基于趋化-改进粒子群算法的SVR光合速率预测模型

针对回归型支持向量机(SVR)参数选取影响模型性能的问题,提出融合细菌觅食算法趋化操作的改进粒子群混合算法(C-IPSO),以优化SVR的惩罚参数和核参数。同时,为了实现对温室环境的精细控制,结合温室作物生长环境因子,建立一种基于趋化-改进粒子群算法优化的回归型支持向量机温室光合速率预测模型。以温室番茄幼苗期、开花期、结果期为例,与支持向量机和基本粒子群算法优化的支持向量机分别建立的模型进行实验对比。结果发现:建立的三个生长期光合速率预测模型的光合速率实测值和预测值的决定系数分别为0.954 8、0.985 4和0.951 5,均比另外两个预测模型更接近于1,表明该模型预测效果均更佳,并证明了所提算法的有效性,为指导温室环境根据作物光合需求进行精准调控提供了理论基础。

Prediction Model of SVR Photosynthetic Rate Based on Chemotaxis-Improved Particle Swarm Optimization

In order to solve the problem that SVR parameters affected the model performance, this paper put forward a hybrid particle swarm optimization algorithm (PSOA) based on chemotactic operation of bacteria foraging optimization algorithm (BFOA) to optimize the SVR penalty parameters and kernel parameters. At the same time, in order to achieve fine control of the greenhouse environment, combining with the growth factors of greenhouse crops, this paper established a regression model based on chemotactic-improved particle swarm optimization to predict the photosynthetic rate in greenhouse. Taking greenhouse tomato at seedling stage, flowering stage and fruiting stage as examples, this paper carried out experiments and compareison support vector regression (SVR) with support vector regression based on particle swarm optimization(PSO-SVR). The results showed that the determination coefficients of photosynthesis rate and the predicted value of the photosynthesis rate in these 3 growth stages were 0.954 8, 0.985 4 and 0.951 5, respectively, which were closer to 1 than the other two prediction models, indicating the prediction effects of constructed model were better and the proposed algorithm was valid. These results provided theoretical basis for precise controlling and regulating the greenhouse environment according to the needs of crop photosynthesis.