基于模型预测控制与粒子群优化的雾化喷药控制方法研究

A Model Predictive Control-Particle Swarm Optimization Integrated Approach for Atomized Spraying Control

摘要: 传统喷药系统存在控制精度低、响应滞后及抗扰动能力不足等问题，本研究提出了一种基于模型预测控制(Model Predictive Control, MPC)与粒子群优化(Particle Swarm Optimization, PSO)融合的智能雾化喷药控制系统。首先，将该系统建立为二阶动力学模型，建立其状态空间模型并描述喷雾浓度的动态变化过程；其次，构建MPC控制器，通过滚动优化策略实现对系统输出的动态调节；在此基础上，构建多性能指标，诸如均方误差、超调量和控制能量。引入PSO算法对MPC关键参数进行全局优化。最后，在MATLAB环境下进行仿真实验，结果表明，该方法在跟踪精度、超调量及控制平滑性方面均优于单一MPC及开环控制策略，但在控制能量方面略有增加。MPC + PSO的均方误差(MSE)为0.0409，单一MPC和开环控制分别为0.0482和0.0655；MPC + PSO的超调量(overshoot)为21.65，单一MPC和开环控制分别为25.37和35.12；MPC + PSO的能量(energy)为30.26，单一MPC和开环控制分别为29.54和28.47。

Abstract: Traditional spraying systems suffer from problems such as low control accuracy, response delay, and insufficient disturbance rejection capability. To address these issues, this study proposes an intelligent atomization spraying control system based on the integration of Model Predictive Control (MPC) and Particle Swarm Optimization (PSO). First, the system is modeled as a second-order dynamic system, and a state-space model is established to describe the dynamic variation process of spray concentration. Second, an MPC controller is designed to achieve dynamic regulation of system output through a rolling optimization strategy. On this basis, multiple performance indices, including mean square error, overshoot, and control energy, are constructed. The PSO algorithm is then introduced to globally optimize the key parameters of the MPC controller. Finally, simulation experiments are conducted in the MATLAB environment. The results demonstrate that the proposed method outperforms both the standalone MPC and open-loop control strategies in terms of tracking accuracy, overshoot suppression, and control smoothness, although it slightly increases control energy consumption. Specifically, the mean square error (MSE) of MPC + PSO is 0.0409, compared with 0.0482 for standalone MPC and 0.0655 for open-loop control. The overshoot of MPC + PSO is 21.65, while standalone MPC and open-loop control achieve 25.37 and 35.12, respectively. In terms of energy consumption, MPC + PSO reaches 30.26, whereas standalone MPC and open-loop control consume 29.54 and 28.47, respectively.