Power Battery Scheduling Optimization Based on Deep Deterministic Strategy Gradient Algorithm With Bayesian Optimization

Authors:

Abstract
Under the promotion of the national new energy strategy, the utilization rate of renewable energy is steadily increasing, and the application scale of power batteries is growing. The scientific scheduling optimization algorithm can fulfill the application requirements of power batteries across diverse scenarios and enhance resource utilization. In the multi-objective optimization scenario, traditional algorithms cannot effectively address the challenge of parameter tuning due to an explosion of parameter combinations, nor can they adapt the scheduling strategies dynamically. Therefore, this paper introduces a depth deterministic strategy gradient scheduling optimization algorithm utilizing Bayesian optimization. It defines a multi-objective optimization function and employs Bayesian optimization to automatically adjust the weight coefficients in diverse application scenarios, thereby expediting the problem-solving process. The experiment demonstrates that the trained model satisfactorily meets the criteria for cost, energy density, cycle life, temperature, and internal resistance simultaneously. It exhibits rapid convergence, timely provision of scheduling strategies, and successful resolution of the multi-objective optimization challenge. This enhancement increases the efficiency of adjusting the weight coefficient and improves the algorithm’s adaptability to individualized requirements.
Funding Information
  • Guizhou Power Grid Company Ltd. Science and Technology Project (GZKJXM20232475)

This publication has 41 references indexed in Scilit: