Research On Tomato Yield Forecast Method In Northern Solar Greenhouse Based On BA-ELM

Tomato is the main economic crop in our country.It has rich nutritional value and good storage resistance,which is an important agricultural product in the vegetable market.In order to overcome the influence of climatic conditions and weather changes,greenhouse planting technology has been widely used in tomato planting.Accurate greenhouse tomato yield forecasts can help produce growers adjust their greenhouse tomato management plans.Therefore,the research on the yield prediction of greenhouse tomato has important theoretical and practical significance.This paper takes solar greenhouse tomatoes in northern China as the research object.Aiming at the problem of large prediction errors and high computational costs in traditional prediction models,it is difficult to meet the precision requirements of tomato yield prediction.Therefore,a BA-ELM neural network for greenhouse tomato yield prediction model is established in this paper,and the prediction accuracy of the model is verified by experimental data.This thesis carries out related research with the background of the National Natural Science Foundation of China"Research on Intelligent Optimal Control Method of Greenhouse System"（Grant No.61673281）.The main research work of this paper is as follows:（1）Through qualitative analysis and grey correlation analysis,this paper first selects 8main influencing factors:ambient temperature,humidity,irrigation amount,nitrogen fertilizer,phosphorus fertilizer and potassium fertilizer application amount,CO₂ concentration,light intensity are the factors that affect the yield of greenhouse tomato.Based on the above input variables,a model was constructed and a prediction of tomato yield was made.Firstly,the BP neural network prediction model is constructed,Then,in order to realize the optimization of the BP model,the sample training error is reversely transmitted to each neural network node.The results show that the mean absolute percentage error MAPE of the BP model is 2.353%,the mean absolute error MAE is 4.983 t·hm^-2,and the root mean square error RMSE is 5.919t·hm^-2.The output is predicted,but the model complexity is high and the convergence speed is slow and needs to be further optimized.（2）In view of the shortcomings of the BP model,a more flexible ELM single-hidden layer neural network is proposed.The sig function with higher smoothness is introduced as the activation function of the ELM,and the ELM model is established through the generalized inverse matrix and parameters are optimized.The calculated mean absolute percentage error MAPE of the model is 1.677%,the mean absolute error MAE is 3.545t·hm^-2,and the root mean square error RMSE is 3.676t·hm^-2.Compared with the BP model,the convergence speed of the ELM model is accelerated,and the flexibility and robustness are further improved.The ELM neural network model significantly improved the prediction accuracy of greenhouse tomato yield,but the ELM model was less stable.（3）The ELM model oversimplifies the parameter optimization process to improve training efficiency,which leads to insufficient stability of the ELM model.Therefore,in order to solve the existing defects and deficiencies of the ELM model,it is proposed to use BA to optimize the ELM model.The model makes full use of the global optimization ability of the BA algorithm,which combines the generalized inverse operation of the ELM model to realize the optimization and parameter selection of the BA-ELM combined model.The calculated mean absolute percentage error MAPE of the model is 1.048%,the mean absolute error MAE is 2.215 t·hm^-2,and the root mean square error RMSE is 2.644t·hm^-2.Compared with the BP model and the ELM model,the three indexes of MAPE,MAE and RMSE of the BA-ELM model have decreased.The convergence effect of the model was further optimized,and the prediction accuracy of tomato yield was improved.The validity and rationality of the BA-ELM model in predicting the yield of greenhouse tomato was proved,and a high prediction accuracy was achieved.