Data-Driven Sensitivity Coefficients Estimation for Cooperative Control of PV Inverters

This paper introduces a data-driven method to compute voltage magnitude, and power loss sensitivity coefficients in unbalanced low voltage network with high photovoltaic (PV) distributed generation penetration. Implementation of the proposed method is based on the Least-Squares estimator and does not require knowledge of network model parameters, but only requires measurement of voltage magnitude, active, and reactive power. To overcome the data collinearity problem, the sensitivity coefficients are estimated using ridge regression. To track changes in the network operating conditions, the coefficients are updated using the Recursive Least-Squares and to avoid instability in the estimation process during low excitation periods, the method uses the concept of directional forgetting in which old data are discarded only when new information are available. Effectiveness of the proposed method is demonstrated by a case study where the impact of active and reactive power injection is estimated. Besides that, it is also shown how sensitivity coefficients can be useful to establish how each photovoltaic inverter can contribute to set a cooperative control, by achieving losses minimization and ensuring voltage level within limits set by the standards.