nc) : ∆Qg(k + i) = Qg(k + i) − Qmg (k) ∆Qg(k + i) = SQ ∆Qcor(k + i) ∆Qg : DGU reactive power change
ε = [ε1, ε2]T : slack variables to relax the inequality con-straints in case of infeasibility
Qmg : last measured values of DGU reactive power W , S : weighting and penalizing matrices
∆Qcor : DGU reactive power correction by controller SQ : sensitivity matrix of DGU reactive power to re-quested reactive power correction
for i = 1, . . . , Np :
V (k + i | k) = V (k | k) + SV ∆Qg(k + i − 1)
V (k + i|k) : bus voltages predicted at time k + i given the measurements at time k
V (k | k) : last measured values of voltages
SV : sensitivity matrix of voltages to reactive power changes for i = 1, . . . , Np : (−ε1 + Vmincnt) 1 ≤ V (k + i | k) ≤ (Vmaxcnt + ε2) 1 for i = 0, . . . , Nc − 1 : Qming (k) ≤ Qg(k + i) ≤ Qmaxg (k) ∆Qming (k + i) ≤ ∆Qg(k + i) ≤ ∆Qmaxg (k + i)
Qming , Qmaxg ,∆Qming and ∆Qmaxg : lower and upper limits on the DGU outputs and on their rates of change
1 : unit vector
Vmincnt , Vmaxcnt : voltage limits considered at upper level Non-updated sensitivity matrices
• to minimize information exchanges, the optimization is performed under the assumption that all DGUs operate on the sloping part of their VQ characteristics
• however, those DGUs operating in the dead-band of their VQ characteristic do not respond to a correction ∆Qcor with the expected additional reactive power
• this is corrected by the closed-loop model predictive control scheme (other DGUs compensate for the non responding ones)
• by not updating the sensitivities, the voltage correction takes some more time.
On the control of transformer ratios
• Increasing the number of tap changes reduces the life-time of the transformer Load Tap Changer (LTC)
• hence, the latter has not been considered (assuming there is enough controllability from DGUs)
• but the formulation can be easily extended to include the LTC voltage set-point of as a control variable with proper associated “cost”.
Simulation results using hybrid scheme
G G G G G G G G G G G G G G Ext. Grid 1000 1100 1114 1112 1113 1111 1110 1107 1108 1109 1104 1105 1106 1101 1102 1103 1115 1117 1122 1123 1124 1118 1119 1120 1121 1125 1151 1152 1153 1167 1154 1155 1169 1156 1157 1158 1159 1168 1170 1161 1162 1163 1164 1165 1166 1175 1174 1172 1160 1127 1128 1129 1134 1130 1131 1132 1136 1149 1144 1143 1142 1148 1147 1140 1133 1141 1135 1139 1138 1146 1137 1171 G Voltage measurement Power flow measurement G G 1116 G 1126 G G G 1150 1145 G G
DGU under local control only with P, Q, V measurements G
1173
DGU under both local and centralized control
with P, Q, V measurements
75-bus, 11-kV distri-bution network
38 static and 15 dy-namic loads Upper level : Nc = Np = 3 measurements re-ceived every 10 s : (P, Q, V ) on 11-kV side of transformer, (P, Q, V ) at the termi-nal of each DGU, V at 4 non-DGU buses. reactive power cor-rections sent every 10 s
22 DGUs : 3.3-MVA doubly fed induction generators driven by wind turbines and 3-MVA synchronous geners. 10 DGUs under local control only;
12 DGUs under both local and centralized control.
Correction of voltages after a 0.05 pu voltage drop taking place in the external grid, at t = 10 s.
Under the effect of local control, DGUs with terminal volt-age outside the dead-band inject reactive power right af-ter the disturbance. The voltages are rapidly but partly corrected, leading to fewer buses in low voltage situation.
0.96 0.97 0.98 0.99 1 0 10 20 30 40 50
Lower voltage limit
1152 1145
Voltage, pu
Time, s
At t = 20 s, the upper level, coordinated controller has sensed the unsatisfactory voltages and applies the cor-rective actions ∆Qcor on the 12 DGUs under its control. The remaining voltage violation is cleared in two steps.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 10 20 30 40 50 1152 1152 1155 1162
Reactive power, Mvar
Time, s
The 10 DGUs under local control only participate in the initial correction of the voltages. They are reset to their initial reactive power after the upper level has acted.
Related publications and acknowledgement
• H. Soleimani Bidgoli and T. Van Cutsem. “Voltage Profile Correc-tion in DistribuCorrec-tion Grids Combining Single- and Two-level Con-trollers“, to be presented at the IEEE PES PowerTech Conferencc, Manchester, June 2017
• H. Soleimani Bidgoli and T. Van Cutsem. “Combined Local and Centralized Voltage Control in Active Distribution Networks“, sub-mitted for publication in IEEE Trans. on Power Systems, 2017 (manuscript under second review)
Work supported by Public Service of Wallonia, Dept. of Energy and Sustainable Building, within the framework of the GREDOR project.
Les recherches menant aux pr ´esents r ´esultats ont b ´en ´efici ´e d’un soutien financier de la R ´egion Wallonne au titre du programme mobilisateur RELIABLE - R ´eseaux ELectriques Intelligents et durABLEs - organis ´e par la Direction g ´en ´erale op ´erationnelle de l’Am ´enagement du Territoire, du Logement, du Patrimoine et de l’Energie - D ´epartement de l’Energie et du B ˆatiment durable en vertu de la convention No 1250487.