Power station operation state monitoring system and method based on multivariate time series prediction

The invention belongs to the technical field of state monitoring, and discloses a power station operation state monitoring system and method based on multivariate time sequence prediction. Collecting m groups of historical power station characteristic data, wherein the historical power station characteristic data comprises power station data, power grid data and environment data; according to the historical power station feature data, calculating a corresponding historical feature coefficient, and constructing a corresponding historical feature coefficient time sequence; training a coefficient prediction model for predicting historical feature coefficients of future time points according to the historical feature coefficient time sequence; inputting the historical characteristic coefficient time sequence into the coefficient prediction model, obtaining a characteristic coefficient of a future time point, and marking the characteristic coefficient as monitoring data; judging whether a fault instruction is gene