Method, device, equipment and storage medium for constructing forecast model of complex industrial system based on deep learning

Abstract

The invention provides a complex industrial system forecast model construction method, device, equipment and storage medium whose model structure, parameters and variable order are unknown. The complex industrial system forecast model construction method includes: establishing a dynamic model of the industrial system, determining input variables and output variables of the dynamic model, and the output variable is a variable to be predicted; using LSTM to establish a forecast model, and combining the dynamic model The input variable is input as a single neuron of LSTM, the output data of the dynamic model is used as label data, and the unknown variable order of the dynamic model is represented by the number of neurons of the LSTM, through the training algorithm, according to the The error between the label data and the predicted model output determines the model parameters of the LSTM. Aiming at the problem that it is difficult to establish a dynamic mathematical model for complex industrial systems, the dynamic system modeling method is combined with the deep learning technology of complete information space to realize the construction of a complex industrial dynamic system forecast model.

Description

technical field [0001] The invention belongs to the technical field of industrial artificial intelligence, and relates to a method, a device, a device and a storage medium for constructing a prediction model of a complex industrial dynamic system based on deep learning. Background technique [0002] Since it is difficult to establish a dynamic mathematical model for complex industrial systems, its operational decisions still rely on knowledge workers to make manual decisions based on experience and knowledge to optimize the operational decisions of industrial systems. The key is to predict the product quality of the system reflected in the operational actions of the decision. , efficiency, consumption of production indicators and key process parameters are within the target value range. Due to the complexity of the manufacturing process, most production indicators and key process parameters cannot be accurately measured online, such as the concentration and particle size of ...

AI Technical Summary

Problems solved by technology

Due to the complexity of the manufacturing process, most production indicators and key process parameters cannot be accurately measured online, such as the concentration and particle size of the grinding process, the detection of key process parameters such as the caustic concentration in the alumina production process, and the production of fused magnesia. The rate of change of demand in the process, etc., can only be obtained through manual testing or post-event statistics

Therefore, there is a serious lag, and knowledge workers can only make decisions based on experience and knowledge based on lagging information, resulting in problems such as unstable product quality and high consumption

To achieve accurate forecasting of production indicators and key process parameters, it is necessary to solve the modeling problem that existing system identification methods and deep learning methods cannot be used for complex dynamic system forecasting models with unknown model structure and unknown order of input and output variables.

Images