SOC estimation method based on fusion DAE-Bi-LSTM

The invention provides an SOC estimation method based on fusion DAE-Bi-LSTM, and the method comprises the steps: collecting the collection data of a lithium ion battery, and obtaining the training data according to the collection data; performing data preprocessing on the training data to obtain preprocessing result data; constructing a noise reduction auto-encoder, inputting the preprocessed battery data, reducing noise in the battery data, and extracting feature information of the battery data; constructing a plurality of Bi-LSTM models, initializing the weight of each model, inputting the features extracted by the noise reduction auto-encoder, mapping the features into final output after nonlinear variation, and obtaining a prediction result; and constructing a fusion DAE-Bi-LSTM model, updating weight distribution according to an error of each Bi-LSTM model by using an Adaboost algorithm, and fusing prediction results of the plurality of Bi-LSTM models by using weights to obtain a final estimated SOC valu