Predictive and disturbance rejection control for parabolic trough solar field under multiple disturbances and operating points using machine learning and nonlinear disturbance observer
{"title":"Predictive and disturbance rejection control for parabolic trough solar field under multiple disturbances and operating points using machine learning and nonlinear disturbance observer","authors":"Zhuo Chen , Yong-Sheng Hao , Ming Xiao , Guoquan Wu , Zhe Wu","doi":"10.1016/j.solener.2025.113998","DOIUrl":null,"url":null,"abstract":"<div><div>This paper presents a new machine learning predictive and disturbance rejection control scheme for the parabolic trough solar field (PTSF) system which suffers from multiple field disturbances, such as direct normal irradiance variations, heat transfer fluid flow fluctuations, and atmosphere temperature changes. The dynamic model and the disturbance response characteristics of the PTSF system are first investigated to facilitate the controller design. Then, a recurrent neural network (RNN) model is developed to approximate the nonlinear dynamics of PTSF system using the input–output data. Thereafter, a novel RNN-based nonlinear disturbance observer (RNN-NDOB) design is presented to handle multiple disturbances in the PTSF system. An NDOB-assisted RNN-based MPC (NDOB-MPC) is also developed to regulate the PTSF outlet temperature to track the set-point under multiple field disturbances and operating point switching conditions. Finally, a series of simulations on the PTSF system are presented to validate the RNN modeling accuracy, RNN-NDOB disturbance rejection performance, and closed-loop control performance of the proposed NDOB-MPC scheme, respectively. It shows that RNN can approximate the nonlinear dynamics of PTSF system well, and multiple field disturbances can be rejected by RNN-NDOB in real time. Furthermore, the proposed NDOB-MPC scheme exhibits superior control performance compared with two benchmark controllers, a well-tuned two-degree-of-freedom PID controller and a standard linear model predictive control. This superiority was validated across three challenging closed-loop control cases, i.e., step-type set-point tracking with input disturbance, regulation under parameter uncertainty and solar irradiance variations, and operation point switching amidst multiple simultaneous disturbances. The significant performance improvement of the proposed control scheme in these challenging cases demonstrates its promising prospect for enhancing solar energy utilization.</div></div>","PeriodicalId":428,"journal":{"name":"Solar Energy","volume":"302 ","pages":"Article 113998"},"PeriodicalIF":6.0000,"publicationDate":"2025-09-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Solar Energy","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0038092X25007613","RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENERGY & FUELS","Score":null,"Total":0}
引用次数: 0
Abstract
This paper presents a new machine learning predictive and disturbance rejection control scheme for the parabolic trough solar field (PTSF) system which suffers from multiple field disturbances, such as direct normal irradiance variations, heat transfer fluid flow fluctuations, and atmosphere temperature changes. The dynamic model and the disturbance response characteristics of the PTSF system are first investigated to facilitate the controller design. Then, a recurrent neural network (RNN) model is developed to approximate the nonlinear dynamics of PTSF system using the input–output data. Thereafter, a novel RNN-based nonlinear disturbance observer (RNN-NDOB) design is presented to handle multiple disturbances in the PTSF system. An NDOB-assisted RNN-based MPC (NDOB-MPC) is also developed to regulate the PTSF outlet temperature to track the set-point under multiple field disturbances and operating point switching conditions. Finally, a series of simulations on the PTSF system are presented to validate the RNN modeling accuracy, RNN-NDOB disturbance rejection performance, and closed-loop control performance of the proposed NDOB-MPC scheme, respectively. It shows that RNN can approximate the nonlinear dynamics of PTSF system well, and multiple field disturbances can be rejected by RNN-NDOB in real time. Furthermore, the proposed NDOB-MPC scheme exhibits superior control performance compared with two benchmark controllers, a well-tuned two-degree-of-freedom PID controller and a standard linear model predictive control. This superiority was validated across three challenging closed-loop control cases, i.e., step-type set-point tracking with input disturbance, regulation under parameter uncertainty and solar irradiance variations, and operation point switching amidst multiple simultaneous disturbances. The significant performance improvement of the proposed control scheme in these challenging cases demonstrates its promising prospect for enhancing solar energy utilization.
期刊介绍:
Solar Energy welcomes manuscripts presenting information not previously published in journals on any aspect of solar energy research, development, application, measurement or policy. The term "solar energy" in this context includes the indirect uses such as wind energy and biomass