Multi-scale state monitoring method based on big data

The invention relates to a multi-scale state monitoring method based on big data. The multi-scale state monitoring method comprises the following steps of S1, performing data preprocessing on data ina data warehouse; s2, establishing a reference model; s3, asynchronous information fusion: importing actual data detected by outer-layer equipment into the dynamic working condition reference model and the steady-state working condition reference model, and calculating and obtaining a dynamic residual error and a steady-state residual error; dividing the dynamic residual errors and the steady-state residual errors of a plurality of adjacent corresponding time into a plurality of information particles in an information granulation mode; performing information fusion of the dynamic residual error and the steady-state residual error by utilizing Kalman filtering to obtain a fusion residual error; and S4, performing scale analysis on the fusion residual error to construct a state detection signal. Technical bases can