优化RBF神经网络控制水厂混凝剂投加的研究

Optimized RBF Neural Networks Predicts Coagulation in Waterworks

净水厂混凝剂投加受多重进水因素影响,且变化规律呈现高度的非线性模式,难以控制,采用PSO(粒子群算法)对RBF(径向基神经网络)进行优化,建立误差反向传播的非线性高维映射水厂投药量动态模型。相较于单一RBF模型,优化后的RBF模型平均相对误差降低了3.05%、最大相对误差降低了0.1986,迭代收敛速度快,对不同的水厂也具有良好的适应性。

The amount of coagulation in the process of water purification in water plants is influenced by multiple water ingress factors,and the law of change presents a highly nonlinear pattern difficult to control.Based on the RBF optimized by PSO,a dynamic model of nonlinear high-dimensional mapping water plant with error reverse propagation is established.Compared with the single RBF model,the average relative error is reduced by 3.05%,the maximum relative error is reduced by 0.1986,and has a faster iterative convergence speed,and also has good adaptability to the water plant of different processes.