Analysis of methods for optimizing the parameters of PID controllers to improve the accuracy of industrial automation systems

Abstract

In modern conditions, industrial automation requires high precision and reliability of process control. One of the key tasks is to adjust the parameters of proportional-integral-differential (PID) regulators, which are widely used in automatic control systems. However, traditional tuning methods often do not provide the necessary accuracy and require considerable time. The article analyzes modern methods for optimizing the parameters of PID controllers, including mathematical modeling, the use of evolutionary algorithms and machine learning. Grid search algorithms, genetic algorithms, and neural network-based methods are considered. Both classical approaches (the Ziegler-Nichols method) and methods based on automated optimization criteria, such as the absolute error integral (IAE) and the quadratic error integral (ISE), are used for comparative evaluation. The experiments have shown that the use of modern optimization methods based on artificial intelligence algorithms can significantly improve the accuracy of PID control settings compared to traditional approaches. For example, the use of genetic algorithms reduced the average adjustment error by 15%, and the use of machine learning methods by up to 20%. Limitations have also been identified in the application of some methods for tasks with complex process dynamics. The results of the study emphasize the importance of choosing the appropriate optimization methodology, depending on the conditions and specifics of the task. Despite the high efficiency of modern algorithms, their application requires additional computing resources and time to train models. The analysis showed that the integration of modern methods for optimizing the parameters of PID controllers significantly improves the accuracy and reliability of industrial automation systems. Further research involves the development of hybrid methods that combine the advantages of various technologies.