Adaptive Chance Constrained MPC under Load and PV Forecast Uncertainties

Abstract

The recent increase in the intermittent variable renewable energy sources (VRES) results in mismatches between demand and supply that can cause grid instability. These issues can be mitigated with battery energy storage systems (BESS). However, BESS are generally dispatched conservatively to manage uncertainties in VRE forecast. Therefore, this paper proposes an online adaptive stochastic model predictive control (A-SMPC) based approach that minimizes electricity costs by expanding the BESS state of charge (SOC) limits beyond the nominal range of 20% – 80%. Allowing the SOC limits to expand, results in violation of the nominal SOC constraints. Chance constraints are implemented in the proposed A-SMPC method that guarantee that the probability of violating nominal SOC constraints remains below a desired value. Furthermore, the A-SMPC cost function includes time-of-use demand charges that have not been considered before in this type of model. Simulations based on historical load and PV generation data from the Port of San Diego for January 2019 shows that the proposed formulation outperforms the traditional MPC formulation, that does not include nominal SOC constraint violation, by reducing the monthly electricity costs by 7%. The proposed A-SMPC method results in 8% higher BESS utilization which translates to about 1 extra charging/discharging cycle during the analyzed month which is unlikely to have a significant impact on BESS lifetime.