Energy management, optimization, and control are significant subjects that have garnered considerable attention across agricultural and industrial sectors. The relevance of productivity in agricultural output and effective energy management within greenhouse systems has become increasingly evident as global population growth intensifies the demand for food resources. In response to these challenges, this thesis develops and critically evaluates three distinct yet complementary control frameworks for smart greenhouse energy management situated within the broader contexts of smart farming and power grid integration. The first proposed framework introduces a centralized energy management platform based on a dynamic nonlinear quantum particle swarm optimization algorithm embedded within a predictive control structure. This platform orchestrates renewable generators, HVAC systems, energy storage units, water reservoirs, CO$_2$ injectors, dehumidifiers, and smart metering infrastructure to regulate indoor microclimate variables while actively minimizing power exchange with the external grid during periods of low renewable generation. Comparative performance validation is conducted against classical particle swarm optimization and the Pyomo optimization environment solved with IPOPT. In a subsequent contribution, a nonlinear model predictive control methodology is formulated for comparable greenhouse assets, with particular emphasis on adaptability across a comprehensive crop growth model that considers all plant organs. This model dynamically computes dry matter accumulation partitioned among leaves, stems, roots, fruits, and mature fruits. The control strategy prioritizes accurate regulation of temperature, relative humidity, and CO$_2$ concentration trajectories while reducing overall energy consumption. The fourth section of this thesis expands the scope toward interconnected net zero-energy farm microgrids, wherein smart greenhouses operate alongside residential zones and animal housing facilities. A distributed optimization algorithm grounded in the alternating direction method of multipliers is developed to coordinate power exchanges among individual farms as well as between the farm cluster and the main utility grid. This architecture maximizes localized renewable energy utilization while preserving desirable microclimates across all building typologies. Collectively, these contributions provide a progressive trajectory from centralized optimization to distributed coordination and hybrid renewable scheduling, offering practical and theoretical insights for scalable, energy-efficient, climate-resilient greenhouse systems and power grid integration. The final component of this thesis addresses hybrid renewable energy systems combining wind farms (WF) with pumped hydroelectric energy storage (PHES) for peak load reduction. A finite horizon scheduling optimization problem is formulated to control real-time WF-PHES operation using predictions of wind speed and power load. A three-layered ANN forecasts these variables from historical data, while a mathematical decision model accounts for seasonal (summer/winter) load-shaving requirements. The proposed framework is validated on a case study and compared against Pyomo with IPOPT and a Genetic Algorithm (GA), demonstrating its effectiveness through extensive numerical simulations.
Distributed Learning-Assisted Model Predictive Control (MPC) and Intelligent Energy Management for Interconnected Smart Greenhouses Integrated with Microgrids
ZAHMOUN, SAID
2026-06-22
Abstract
Energy management, optimization, and control are significant subjects that have garnered considerable attention across agricultural and industrial sectors. The relevance of productivity in agricultural output and effective energy management within greenhouse systems has become increasingly evident as global population growth intensifies the demand for food resources. In response to these challenges, this thesis develops and critically evaluates three distinct yet complementary control frameworks for smart greenhouse energy management situated within the broader contexts of smart farming and power grid integration. The first proposed framework introduces a centralized energy management platform based on a dynamic nonlinear quantum particle swarm optimization algorithm embedded within a predictive control structure. This platform orchestrates renewable generators, HVAC systems, energy storage units, water reservoirs, CO$_2$ injectors, dehumidifiers, and smart metering infrastructure to regulate indoor microclimate variables while actively minimizing power exchange with the external grid during periods of low renewable generation. Comparative performance validation is conducted against classical particle swarm optimization and the Pyomo optimization environment solved with IPOPT. In a subsequent contribution, a nonlinear model predictive control methodology is formulated for comparable greenhouse assets, with particular emphasis on adaptability across a comprehensive crop growth model that considers all plant organs. This model dynamically computes dry matter accumulation partitioned among leaves, stems, roots, fruits, and mature fruits. The control strategy prioritizes accurate regulation of temperature, relative humidity, and CO$_2$ concentration trajectories while reducing overall energy consumption. The fourth section of this thesis expands the scope toward interconnected net zero-energy farm microgrids, wherein smart greenhouses operate alongside residential zones and animal housing facilities. A distributed optimization algorithm grounded in the alternating direction method of multipliers is developed to coordinate power exchanges among individual farms as well as between the farm cluster and the main utility grid. This architecture maximizes localized renewable energy utilization while preserving desirable microclimates across all building typologies. Collectively, these contributions provide a progressive trajectory from centralized optimization to distributed coordination and hybrid renewable scheduling, offering practical and theoretical insights for scalable, energy-efficient, climate-resilient greenhouse systems and power grid integration. The final component of this thesis addresses hybrid renewable energy systems combining wind farms (WF) with pumped hydroelectric energy storage (PHES) for peak load reduction. A finite horizon scheduling optimization problem is formulated to control real-time WF-PHES operation using predictions of wind speed and power load. A three-layered ANN forecasts these variables from historical data, while a mathematical decision model accounts for seasonal (summer/winter) load-shaving requirements. The proposed framework is validated on a case study and compared against Pyomo with IPOPT and a Genetic Algorithm (GA), demonstrating its effectiveness through extensive numerical simulations.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
