An improved photosynthesis prediction model based on artificial neural networks intended for cucumber growth control

Abstract. The existing photosynthetic rate prediction models consider only a single growing season. However, a photosynthetic rate prediction model intended for full growth of crops is needed. Therefore, a photosynthetic rate prediction model based on artificial neural networks (ANN), which establishes the prediction of the entire photosynthetic process, is presented in this paper. The proposed model was developed using the multi-factor photosynthetic rate data obtained by experiments on cucumber seedlings. The ANN model was trained with the Levenberg-Marquardt (LM) training algorithm. In contrast to the single-phase photosynthetic rate prediction models, in the proposed model a fusion of parameters of all growing stages was applied, whereat all growing parameters were merged into one six-dimensional input signal. Verification of model accuracy and performance has shown that merging of growing parameters has obvious advantage. Moreover, the proposed model satisfied the requirement in terms of training error, namely the training error was less than 0.0001. In addition, the correlation between measured and estimated values was 0.984, thus, good correlation and estimation were achieved. Besides, the test error was less than 5.166%, which proves a high accuracy of the proposed model. Therefore, the proposed prediction model can provide the theoretical basis for the facilities light regulation and technical support.