Due to the rapid increase in residential PV capacity, it gets more and more difficult for the grid to absorb power at peak hours due to overvoltage issues, causing concerns to DSOs and prosumers who are temporarily unable to inject power into the grid.

In collaboration with Belgian DSOs and via the use of metered consumption and injection data, dedicated Monte-Carlo analyses were performed to investigate mitigation measures, among which the PQ control of residential inverters.

The principle of this mitigation measure is to apply a dedicated control to the inverters since they can control their active and reactive power output at will. By either reducing their active power output or absorbing some reactive power in parallel of the active power production, the voltage level can be reduced and inverter tripping due to overvoltages can be avoided.

Several types of PQ controls are possible and have been investigated in this study:

- Constant cos(phi) : absorbing reactive power in parallel of the active power injected, following a constant cos(phi) whatever the active power level;

- Cos(phi) = f(P) : the cos(phi) depends on the active power output (typically, cos(phi) = 1 for low active power level and linearly decreases starting from 50% P);

- Droop control active power P = P(U) : reduce the active power output following a droop curve based on the voltage level measured locally;

- Droop control active power Q = Q(U) : absorb reactive power following a droop curve based on the voltage level measured locally;

- Droop controls active and reactive power in parallel : Q(U) with P(U) as backup if the action of Q(U) is not sufficient to reduce the voltage

Even though only some countries already have a clearly defined regulatory framework for such use of residential PV inverters, it is worthy to analyze the potential benefits of these controls.

In order to analyze and compare the relative efficiencies of the different controls, a dedicated parametrized LV feeder model was developed in Matlab Simulink and Monte-Carlo simulations were performed, with variations of different input parameters and randomly assigned prosumer profiles, based on existing DSO datasets. To account for the disconnection occurrences and corresponding production losses, the inverters’ dynamic behavior was also modelled, down to a time resolution of 10 seconds. The disconnection and reconnection of the inverters were thus accurately modelled and their impact on the production losses accounted for.

Thanks to hundreds of scenarios simulated, statistical insights could be obtained on the efficiencies of the different control strategies, with respect to different KPI’s of interest for the DSOs:

- Global production losses;

- Equity among prosumers;

- Global reactive power flow;

- Impact on the maximum current in the feeder (cf. ampacity);

- Impact of having different penetration level of inverters with and without controls (relevant if only recent inverter have PQ controls activated by default).