PID离线优化算法库数据集

本数据集针对PID离线优化算法库的算法研究，其中数据均为实际水泥生产线中所产生以及计算衍生数据。在取得水泥产线中的模型参数和对应PID后，通过MATLAB设计神经网络，计算迭代得到优化后的PID及相关数据，优化后的PID相较于初始的PID拥有更好的控制效果。该数据集主要包含BP-PID优化，RBF-PID优化，RNN-PID优化等三种神经网络优化方法的数据，其中BP-PID又根据激活函数的种类和放大系数的不同分为以下四种：输出层激活函数为ArcTan函数，放大系数分别为1x，10x，20x的三种，输出层激活函数为LeakyReLU函数，放大系数为1x的一种；其中RNN-PID又根据学习率和动量因子更新方式的不同分为以下两种：定学习率和定动量因子的一种，同时更新学习率和动量因子的一种；总计七项数据。其中每项数据又包含系统参数，初始PID，PID过程参数，控制量过程参数，输出量和跟总量等，数据量60KB。

This dataset is tailored for algorithmic research on the PID (Proportional-Integral-Derivative) offline optimization algorithm library. All data are either collected from real cement production lines or derived via computational processes.
After acquiring the model parameters and corresponding PID controllers from the cement production line, we designed neural networks using MATLAB and conducted iterative computations to obtain optimized PID controllers and their associated data. The optimized PID controllers demonstrate superior control performance compared to the initial ones.
This dataset primarily covers data from three neural network-based PID optimization methods: BP-PID (Back Propagation Neural Network-based PID), RBF-PID (Radial Basis Function Neural Network-based PID), and RNN-PID (Recurrent Neural Network-based PID).
Specifically, BP-PID is categorized into four subtypes based on the output layer activation function and amplification factor: three subtypes using the ArcTan activation function with amplification factors of 1×, 10×, and 20× respectively, and one subtype using the LeakyReLU activation function with an amplification factor of 1×.
For RNN-PID, it is divided into two subtypes based on the update strategies of learning rate and momentum factor: one with fixed learning rate and fixed momentum factor, and another that updates both learning rate and momentum factor simultaneously. In total, the dataset includes seven subsets of data.
Each subset contains system parameters, initial PID controllers, PID process parameters, control quantity process parameters, output values, and total tracking quantities, with a total data size of 60 KB.

搜集汇总

背景与挑战

背景概述

PID离线优化算法库数据集包含实际水泥生产线中的模型参数和计算衍生数据，通过MATLAB设计的神经网络进行优化，涵盖BP-PID、RBF-PID、RNN-PID三种优化方法的七项数据，总数据量为73.89KB。