Deep Learning Driven Rolling Evaluation and Optimization Method of Distribution Network Asset Utilization Efficiency

With the development of distribution network and the emergence of new load areas, scientific methods are needed to guide the expansion and planning of distribution network. Traditional investment methods are based on human experience, which tends to cause confusion in distribution network planning and low investment efficiency. In this paper, a deep learning driven rolling evaluation and optimization method of distribution network asset utilization efficiency is proposed. Stacked autoencoder (SAE) is used to evaluate the asset utilization efficiency of distribution network. The input of SAE is the current load level and investment strategy, and the output of SAE is asset utilization efficiency. SAE greatly accelerate the evaluation of asset utilization efficiency. Considering the multiyear rolling investment, the investment in the next year will be evaluated based on the investment results of the previous year, so as to achieve the optimal asset utilization efficiency of multi-year investment. The proposed method is tested on a modified IEEE 34-bus system, which shows the feasibility of the proposed method.