Distributed and Risk-Averse ADP Algorithm for Stochastic Economic Dispatch of Power System with Multiple Offshore Wind Farms

Abstract

With more and more offshore wind power being increasingly connected to power grids, fluctuations in offshore wind speeds result in risks of high operation costs. To mitigate this problem, a risk-averse stochastic economic dispatch (ED) model of power system with multiple offshore wind farms (OWFs) is proposed in this paper. In this model, a novel GlueVaR method is used to measure the tail risk of the probability distribution of operation cost. The weighted sum of the expected operation cost and the GlueVaR is used to reflect the risk of operation cost, which can consider different risk requirements including risk aversion and risk neutrality flexibly by adjusting parameters. Then, a risk-averse approximate dynamic programming (ADP) algorithm is designed for solving the proposed model, in which multi-period ED problem is decoupled into a series of single-period ED problems. Besides, GlueVaR is introduced into the approximate value function training process for risk aversion. Finally, a distributed and risk-averse ADP algorithm is constructed based on the alternating direction method of multipliers, which can further decouple single-period ED between transmission system and multiple OWFs for ensuring information privacy. Case studies on the modified IEEE 39-bus system with an OWF and an actual provincial power system with four OWFs demonstrate correctness and efficiency of the proposed model and algorithm.

Keywords

• Approximate dynamic programming (ADP)
• alternating direction method of multipliers
• GlueVaR
• offshore wind farm
• risk-averse stochastic optimization