Bi-Level Planning Method for Distributed Energy Storage Siting and Sizing Considering Demand Response

A bi-level planning method is proposed for distributed energy storage (DES) siting and sizing considering demand response. The upper level model aims to minimize electricity cost of users and demand response frequency of DES with DES participating in demand response (DR) program. Deep reinforcement learning (DRL) algorithm using dueling network architecture is used to solve the optimal combined sequential response strategy (CSRS) of DES. The immediate reward function is defined to evaluate the effectiveness of DES response strategy considering voltage amplitude, DES response frequency and electricity cost. The lower level model is deployed to determine the DES allocation scheme based on the optimal CSRS from the upper model. The planning scheme with the maximum return on investment is obtained by iterative calculation. The proposed planning method is conducted in the modified IEEE 33-node system. In time of use pricing mechanism, the proposed method brings significant economic benefit to users as well as operating benefit including voltage regulation and peak load shifting via sequential combinatorial optimization charge-discharge strategy of DES.