Abstract: 【Objective】 To study the dynamic prediction model of tomato fruit diameter variation in greenhouse, which can provide certain decision support for the law of water and fertilizer required by tomatoes. 【Methods】 Based on the fruit diameter as the research object, with five of tomato fruit enlargement period data as a model, using principal component analysis (PCA) to perform the plant physiological ecology and environment information analysis, extract the main ingredients, again with the principal components as independent variables, output variables as the dependent variable, to establish BP neural network regression dynamic prediction model including air temperature, air humidity, soil moisture content, leaf temperature and fruit diameter. In addition, the data measured in the fruit expansion period of 3 tomato plants were used as the test data to predict and compare the predicted and measured values. 【Results】 The first decision coefficient for tomato plant predicted and the measured values (R²) was 0.964, and the root mean square error (RMSE) was 0.238; The second tomato plant decision coefficient of the predicted values and the measured values (R²) was 0.960, and the root mean square error (RMSE) was 0.051; The first three decision coefficient of tomato predicted and the measured values (R²) was 0.951, and the root mean square error (RMSE) was 0.047. 【Conclusion】 The model can predict the change of tomato fruit diameter in the greenhouse in a short time, and it can be used to predict the change of tomato fruit diameter in the autumn in the greenhouse in Xinjiang. The water and fertilizer can be fine-tuned according to the difference between the predicted value and the actual measurement.