基于BP神经网络建立杏鲍菇生长的CO_2浓度预测模型及其调控策略

Loop Control Stategies of CO_2 Concentration Based on BPNN for Pleurotus eryngii in a Factory Farm

鉴于杏鲍菇工厂化栽培过程中开关调控滞后的现状,在温湿、光照参数适宜的情况下,以历史CO_2浓度和相对生长时间为输入参数,建立基于BP神经网络的群落式杏鲍菇生长阶段CO_2预测模型,并基于该预测模型提出不同生长阶段CO2浓度调控策略。仿真结果显示,当杏鲍菇处于原基生长阶段时,在CO_2为1 964.3 mg·m^(-3)时密封计时58 min后通风6 min;当子实体处于生长阶段时,在CO_2为4 910.7 mg·m^(-3)时密封计时36 min后通风3.8 min;当子实体成熟时,在CO_2为6 875 mg·m^(-3)时密封计时70 min后通风7.5 min。模型训练集和测试集的相关系数达到0.98,预测精度较高。

Given the status quo that the lag of switch control in the factory process of cultivation of Pleurotus eryngii, according to the experimental CO2 data of P. eryngii under appropriate temperature, humidity and light setting environment in the fall, BP neural network was used to create latest CO2 concentration prediction model with initial CO2 concentration and the relative growth time as the input. In terms of the above model, a control strategy of CO2 in greenhouse control system was put forward. The simulation resuhs show that： when the concentration of CO2 is 1 964.3 mg ·m^-3, it should be seal up rooms for 58 minutes and turn on external ventilation for 8 min in the period of anlage growth; when the concentration of CO2 is 4 910.7 mg·m^-3, it should be seal up rooms for 36 minutes and turn on external ventilation for 3.8 minutes in the early fruiting body; when the concentration of CO2 is 6 875mg·m^-3, it should be seal up rooms for 70 minutes and turn on external ventilation for 7.5 minutes in the mature period of fruiting body. The prediction model created by training group and testing group was accurate both with correlation coefficient of 0.98. Therefore, the prediction model had high accuracy and certain universality.