基于Dropout优化算法和LSTM的铅酸蓄电池容量预测

针对变电站铅酸蓄电池容量预测模型存在的预测准确率低、泛化能力差等问题，提出一种基于Dropout优化算法和长短期记忆LSTM（long short-term memory）神经网络相结合的容量预测模型。该模型以LSTM神经网络为基础，结合变电站蓄电池充放电特性，将长时间跨度的蓄电池运行数据作为模型的输入，建立多层级LSTM预测模型来提升预测结果的准确率。同时基于Dropout优化算法完成LSTM预测模型的训练，提升模型的泛化能力。工程实际应用表明，相较于传统的LSTM神经网络和BP神经网络，改进模型在长时间跨度预测时具有更高的准确率和更好的泛化能力。

Prediction of Lead-acid Battery Capacity Based on Dropout Optimization Algorithm and LSTM

Aimed at the problems in the capacity prediction model for lead-acid batteries in substations such as a low prediction accuracy and a poor generalization capability, a capacity prediction model based on the Dropout optimization algorithm and long short-term memory(LSTM) neural network is proposed. Based on the LSTM neural network and combined with the charging and discharging characteristics of batteries in substations, the long-term battery operation data is taken as the model input, and a multi-level LSTM prediction model is established to improve the accuracy of prediction results. Meanwhile, the training of the LSTM prediction model is completed based on the Dropout optimization algorithm, thus improving the model''s generalization capability. A practical engineering application shows that compared with the traditional LSTM neural network and BP neural network, the improved model has a higher accuracy and a better generalization capability when predicting the capacity in a long term.