Industrial pollutant anomaly detection method based on TCN-BP

The invention relates to an industrial pollutant anomaly detection method based on TCN-BP. The method comprises the following steps: firstly, carrying out time mark alignment, data cleaning and normalization processing on industrial pollutant concentration and power data, then inputting the data into a TCN model in a vector form to carry out feature extraction, and then respectively inputting time sequence features of the power data and the pollutant concentration data into a full-connection network to integrate the time sequence features; and the residual error of the real data is used as the input of the BP model to train the model. And finally, performing pollutant concentration data prediction by using the trained model, and performing abnormal value detection on the pollutant concentration data according to a prediction result based on a statistical method. According to the method, the feature extraction capability of the TCN model for the time series data is fully fused, and the training of the model is