Method for estimating residual life of new energy vehicle-mounted battery based on model learning

Abstract

The invention relates to a method for estimating the residual life of a new energy vehicle-mounted battery based on model learning. The method specifically comprises the steps that capacity variablesreflecting the residual service life of the vehicle-mounted battery are selected and determined, and periodic variation data of the battery capacity are collected; the capacity data are processed necessarily so as to meet the input-output learning relationship required by a Gaussian process regression algorithm; the processed data are learned through the Gaussian process regression algorithm, andhyper-parameters are solved through a conjugate gradient algorithm; and a root-mean-square unscented Kalman filter algorithm is applied to a learned model, and the estimating accuracy of the capacityis improved through the time updating and measuring updating stages of the algorithm. According to the method for estimating the residual life of the new energy vehicle-mounted battery based on modellearning, the problem of complex mechanism analysis in the vehicle-mounted battery is avoided, and the real-time estimating precision of the residual service life of the vehicle-mounted battery is further improved through improving of algorithm numerical stability and according to the real-time capacity measuring data.

Description

technical field [0001] The invention relates to the field of estimating the remaining life of a vehicle battery, in particular to a model learning method for estimating the remaining life of a new energy vehicle battery. Background technique [0002] The core difference between domestic new energy vehicles and traditional vehicles is that they rely on the power system of energy storage batteries. With the promotion and popularization of new energy vehicles in China, there are also some problems in the application of new energy vehicles. For example, the vehicle battery is easy to Affected by temperature, voltage and current, charge and discharge times, etc., the service life of the battery is reduced. This not only needs to be solved from the perspective of physical and chemical mechanisms such as material research and development, product innovation, but also needs to be improved from the perspective of functional assistance. The ability to predict the health of the battery...

AI Technical Summary

Problems solved by technology

[0002] The core difference between domestic new energy vehicles and traditional vehicles is that they rely on the power system of energy storage batteries. With the promotion and popularization of new energy vehicles in China, there are also some problems in the application of new energy vehicles. For example, the vehicle battery is easy to Affected by temperature, voltage and current, charge and discharge times, etc., the service life of the battery is reduced. This not only needs to be solved from the perspective of physical and chemical mechanisms such as material research and development, product innovation, but also needs to be improved from the perspective of functional assistance. The ability to predict the health of the battery, such as the prediction of the remaining service life of the on-board energy storage battery. Therefore, whether the prediction accuracy of the remaining service life of the on-board battery can be improved will be directly related to the performance indicators of the new energy vehicle itself, and it is related to the domestic market of new energy vehicles. foreign market share and the international status of the country's manufacturing power

[0003] In the past, people's prediction of the remaining service life of the vehicle battery relied on the establishment of a physical model of the vehicle battery, such as directly testing the internal resistance loss of the vehicle battery, testing the capacity of the vehicle battery, and the electrochemical mechanism, etc., in order to obtain information that can reflect its remaining life. Specific mathematical models of life changes, but there are often large errors in this way. Even if the difficulties in the process of model establishment, especially in the process of testing these data are overcome, it is difficult to establish a model that is close to the real remaining service life

Images