Genetic particle filtering algorithm for interactive multi-model switching and parameter online identification

Abstract

The invention discloses a genetic particle filter algorithm for interactive multi-model switching and parameter online identification, which comprises the following steps: step 1, state estimation: carrying out state estimation by adopting a particle filter algorithm, and replacing resampling of traditional particle filtering with a genetic algorithm selection, crossover and variation method; firstly, inputting a previous state estimation result set, and performing input interaction operation based on a battery model set; inputting the measured value of the battery terminal voltage to update the predicted value of the state estimation, and calculating a likelihood function to update and calculate the model probability; and finally, performing output interaction operation on the state estimation according to a model probability calculation result to obtain a final state estimation result; step 2, respectively estimating the SOC value and the SOH value by using the state estimation method in the step 1; and step 3, updating the SOC value and the SOH value. The algorithm provided by the invention can keep good estimation precision in a nonlinear environment, and has good robustness.

Description

technical field [0001] The invention mainly relates to the technical field of battery state estimation algorithms, in particular to a genetic particle filter algorithm for interactive multi-model switching and parameter online identification. Background technique [0002] At present, most algorithms for electric vehicle power battery state estimation use Kalman filter method, improved Kalman filter method or standard particle filter algorithm when estimating SOC, and ignore the relationship between SOC and SOH when estimating SOH. It mainly has the following three problems: 1. For SOC estimation, the Kalman filter algorithm is only suitable for the environment where the noise is a Gaussian density distribution linear system, while the working environment of the electric vehicle power battery is mostly a nonlinear environment; the improved The current Kalman filter algorithm is mainly to linearize its nonlinear part, which is difficult to meet the current industry requirement...

AI Technical Summary

Problems solved by technology

It mainly has the following three problems: 1. For SOC estimation, the Kalman filter algorithm is only suitable for the environment where the noise is a Gaussian density distribution linear system, while the working environment of the electric vehicle power battery is mostly a nonlinear environment; the improved The Kalman filter algorithm is mainly to linearize its nonlinear part, which is difficult to meet the current industry requirements for the accuracy of power battery state estimation

The standard particle filter algorithm has the risk of particle degradation. If it continues to iterate, a large amount of resources will be consumed in processing trivial particles, which will cause waste of resources and affect the estimation results; 2. When estimating the battery state, if only a single The battery model is estimated, and the estimation accuracy will gradually deteriorate; 3. The interaction between SOC and SOH when the power battery is working, if not suppressed, will further reduce the estimation accuracy

Images