Greenhouse Coordinated Control Strategy Based On Deep Reinforcement Learning

The growth of crops in greenhouse is closely related to environmental factors,and it is important to develop a reasonable greenhouse environmental factor control strategy for efficient greenhouse production.To solve the problem of uncoordinated control of temperature,light,and water and fertilizer environmental factors by existing greenhouse environmental control methods,a coordinated control decision method for greenhouse environmental factors based on deep reinforcement learning is proposed,and a coordinated control decision system for greenhouse is designed and developed based on this method.The main research works and conclusions are:(1)A greenhouse actuator control method based on deep reinforcement learning.A deep reinforcement learning interaction environment was constructed,deep reinforcement learning action,state,and reward functions were designed,and the actuator control decision model was trained by collecting environmental factor data inside and outside the greenhouse.The final reward values were 78.1,69.1,and 77.9for the DDPG,SAC,and PPO deep reinforcement learning algorithms,respectively;the PPO algorithm converged faster and took less time to compute.The simulation experimental results show that the root mean square error of temperature control using deep reinforcement learning actuator control decision model is 0.79°C.(2)Coordinated control strategy of greenhouse temperature and light based on deep reinforcement learning.The temperature and light coordinated control model action,state and reward functions are designed with the crop photosynthesis rate and energy consumption as the optimization objectives,the LSTM network is introduced to optimize the stability of the deep reinforcement learning decision,and the PPO algorithm is used to train the greenhouse temperature and light coordinated control decision model.The simulation experimental results show that compared with the traditional control method,the temperature and light coordination control method based on deep reinforcement learning increases photosynthesis rate by 5.8% and reduces regulation energy consumption by 11.3%.The final selection of the corresponding nutrient solution irrigation volume for group T4 was used as the irrigation coordination control volume.(3)Nutrient solution irrigation volume regulation strategy based on total crop photosynthesis.The information of greenhouse crop leaf area and environmental factors such as temperature and light is collected to calculate the total photosynthetic amount of crops during the time period,and the amount of nutrient solution irrigation is dynamically adjusted according to the total photosynthetic amount to establish the coordination relationship between temperature and light environment and the amount of water and fertilizer irrigation.There were five levels of nutrient solution irrigation in the experimental group: T1(50%w),T2(75%w),T3(100%w),T4(125%w)and T5(150%w).The experimental results showed that T4 grew best in the experimental group,and the volume of nutrient solution irrigation was basically the same as that of the control group,and the cumulative growth of plant height increased by 3.14% and the cumulative growth of stem diameter increased by 10.23% compared with the control group.(4)Greenhouse environmental coordination control system development.Combining the deep reinforcement learning temperature and light coordination control strategy and nutrient solution irrigation strategy,the greenhouse temperature and light environment and fertigation coordination control method was proposed.A deep reinforcement learning-based greenhouse environmental coordinated control system was developed,which includes modules of environmental factor acquisition,humancomputer interaction,greenhouse control decision and actuator control.The experimental results show that compared with the traditional greenhouse environmental factors and fertigation strategy,the greenhouse temperature,light,water and fertilizer environment coordinated control method reduces energy consumption by 8.1%,increases crop height growth by 16%,increases stem diameter growth by 27.6% and reduces nutrient irrigation by 7.9%.The deep reinforcement learning greenhouse execution mechanism control decision-making model established by the above research can accurately regulate greenhouse temperature.The use of deep reinforcement learning greenhouse environment coordination control decision-making method enables crops to achieve better growth,reduce greenhouse regulation energy consumption and fertigation volume.Coordinated greenhouse control decisions based on deep reinforcement learning technology can achieve efficient use of energy and water and fertilizer,which is conducive to the green and sustainable development of facility agriculture and improve the level of intelligent greenhouse environmental control.