Energy Efficiency Optimization Scheduling For Stereoscopic Seedbeds Based On Multi-Scale Solar Radiation Rolling Prediction With Machine Learning

The plant factory represented by the multi-layered stereoscopic cultivation mode has the advantages of high production efficiency,short crop growth cycle,excellent quality,and high resource utilization,and is an important way to solve the increasingly severe food production problems.However,the main problems are low automation,difficult intelligent scheduling,and low resource utilization.This paper studies the energy efficiency optimization scheduling method of stereoscopic seedbeds,based on multi-scale solar radiation rolling prediction,with the aim of improving the intelligence of sunlight-utilizing plant factories and reducing their energy consumption.The research contents are as follows.The semi-structured environment of plant factories is difficult to explain with data.This paper studies heterogeneous environmental information,equipment spatio-temporal information acquisition methods and unattended closed-loop monitoring methods.First,an environmental data acquisition sensor network was established,and a miniature multimode gateway was developed.In addition,digital abstraction modeling is performed on the environment,crop state,and equipment,and a database including those information is established on the server.Finally,a remote monitoring platform was developed,which integrated the control function of the environmental regulation actuator.Those above finally realize an unattended closed-loop monitoring system,which include information collection,transmission,analysis,decision-making and control.It is difficult for a sunlight-utilizing plant factory to control the light received by plants precisely.This paper studies the accurate prediction method of hourly solar radiation,establishes solar radiation models including linear components,non-linear components and time-series components.And then,the prediction models,based on GBDT and LSTM,are established for the characteristics of different components.Using multisource data of historical weather,real-time weather forecast and local environment information,we establish a fusion prediction model for smallarea,hourly solar radiation.Finally,we validate the prediction model.Experiments show that the model improves the prediction accuracy of hourly solar radiation in a small area compared to other statistical models and single machine learning model.The traditional fixed stereoscopic cultivation method has the problem of uneven light receiving between shelves,which leads to inconsistent crop growth.This paper studies the dynamic optimal scheduling method of seedbeds based on solar radiation prediction,establishes a light receiving model of multi-layer stereo-seedbeds,implements a dynamic real-time scheduling system for seedbeds.Through experiments,the effectiveness of the dynamic seedbed scheduling method is verified.The results show that compared with fixed scheduling,dynamic optimized scheduling effectively reduces the difference in light reception between different layers of seedbeds.This paper establishes a closed-loop monitoring system for stereoscopic cultivation environment,equipment and crop information in plant factories.We study the real-time rolling prediction method for smallarea,hourly solar radiation,and provide a dynamic optimal scheduling scheme for seedbed,which significantly improves the uniformity of light received by the crop on different layers and optimal utilization of solar energy.