首页|期刊导航|机械与电子|面向支柱绝缘子清洗的机器人多模式自适应决策与智能控制方法研究

面向支柱绝缘子清洗的机器人多模式自适应决策与智能控制方法研究OA

Research on Multi-mode Adaptive Decision-making and Intelligent Control Method for Insulator Cleaning Robots

中文摘要英文摘要

针对绝缘子清洗机器人在高空支柱绝缘子清洗作业中存在的作业模式僵化与参数控制静态化问题,提出一种融合感知决策与自适应控制的智能清洗方法.首先,构建基于多维度视觉的污秽特征提取模型,实现对污秽类型、覆盖率、曲率、湿润度与厚度的精准感知;在此基础上,提出一种模糊-贝叶斯决策机制,实现干刷、水洗与混合洗 3 种作业模式的鲁棒自主选择.进一步,设计基于深度强化学习的双层清洗参数自适应控制架构,上层通过离线学习全局优化策略,下层结合实时反馈进行局部补偿,实现清洗参数的动态实时优化.同时,引入基于扰动观测器的本体抗扰控制器,保障绝缘子清洗机器人在清洗扰动下的本体稳定.仿真结果表明,所提方法在清洗清洁率、能量效率、模式适应性和运动稳定性方面均显著优于传统固定策略与单一智能控制方法,平均清洁率达 92.1%,能耗降低 36.6%,验证了其在复杂动态环境中具备良好的自适应能力和工程应用潜力.

To address the problems of rigid operation modes and static parameter control in existing robotic cleaning systems for high-altitude post insulators,this paper proposes an intelligent cleaning framework that integrates perceptual decision-making with adaptive control.First,a multi-dimensional vision-based contamination feature extraction model is constructed to achieve accurate perception of con-tamination type,coverage rate,surface curvature,wettability,and layer thickness.On this basis,a fuzzy-Bayesian decision mechanism is developed to enable robust autonomous selection among three cleaning modes:dry brushing,water washing,and hybrid cleaning.Furthermore,a dual-layer adaptive parameter control architecture based on the Deep Deterministic Policy Gradient(DDPG)algorithm is designed,where the upper layer learns global optimization strategies through offline learning,and the lower layer performs real-time local compensation using feedback signals,thereby enabling dynamic and online optimization of cleaning parameters.Concurrently,a disturbance-observer-based anti-disturbance controller is intro-duced to maintain the body stability of the robot under cleaning-induced disturbances.Simulation results demonstrate that the proposed method significantly outperforms conventional fixed strategies and single in-telligent control methods in terms of cleaning efficiency,energy consumption,mode adaptability,and mo-tion stability,achieving an average cleaning rate of 92.1%and a 36.6%reduction in energy consumption.These results validate its strong adaptive capability and promising engineering application potential in com-plex dynamic environments.

裴文龙;徐塑哲;张佰庆;衡可;郭明鑫;季洪鑫

国网徐州供电公司,江苏 徐州 221000国网徐州供电公司,江苏 徐州 221000国网徐州供电公司,江苏 徐州 221000国网徐州供电公司,江苏 徐州 221000国网徐州供电公司,江苏 徐州 221000中国矿业大学电气工程学院,江苏 徐州 221116

信息技术与安全科学

绝缘子清洗机器人模糊-贝叶斯决策深度强化学习自适应控制扰动观测器

insulator cleaning robotfuzzy-Bayesian decisiondeep reinforcement learningadaptive controldisturbance observer

《机械与电子》 2026 (9)

44-56,13

国网江苏省电力有限公司科技项目资助(J2025150)

评论