Prediction method and system for industrial time series data

Abstract

The invention relates to a prediction method and system for industrial time series data. The method comprises the following steps: acquiring historical record data monitored in a production process; establishing a prediction correction model composed of a prediction network and a filtering algorithm, and training the prediction network by using the obtained historical record data; predicting approximate Gaussian distribution of a target variable at the current moment through the trained prediction network; according to a monitoring value and monitoring error distribution of a target variable at the current moment in the production process; and calculating posteriori distribution of the target variable at the moment through an extended Kalman filtering algorithm in combination with distribution of the target variable, taking a correction result as correction data, putting the correction result back to the prediction network for prediction at the next moment, and iteratively predicting and correcting the target variable. The method does not depend on an accurate mathematical model, filtering correction can be carried out for the problem that errors exist in sensor monitoring values,and better data support is provided for prediction at the later moment.

Description

technical field [0001] The invention relates to the field of industrial production, in particular to a prediction method and system for industrial time series data. Background technique [0002] In industrial production scenarios, production machines must operate under safe production indicators. If the state quantity during the production process is not within the threshold range, it is likely to cause serious production accidents. Therefore, it is very important to give early warning when the machine is not operating normally. necessary. [0003] In the process of industrial production, in order to avoid potential safety hazards, it is necessary to grasp the changing laws of the state quantities of production machines and make accurate predictions. However, the mechanism of most production machines is complex, it is difficult to establish a mathematical model, and there is often a large time delay in predicting the statistical relationship between the target variable and ...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a prediction method and system for industrial time series data to solve the problems that the existing industrial time series data prediction method is difficult to establish a mathematical model, there are errors in sensor data and the target variable is difficult to predict

Images