Strategy model training method and device, electronic equipment and storage medium

Abstract

The invention provides a strategy model training method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a historical operation data set of industrial control scene information for pre-training, and obtaining a system dynamic characteristic model, a pre-training income model, a pre-training loss model and a pre-training strategy model, mixing a simulation operation data set obtained by the pre-training strategy model and the system dynamic characteristic model with a historical operation data set to obtain a mixed data set, and performing joint training based on the pre-training strategy model, the pre-training income model and the pre-training loss model to obtain a training result; enabling loss data output by the trained loss model to be smaller than a preset loss threshold value and enabling income data output by the income model to reach the maximum value, and therefore the simulation operation data set and the historical operation data set are mixed, mixed training is conducted on the obtained mixed data set, offline reinforcement learning is achieved, and the control strategy optimization effect in the industrial control scene is better.

Description

technical field [0001] The present application relates to the field of computer technology, and in particular to a training method, device, electronic equipment and storage medium of a strategy model. Background technique [0002] At present, there are a large number of production control links in the industrial system. In related technologies, the operation control of each production control link in the industrial system is usually based on the control strategy model learned by using the online reinforcement learning framework. However, in training the control strategy model, the online reinforcement learning framework requires a large amount of interaction and trial and error with a high-fidelity simulator or real system environment, and collects system state and action data for model optimization training. However, due to the complexity of industrial systems, the cost of obtaining high-fidelity simulators corresponding to real industrial systems is extremely high, and the...

AI Technical Summary

Problems solved by technology

However, due to the complexity of industrial systems, the cost of obtaining high-fidelity simulators corresponding to real industrial systems is extremely high, and there is no virtual environment close to the real scene for interaction, resulting in the trained control strategy model, which is aimed at the control effect of industrial systems. bad

Images