Fairness-aware Photovoltaic Generation Limits for Voltage Regulation in Power Distribution Networks using Conservative Linear Approximations

TL;DR

The paper tackles over-voltage in PV-heavy distribution networks by computing day-ahead, fairness-aware PV generation limits. It combines scenario-based stochastic optimization with Conservative Linear Approximations (CLA) to linearize AC power-flow constraints and introduces a fairness objective that minimizes disparities in PV curtailment across owners. The method yields a linear program that can be solved offline and communicated to inverters daily, demonstrated on a CIGRE benchmark network with multiple PV plants. Results show that incorporating fairness reduces cross-plant disparities (as measured by Jain's Fairness Index) but increases total curtailed energy, and CLA-based voltages closely align with true AC solutions, validating the approach's practical viability.

Abstract

This paper proposes a framework for fairly curtailing photovoltaic (PV) plants in response to the over-voltage problem in PV-rich distribution networks. The framework imposes PV generation limits to avoid overvoltages. These limits are computed a day ahead of real-time operations by solving an offline stochastic optimization problem using forecasted scenarios for PV generation and load demand. The framework minimizes the overall curtailment while considering fairness by reducing disparities in curtailments among different PV owners. We model the distribution grid constraints using a conservative linear approximation (CLA) of the AC power flow equations which is computed using a set of sampled power injections from the day-ahead predicted scenarios. The proposed framework is numerically validated on a CIGRE benchmark network interfaced with a large number of PV plants. We compare the performance of the proposed framework versus an alternative formulation that does not incorporate fairness considerations. To this end, we assess tradeoffs between fairness, as quantified with the Jain Fairness Index (JFI), and the total curtailed energy.

Figures (8)

• Figure 1: Flow chart for the computation of day-ahead PV generation limits.
• Figure 2: CIGRE benchmark low-voltage distribution grid with multiple PV plants that result in over-voltage problems.
• Figure 3: Day-ahead scenarios of (a) PV generation, (b) active power demand, and (c) reactive power demand. The tags 'low' and 'up' correspond to the 10-th and 90-th percentile of the predictions in each case.
• Figure 4: Simulation results for the case when fairness is not considered i.e., ($\alpha_2 = 0$).
• Figure 5: Simulation results for the case with fairness ($\alpha_2 = 7$).