With the rapid development of energy Internet, the side of terminal users energy consumption data presents explosive growth.SA large amount of abnormal load data may occur in the collected massive data due to device faults or external environmental factors.SThis paper presents a method for detecting and correcting abnormal load data of energy Internet based on PSO-BiLSTM neural network.SFirstly, a large amount of normal load data is used to train the bidirectional LSTM model, and PSO optimization algorithm is selected to optimize the parameters of the prediction model, and the optimized bidirectional LSTM model is used for load prediction.SBased on the load prediction results, the abnormal load data in the load curve are detected by error analysis and outlier judgment criterion. Finally, the abnormal load data detected are corrected by the prediction results.SExperimental results show that this method has good effect on abnormal load data detection, and it is easy to train, and the error rate of abnormal load data detection is low.