[2Win5] Poster session 2

[2Win5-75] Model Predictive Control Based on Machine Learning and Internal Plant Control Logic

〇Satoshi Kiryu¹, Satoshi Suzuki¹, Tatsuya Iizaka¹, Tetsuro Matsui¹ (1.Fuji Electric Co., Ltd.)

Keywords:AI, Machine Learning, Optimization, Process Control

In recent years, research on operational support technologies utilizing machine learning techniques has become increasingly active in the industrial sector, particularly in plant operations. In such plants, control logic is generally implemented internally to follow the set values configured by operators, and changes to these set values can significantly alter the plant's behavior. As a result, simply applying existing machine learning models often makes it challenging to predict future states, and, in some cases, it becomes impractical to provide operational support for the plant based on such predictions. The objective of this study is to establish a method for calculating appropriate set values for plants with internal control logic. To achieve this, we propose a predictive method that combines machine learning models with the plant's control logic, as well as a method for calculating optimal set values utilizing this approach. To evaluate the effectiveness of the proposed method, we created a case study by incorporating custom internal control logic into a cart-pole simulation environment. By applying the proposed method to this case study, we confirmed that it is possible to achieve more accurate predictions and calculate set values that enable stable control.