A Supervised-Learning Assisted Computation Method for Power System Planning

Abstract

To achieve a low-carbon economy, the comprehensive planning of the early retirement of coal-fired power plants (CFPPs), the construction of renewable energy plants, and the investment in energy storage systems (ESSs) are critical. However, the nonlinear characteristic of the alternating current power flow (ACPF) brings difficulties in solving the long-term planning problem efficiently. Hence, in this article, a learning and optimization integrated framework is presented to help the power sector to reach the emission reduction goal economically while guaranteeing system security and reliability. First, the electricity network transition from the fossil fuel-dominated system to the low-carbon-oriented system is planned. Then, a data-driven method is utilized to regress the power flows, and a method is proposed to integrate the data-driven model into an optimization problem to alleviate the heavy computational burden. Because the penetration of intermittent renewable energy is high in a low-carbon energy system, the security and reliability of the planning decisions are verified in the second level by the N-1 security check. Simulation results reveal that the data-driven method can calculate power flows more accurately than direct current power flow (DCPF) and linearized ACPF models. Also, the supervised-learning method can help reduce computing time. The proposed planning model is verified on the IEEE 30-bus system. Through case studies, it can be concluded that our proposed method can reach the low-carbon goals with the lowest cost, and reliability can be ensured at the same time. Because of the coordinated retirement of CFPPs and the proper planning of ESS, the planned electricity system can cope with uncertainties better.