Research on Wind Power Output Sequence Simulation Method Based on Multi-layer convolutional Neural Network

Generating a large amount of simulation data with actual statistical properties from historical wind power output series is the basis for wind power grid integration analysis and long-term planning of new power systems. Since the simulation method of MCMC mainly focuses on the overall probability distribution of wind power output, it is easy to ignore the local characteristics, which results in the simulation data not being able to well capture the multi-scale structure or local patterns existing in the wind power data. Therefore, this article proposes a new method for simulating wind power output sequences based on multi-layer convolutional neural networks. Through multi-layer convolution operations, this method can capture patterns at different scales. Through practical case verification, the method proposed in this paper has significantly improved the probability density function and auto-correlation function compared to the K means MCMC method based on particle swarm optimization.