Research On Multi-objective Optimization For Environmental Control Of A Greenhouse Based On Neural Network And NSGA ?

Greenhouse environment was one of the key factors affecting tomato crop yield and quality,there were multi-objective interrelated relationships in complex greenhouse systems,with significant differences in greenhouse costs.Previous research overemphasized addressing single variables that controlled the greenhouse environment,failure to fully consider the problems caused by conflicts between multiple targets for greenhouse energy costs.Therefore,how to reduce greenhouse energy consumption while meeting environmental control,achieve high quality and high yield of tomato crops and the goal of energy saving and efficiency improvement was an important issue to be solved urgently.To better create crop growth conditions and achieve savings in electricity costs.This paper focus on greenhouse tomato crops.Using EMD?RBF neural network to mine potential changes in the greenhouse environment,establish a greenhouse temperature error model,build a multiobjective optimization function for energy consumption,and use Non-Dominated Sorted Genetic Algorithm-II(NSGA?)with the elite strategy to optimize the corresponding decision-making parameters to achieve multi-objective regulation of complex greenhouses.Research content and results:(1)According to the spatial characteristics of the greenhouse environment,multi-source information collection of greenhouse tomato crops,multi-point collection of temperature and light intensity information in the greenhouse.WSN sensor nodes collect real-time environmental data inside and outside the greenhouse and control equipment status data,with wireless sensor networks(WSN)as the main core.Empirical Mode Decomposition(EMD)is used to process the collected data,further reduce the noise of the collected data signals,to improve the accuracy of the prediction,and to fully prepare for the greenhouse temperature prediction modeling.(2)According to the goals of controlling the greenhouse environment and energy cost,build a temperature prediction model,temperature control model,and energy consumption model.Based on the EMD?RBF model and time series model,the input amount is the perturbation parameter and decision parameters that affect the indoor temperature,and the output amount is the temperature in the greenhouse.Simulation results show that EMD?RBF can determine the nonlinear relationship between the disturbance parameters and decision parameters and the greenhouse temperature,and the prediction accuracy is better than the time series model;Calculate the actual working time of continuous control equipment with greenhouse temperature control model;Take the total power consumption of continuous control equipment and discontinuous control equipment as the energy consumption model.(3)Multi-objective optimization control of algorithm.Based on the greenhouse temperature error model and energy consumption model as optimization targets,multi-objective control decision parameters using the NSGA? algorithm and NSGA algorithm.Simulation results show that the NSGA? algorithm has a better ability to search Pareto solution and better control performance for the objective function than the NSGA algorithm.It is found that the former can not only meet the best growing conditions of tomato crops,but also save electricity costs to a large extent.The proposed EMD?RBF neural network is used to construct a greenhouse temperature prediction model and applied to the multi-objective control of NSGA? algorithm.Not only optimizing the combination of decision parameters based on controlling temperature accuracy reduces energy consumption,but also reduces the complexity of multi-objective control.An indepth study of multi-objective optimization control brings greenhouse high tomato crop yields and improves economic benefits.It provides a reasonable way for the greenhouse to determine the intelligent control strategy,and provides an effective method for achieving multi-objective optimal control.