制冷站系统负荷和能效比改进T-S模糊模型构建

Construction of Improved T-S Fuzzy Models for Load and Energy Efficiency Ratio of Refrigeration Station Systems

制冷站具有非线性、强耦合等特点,导致传统机理的建模面临困难。为实现系统节能优化控制,并改善模型在线修正和移植性能,提出一种基于改进Takagi-Sugeno(T-S)模糊模型的制冷站负荷和能效比动态建模方法,前件结构辨识设计改进天牛须搜索算法,以改善模糊C-均值聚类方法对初值敏感和易陷入局部最优的问题;为实现非线性模型辨识,并降低现场测试数据噪声的影响,设计自适应扩展卡尔曼滤波算法实现模型后件参数辨识和在线修正。实验结果表明,所建立的负荷和能效比预测模型在广州某建筑上运行时,相对误差分别为0.63%和1.49%;使用广州市另一座建筑的数据进行模型可移植性验证,经过500步在线训练,新模型成功收敛,证明所构建模型具备良好的可移植性和适应性。

The characteristics of chiller plants,such as nonlinearity and strong coupling,present challenges for traditional mechanism modeling.To achieve energy-saving optimization control and enhance model online correction and portability,a dynamic modeling method for chiller plant load and energy efficiency ratio based on an improved Takagi-Sugeno(T-S)fuzzy model is proposed.The antecedent structure identification utilizes an improved beetle antennae search algorithm to address the initial value sensitivity and local optima issues in fuzzy C-means clustering.An adaptive extended Kalman filter algorithm is designed for nonlinear model identification and to mitigate the impact of noise in field test data,enabling parameter identification and online correction of the model’s consequent part.Experimental results demonstrate that the established load and energy efficiency ratio prediction models,when applied to a building in Guangzhou,achieve relative errors of 0.63%and 1.49%,respectively.The model’s portability was verified using data from another building in Guangzhou,where the new model successfully converged after 500 steps of online training,proving the model’s good portability and adaptability.