44results about How to "Increase training samples" patented technology

The invention discloses a lithium battery SOH long-term prediction method based on multi-battery data fusion. The method comprises the steps of collecting data of the same kind of lithium batteries in a charging and discharging process, preprocessing, constructing an input matrix of multi-battery data fusion, and sending the input matrix into a multiple-input multiple-output long-short-term memory network model for training; preprocessing the data of the predicted batteries in real time and then sending to the multiple-input multiple-output long-short-term memory network model for prediction; collecting a historical prediction result after prediction and historical real data in the charging and discharging process, and training an NARNN model; and taking the prediction result at the current moment as the input of the NARNN model, and outputting a health state parameter SOH among a plurality of times of charging and discharging in the future. The method overcomes the defects that a traditional battery SOH prediction algorithm is only used for modeling the predicted batteries, generalization is weak, and long-term prediction precision is low. Training samples are greatly increased, and model combination is optimized, so that the accuracy of model prediction is improved, and the accuracy of SOH long-term prediction is improved.