An Improved ADP Algorithm for Multistage Unit Commitment of Renewable Energy Integrated Power System

With the continuous penetration of renewable energy sources (RESs), the unit commitment (UC), as a key operational problem in power system, is facing increasingly challenges due to the uncertainties. This paper proposes a decomposed optimization algorithm for the multistage UC with the uncertainties of RESs. The Wasserstein metric-based distribution ally robust optimization (DRO) is extensively applied to handle the uncertainties in UC, and the UC is decomposed into a series of nested subproblems by improved approximate dynamic programming (ADP). Compared with traditional ADP algorithm, the proposed algorithm improves Lagrange cut-based value function to adaptively tackle integer variables, while provides tractable and favorable robust commitment strategies. Moreover, a parallel technique is further used to accelerate the algorithm in both scenario and stage levels. Case studies on several test systems illustrate that the proposed algorithm is favorable in terms of computational accuracy and efficiency.