Fault early warning method for drainage system based on multi-source heterogeneous data

The invention provides a multi-source heterogeneous data drainage system fault early warning method. The method comprises the following steps: acquiring three-phase current, three-phase voltage and power factor after a drainage pump is started, vibration waveform data and sound data when the drainage pump runs, and switching value data of starting and stopping of the drainage pump; historical data of three-phase current, voltage median, power factor, vibration peak-to-peak value and pump starting time 10 seconds after the drainage pump is started are collected, the historical data are adopted, a GBDT gradient decision boosting tree is utilized to train the prediction model, and the three-phase current, voltage median, power factor and vibration peak-to-peak value 10 seconds after the drainage pump is started are calculated. The current value of the pump starting time is predicted, the upper limit and the lower limit of a predicted value are determined through the 3 sigma principle of data normal distribution,