The Research Of The Glass Greenhouse Temperature Model Based On BP Network And The Design Of Monitoring System

In contemporary, facility agriculture has become an impotant sign of agriculturalmodernization； On the one hand,the modern greenhouse is one of the most basic way torealize the technology of facility agriculture； On the other hand, it can provide properenvironment for crops under the various natural conditions. In this way, it will greatly shortenthe growth cycle of crops, and it will achieve the goals that crops can be producted withhigher output, quality and efficiency at same time. In order to achieve maximum economicefficiency, minimal energy consumption and production inputs,it must to control thegreenhouse climate, and change the temperature, humidity, illumination intensity, CO2concentration and other climate factors to get the best conditions for crop growth. Theestablishment of greenhouse microclimate model for meeting the control requirements is thepremise and foundation to achieve optimal control of greenhouse production process, it is ahot issue in the research of greenhouse environment control today, and it is one of mostimportant influence factors on crop growth. However, the current greenhouse temperaturemodel, one is based on the physical mechanism, it has a complex structure, more variables,and many parameters are genrally difficult to determine the values, obtained only byexperience, and it has relatively lower accuracy; The other one is the identification model, theaccuracy of the model is directly dependent on the choice of parameters and the modelstructrue, since the choice of parameters is lack of analysis based on the pysical mechanism,thers is a bigger error of prediction error, input dimensions inccrese and other issues.In this paper, combined with the characteristics of the above two models, through theanalysis of the greenhouse temperature physical model, this paper implemented thescreenning of the main factors that influenced the greenhouse temperature, and then built amultifactorial predictive model of greenhouse temperature based on BP nenural network. Themain contents are shown as the follows:(1)This paper put forward the scheme of greenhouse temperature modeling, the schemedetermined the basis and method of screening factors affecting greenhouse temperatures,designed and built the indoor distributed temperature monitoring system and the outdoor WSN monitoring system for experimental data acquisition. The paper proposed a data dealingmethod and modeling ideas based on BP nenural network, and set the authentication schemeand performance analysis methods of the model finally.(2)The developed temperature monitoring system in greenhouse consists of gatewaymodule, RS485bus communication module, STC12C5A60S2microcontroller temperatureacquisition nodes and1-WIRE BUS temperature acquisition module，it can achieve64different points temperature inside the greenhouse. Moreover, the WSN monitoring systemoutside the greenhouse was designed, it can collect outdoor temperature and sunlight intensity.So both the temperature indoor and outdoor and light intensity which influence temperatureindoor can be achieved through the integration of these two monitoring systems. At last, thecollected date can be send to the upper Web server by the GPRS module.(3)Collected the environmental factors data by environmental monitoring platform,contrasted and analysised formula with relevant papers to determine the empirical values andcalculation methods of clearness index and solar zenith angle, treated the gross errors with3δrule on the data samples of the five types influcing factors which were temperature inside thegreenhouse, temperature outside the greenhouse, sunlight intensity, clearness index and solarzenith angle respectively, and presented a normaliaztion to deal with sample data using thelinear transformation at last.(4)This paper designed the BP network whose structure was5:8:1, built the BP neuralnetwork algorithm via MATLAB, and trained the network with the training set data whichaccounted for75%of the collected data samples of the five types of influcing factor. Theresults showed that the convergence speed was very fast, the model designe was fit for therequirements. Performed simulation by using testing set data which accounted for25%of thecollected data samples, it was found that the simulation results of BP network was very good,and model building process reliability is good too.The validation results of the BP network temperature model established in this papershows that the RMSE of sunny day simulation is0.42℃, the maximum error is1.86℃; theRMSE of cloudy day simulation is0.43℃, the maximum error is2.68℃; the RMSE ofovercast day simulation is0.29℃, the maximum error is1.53℃. The simulated results of thegreenhouse temperature model are quite satisfactory, which are consistent with the actualrequirements. As can be seen, this model will provide the theoretical support and effectivesolution for the implementation of the control system of the greenhouse temperature.