Optimization of Deep Belief Network and Its Application on Equivalent Modeling for Micro-Grid

With the advantages of high energy utilization efficiency, safe and stable work, flexible and convenient installation, microgrid can make up for the defects of traditional power grid. In order to establish the multi-objective grid-connected model and further improve the optimization efficiency of the microgrid, the microgrid is regarded as an integral external system. The microgrid equivalent model is constructed by taking the current and power of the grid-connected access terminal in various operating states of the microgrid as the input and output of the optimization model, respectively. The input weight and hidden layer threshold of the traditional microgrid model are set randomly and difficult to be dynamically adjusted, which leads to the lack of self-adaptability of the limit grid model and affects the optimization accuracy. The depth belief network is applied to the equivalent modeling of micro grid to improve the modeling accuracy and optimization efficiency. Compared with the traditional neural network, the forecast accuracy and fitting accuracy of the proposed network are improved.