Research And System Design Of Intelligent Control Model For The Internal Environment Of Modular Plant Factories

In the face of the intricate challenges posed by climate variability,agricultural land degra-dation,rapid urbanization,an aging workforce,and the mounting demand for safe and nu-tritious sustenance,the imperative to cultivate purer and safer agricultural commodities has acquired heightened urgency.To confront these challenges,contemporary agriculture is un-dergoing a transformative paradigm shift toward facility-based,resource-efficient production technologies.Distributed and modular plant factories have emerged as pivotal conduits and developmental trajectories to augment agricultural production efficiency and curtail energy expenditure.By amalgamating the functionalities of conventional large-scale plant factories into compact,scalable modular systems,modular plant factories offer enhanced flexibility,scalability,and applicability.However,it is noteworthy that this technology is presently in the research and experimental phase.To elevate the levels of automation and intelligence,immedi-ate attention must be directed toward the research and refinement of environmental modeling,control methodologies,and equipment systems.The widespread adoption and implementation of this technology will not only foster the sustainability of agricultural production but also en-gender broader and more efficient production practices and resource utilization.The principal focal points of this thesis are as follows:This dissertation harnesses modularized apparatus within a controlled environment to acquire salient environmental parameters and device metrics throughout the developmental phases of Pleurotus ostreatus.In response to extraordinary data instances,a data preprocess-ing methodology is posited,entwining the mobile median approach and Akima cubic spline interpolation.Comparative empirical findings manifest a conspicuous augmentation in model precision.Subsequent to the preprocessing regimen,the resultant dataset evinces a reduction of 26.05%,49.13%,and 53.01%in root mean square error for indoor temperature,humidity,and CO₂concentration,respectively.Additionally,the mean absolute errors exhibit a decline of 52.91%,52.03%,and 66.64%for the aforementioned variables.To surmount challenges associated with high-dimensional modeling data and overfitting,a feature selection experi-ment employing the random forest algorithm is undertaken.This experiment accomplishes commendable reduction in data dimensionality for the three output variables of temperature,humidity,and CO₂concentration,thus effectively mitigating model inaccuracies.In view of the intricate interplay of environmental factors within modular plant factories,the relationship between the physiological respiration of edible fungi and the complex oper-ational mechanisms,as well as the formidable task of monitoring certain elusive parameters,we propose a modular plant factory multi-environmental variable prediction model grounded in non-linear autoregressive neural networks（NARX）.Through meticulous comparative ex-perimental analysis,the results highlight the exceptional performance of this predictive model in accurately forecasting indoor temperature,humidity,and CO₂concentration.Notably,the model outperforms three alternative prediction models,namely Light GBM,XGBoost,and LSSVM,exhibiting significantly superior average absolute error and root mean square error.This model empowers precise anticipation of forthcoming fluctuations in the internal envi-ronment of modular plant factories.Moreover,when considering the root mean square errors for indoor temperature,humidity,and CO₂concentration over the next 20 minutes,they are recorded as 0.0993,3.4441,and 88.001,respectively.These findings lay a robust theoreti-cal groundwork for achieving intelligent control and effective production management in this domain.Considering monitoring precision and cost-effectiveness,an intelligent environmental pa-rameter collection and internal environment control system for modular plant factories is de-vised,with the development of hardware architecture and the completion of software design tasks.The dissertation delineates the software and hardware design procedures for the in-telligent internal control system,encompassing the hardware circuit design for the primary controller,environmental acquisition terminals,and equipment control terminals,equipment layout wiring,control system software design,and the implementation of model predictive control algorithms.An astute cloud-based monitoring system is established on the One NET platform,with the design of pertinent application interfaces.Ultimately,subsequent to actual testing,the developed system successfully accomplishes real-time data monitoring and smart control of equipment for modular plant factories,furnishing dependable technical support for the tangible production of Pleurotus ostreatus.