Temperature prediction model for solar greenhouse based on improved BP neural network

Abstract

In the production of solar greenhouse, indoor temperature is a very important index, which is closely related to the growth of crops. In order to solve the problem of hysteretic nature in solar greenhouse temperature regulation, a solar greenhouse temperature prediction model was proposed in this paper. The model took three indoor parameters including temperature, humidity and light intensity of current time in solar greenhouse as inputs and was built on the base of a kind of BP neural network algorithm which was improved by nearest neighbor algorithm, and accurate prediction results were obtained. The mean absolute error (MAE), mean relative error (MRE) and maximal absolute error (MaxE) of the predicted value of the model were 0.89°C, 4.66% and 2.23°C, respectively. All of the three indexes were greatly improved compared with the model without improvement. The model was able to predict the indoor temperature of the greenhouse accurately and as well provide technical support and decision support for the temperature regulation system of the solar greenhouse.