基于神经网络PID算法的优化饲料配制系统

Research and optimization of feed preparation system based on neural network PID algorithm

【目的】设计饲料配制控制系统,并采用神经网络PID优化算法实现对配料精度的提高。【方法】以西门子S-200 smart型PLC为主控设计饲料配制控制系统,针对现有常规PID算法的控制策略存在超调大、收敛慢等缺陷和BP神经网络梯度下降过程容易出现局部最小化问题,提出以附加动量项的BP神经网络PID算法实现称重误差的降低。【结果】基于动量项的梯度下降法建立的BP神经网络PID算法模型解决了参数自学习整定问题,在响应速度上该算法与PID算法对比为3∶1,试验后平均精度99.6%。并在收敛速度和改善超调现象具有更高效的表现。【结论】配料系统经算法优化后误差得到有效控制。

【Objective】To design a feed preparation control system in view of the operation requirements of feed weighing and batching in small and medium-sized pastures and the weighing error of feed dynamic weighing by using the neural network PID optimization algorithm to improve the batching accuracy.【Methods】Siemens S-200 smart PLC was taken as the main control,the feed preparation control system was designed,which led to summary that the control strategy of the existing conventional PID algorithm had the defects of large overshoot and slow convergence,and the local minimization problem was easy to occur in the gradient descent process of BP neural network.A BP neural network PID algorithm with additional momentum term was proposed to reduce the weighing error.【Results】The BP neural network PID algorithm model based on the gradient descent method of momentum term solved the problem of parameter self-learning tuning,and had more efficient performance in convergence speed and improving overshoot.【Conclusion】The actual verification shows that the batching error is effectively controlled.