Research On SOC Estimation Algorithm Of Power Battery Based On PSO-BP Neural Network

The state of charge of a power battery,that is,the remaining capacity of the battery,is usually expressed by the state of charge(SOC).the accurate estimation of its value is of great significance.the existing SOC estimation algorithms have other defects such as long calculation period and low accuracy.among many SOC estimation algorithms,the data-driven model can well eliminate some defects of the model driving itself,while the data-driven model can effectively shorten the estimation period and make the estimation more efficient.Firstly,a power battery performance analysis and evaluation platform is built to obtain the test data required by the data-driven model.On this basis,the model-driven algorithm is modeled,parameter identification and algorithm simulation are carried out,so as to analyze the defects of the summarized model-driven algorithm and the internal factors causing large errors,and carry out detailed exploration on the data-driven algorithm represented by neural network algorithm.Then,an estimation model based on traditional BP neural network is established.The terminal voltage,current and temperature of the power battery are selected as network inputs,and the gradient descent method is used to realize algorithm simulation and further verification under HPPC and DST conditions.Finally,the inherent defects and error sources of the traditional BP neural network modeling and simulation results are deeply explored,and an improved scheme for optimizing the traditional BP neural network algorithm by using the particle swarm optimization algorithm is proposed.The defects of the traditional BP neural network algorithm are made up and corrected pertinently,and the optimization algorithm PSO-BP is simulated to verify the scientific rationality of the optimization algorithm by comparison.The research results show that the overall prediction effect of the data-driven algorithm is more efficient than the model-driven algorithm,and the traditional BP neural network algorithm has the defects of low efficiency and precision.The optimization algorithm PSO-BP neural network effectively alleviates and eliminates the prediction defects of the traditional algorithm,with high precision and good robustness.