双时间尺度温室环境最优控制仿真验证OA

To address the problem of control performance degradation caused by long-term weather prediction errors in optimal control of greenhouse climate,a control objective model that represents the profit obtained in the control process was established based on lettuce and energy-related market prices in Beijing.Beijing's annual weather data,collected from the weather station,were used as the reference for long-term weather predictions.Three methods,including Gaussian white noise,time series reversal,and time substitution,were used to generate weather prediction errors.These transformed weather data were then used as real weather data for simulations.A greenhouse-lettuce mechanistic model was taken as the system model for simulations.The control performance was analyzed for the two-time-scale decomposed Receding Horizon Optimal Control(RHOC).Its yield,cost,and profit were compared with those from the traditional Model Predictive Control(MPC).The results showed that:MPC exhibited superior performance only under perfect weather predictions,whereas RHOC improved the profit by 1.24％,20.59％,and 21.32％respectively compared to MPC in scenarios with Gaussian white noise,time series reversal,and time substitution.The quantitative analysis in this study confirmed the superiority of the two-time-scale optimal control of greenhouse climate,and provided data support for the online implementation of control algorithms.