Battery Residual Power Estimation Method Based on NARX-UKF Algorithm

Abstract

The present invention discloses a battery residual power estimation method based on the NARX-UKF algorithm. The existing various methods have various problems. According to the method, voltage and current of a battery under different conditions is firstly measured, then the measured data is pre-processed, an NARX-UKF network is constructed, processed voltage and current training data is input intothe NARX network to train the network, after reaching a training target, test data is input, an input result of the NARX network is an estimated value of an SOC and also is an input quantity of the UKF model, and after the estimated value passes through the UKF model, a state update value obtained is an estimated battery residual power at current time. The method does not need to establish a battery model, and only requires a common measurable quantity to quickly and accurately estimate the battery residual power. The method has the advantages of fast model training, less required parametersand high estimation accuracy.

Description

Technical field [0001] The present invention belongs to the field of battery technology, and proposes a solution for estimating remaining battery power based on NARX (based on nonlinear adaptive regression neural network)-UKF (unscented Kalman filter) algorithm. Background technique [0002] In recent years, with the development of the economy, the number of cars has continued to increase, and environmental problems have become increasingly serious. The use of high-efficiency and environmentally-friendly alternative energy sources is one of the effective ways to solve pollution problems and traditional energy exhaustion problems. It has the advantages of low noise, low pollution, low emissions and high energy efficiency. Therefore, new energy vehicles have been vigorously developed in various countries. The battery management system (BMS, Battery Management System) is the key technology of electric car batteries. The development of battery management system is of great significan...

AI Technical Summary

Problems solved by technology

The noise error of the ampere-hour integration method will cause a large cumulative error under actual application conditions. The open circuit voltage method requires the battery to stand for a long time. The Kalman filter method requires high accuracy in battery modeling. The general neural network method to estimate SOC, the estimation result has a large jump

Images