Simulation of power system planning and investment decision model based on neural network algorithm

Traditional planning and investment decision-making methods have been difficult to meet the needs of modern power systems. In this paper, a simulation method of power system planning investment decision model based on neural network algorithm is proposed to improve the accuracy and interpretability of power system planning investment decision. The model adopts a hybrid neural network architecture which combines physical constraint embedding with data driving. The core modules include physical constraint layer, feature extraction network, multi-objective optimization layer and dynamic strategy adjustment module. By modeling the uncertainty of renewable energy output using a conditional variational autoencoder (CVAE), a comprehensive cost function was established. An improved NSGA-III algorithm was then applied to achieve coordinated optimization of economic, reliability, and low-carbon objectives. In addition, a deep deterministic policy gradient (DDPG)-based reinforcement learning (RL) agent was developed to adapt to changes in external factors such as electricity prices and policies. The simulation results show that compared with the traditional methods, this model has obvious advantages in economy and the investment cost is significantly reduced. In terms of new energy consumption, the power abandonment rate has dropped significantly, showing stronger peak clipping ability and intermittent adaptability to new energy, In terms of low carbon, the intensity of carbon emission is significantly reduced, showing good emission reduction potential and policy adaptability. At the same time, this model achieves a reasonable balance between calculation time and resource consumption in calculation efficiency, and provides an efficient and reliable new method for power system planning and investment decision.