The traditional greenhouse microclimate model at home and abroad is mostly based on the physical laws such as warm transfer and mass procedures, and the crops physics such as transpiration and respiration. The mechanism model for greenhouse microclima is established through energy balance and material balance. The mechanism model has many unknown parameters, and to measure these parameters require expensive instruments, and the test costs are high. The target of some parameters is time-requiring and workwriter, and the target method is not easy to understand. Due to burning or damage to fuel materials, many model parameters are changed with time, and the model itself is also time-response. Therefore, the traditional greenhouse microklyme mechanism model is poor barbarity, poor adaptation and high costs. It's not easy to master of unexpected greenhouse leaders. It's very difficult to establish an advanced and useful computer control system for drives to low costs. As for experimental modelling, scientists entered and abroad have used the neural network method to establish a greenhouse microclimate model, but this method has several major decisions: it requires a large amount of training data, the model exercise time is long, overall, local minimum, Network Structure decisions depend on subobjective experience, etc. These missions limit the use of the neural network method of the computer's true time control of the greenhouse microclimate.<br>
正在翻译中..