Smart Energy Management for Sustainable Greenhouse Operations: Optimization of Grid Energy and Battery Storage

Effective energy management plays a vital role in improving greenhouse (GH) performance and advancing sustainability. This study introduces an innovative multi-objective optimization strategy specifically developed for GH energy management, aiming to reduce reliance on grid energy while maximizing the battery state of charge (SOC) over a defined time period. The optimization framework incorporates decision variables, including network power, battery power, and battery energy, while respecting constraints on battery capacity, initial energy levels, and the minimum and maximum energy limits of the battery storage system. The performance of a smart energy management system (EMS) is assessed by comparing it against conventional optimization techniques. Critical hyperparameters for the optimization process—such as plateau time, prediction time, and optimization time—are identified using the ellipse optimization technique. By modeling the GH as a microgrid, the study examines relevant energy management metrics and load profiles. Simulations performed in MATLAB Simulink (R2021b) reveal substantial improvements, with the smart EMS reducing the objective function by more than 50% relative to traditional EMS approaches. Furthermore, the proposed EMS demonstrates strong robustness under varying load power and irradiation patterns. Even in partial-shading scenarios, the system remains adaptable, with a maximum increase in the objective function of only 0.35553%. Matching the photovoltaic (PV) system output power to realistic field conditions further confirms the EMS’s practical effectiveness. These results highlight the smart EMS’s capability to optimize energy use in GH settings, paving the way for more sustainable energy management solutions. This work advances energy optimization techniques in agricultural contexts, supporting greater resource efficiency and environmental responsibility.

Background and motivations

The integration of renewable energy sources (RESs), such as photovoltaic (PV) systems, wind power, and battery storage, into small- and medium-scale electricity networks offers multiple benefits, including lower energy costs and improved resilience against outages [1, 2]. Greenhouses represent one promising application area for these RES technologies. Greenhouse farming is essential for ensuring a consistent food supply by maintaining a controlled environment that supports ideal plant development. Nevertheless, the high energy requirements of GH operations present a major obstacle, driving the need for creative strategies to boost energy efficiency. The pursuit of sustainability has encouraged researchers to explore the combination of RESs with sophisticated energy management systems (EMSs) [3]. In this context, the present study contributes by introducing a smart EMS tailored to enhance solar energy utilization within GH facilities.

With global food demand continuing to grow and pressure to reduce agriculture’s ecological footprint intensifying, efforts to improve the sustainability of GH operations have gained momentum. Much of the energy consumed in greenhouses comes from traditional grid sources, which significantly affect both operating costs and greenhouse gas emissions. To address this issue, there is rising interest in employing RESs—particularly solar power—to supply energy for GH operations [4]. Effective use of such intermittent renewable sources requires advanced EMSs capable of matching supply with demand while accounting for the variable energy requirements typical of greenhouse settings. Figure 1 illustrates a schematic diagram of RES integration in a smart GH [5].

Literature review of energy management in GHs 

Earlier research has examined multiple dimensions of energy management within greenhouses, placing particular focus on the incorporation of renewable energy sources (RESs). Prominent studies, including those by Smith et al. [6] and Wang et al. [7], have evaluated the viability of geothermal energy for greenhouse use, underscoring its benefits related to long-term sustainability and economic efficiency. Nonetheless, the combined optimization of energy use via intelligent systems alongside geothermal sources continues to receive limited attention.

Within the field of energy management systems (EMS), conventional optimization methods have long been applied to improve energy performance across diverse applications. However, the adoption of advanced smart EMSs—which integrate machine learning and forecasting techniques—is increasingly recognized for their capacity to respond effectively to fluctuating environmental variables. Research by Li and Wang [8] and Chen et al. [9] demonstrates the strong performance of smart EMS across a range of energy optimization contexts, inspiring the current investigation into their suitability for greenhouse environments.

Greenhouse farming, which plays a vital role in global food security, faces the ongoing challenge of reducing energy use to achieve greater sustainability and cost-effectiveness. The adoption of smart EMSs, combined with targeted optimization techniques, represents a highly promising direction, delivering multiple benefits over conventional methods [5]. These benefits can be categorized as follows:

• Enhancing Resource Utilization and Adaptability: Smart EMSs applied in greenhouse settings enable accurate management of resources by synchronizing energy consumption with real-time requirements. This approach optimizes the use of electricity, heating, and other inputs, thereby reducing waste and lowering operating expenses. A major strength of these intelligent systems is their ability to adapt in real time to changing external conditions. Leveraging live sensor data and predictive models, they dynamically modulate energy allocation according to variables such as temperature, humidity, and solar availability, ensuring efficient performance across diverse scenarios [10, 11].

• Optimized Crop Growth and Yield: Smart EMSs support enhanced plant development by customizing the growing environment to match the precise requirements of different crops. Investigations by Sofwan et al. [12] demonstrate the favorable effects of intelligent control systems on agricultural output, thanks to their ability to finely regulate parameters including temperature, lighting intensity, and CO₂ concentration.

• Reduction in Environmental Footprint: By eliminating wasteful energy practices and prioritizing renewable energy sources, smart EMSs help decrease the overall ecological impact of greenhouse gas activities. Incorporating RESs, such as solar energy, supports broader sustainability objectives and reduces the environmental burden associated with GH production [13, 14].

• Cost Savings and Economic Viability: Intelligent EMSs deliver clear financial benefits through improved energy efficiency and decreased reliance on grid-supplied power. Over extended periods, these reductions in energy expenditure improve the profitability and long-term economic stability of greenhouse enterprises [15, 16].

• Mitigation of Energy Price Volatility: The application of optimization techniques reduces exposure to fluctuations in external energy markets by increasing self-sufficiency. This stability is especially valuable for greenhouse managers, as it facilitates more reliable budgeting and financial forecasting [17].

• Real-Time Monitoring and Control: Smart EMSs provide continuous oversight and immediate adjustment of energy flows, enabling rapid responses to changes in demand or unexpected disruptions. Such responsiveness strengthens system robustness and ensures consistent, dependable operation of greenhouse infrastructure [18].

• Integration of Machine Learning for Predictive Analysis: By embedding machine learning within smart EMSs, it becomes possible to forecast future energy needs based on patterns derived from past performance and environmental trends. This forward-looking capability enhances the planning of energy allocation and storage strategies inside the greenhouse [19].

• Improved Operational Efficiency: The detailed regulation offered by smart systems leads to higher overall operational performance. Automating demanding tasks, such as climate control, reduces the need for manual oversight, freeing greenhouse personnel to focus on higher-level planning and crop care [10, 20].

• Technological Innovation and Industry Leadership: Operators who implement smart EMSs establish themselves at the forefront of agricultural technology. By adopting state-of-the-art tools, they help drive progress toward more sustainable farming methods and establish exemplary industry standards [21].

In light of the challenges and opportunities outlined above, this paper addresses the following four key research questions:

• How can solar energy integration be optimized to decrease reliance on traditional electricity grids and improve the environmental sustainability of greenhouse operations?

• What are the primary benefits and possible limitations of deploying a smart EMS for greenhouse energy optimization, especially when compared with conventional optimization techniques?

• In what ways can the careful selection and tuning of critical hyperparameters—such as plateau time, prediction time, and optimization time throughout the day—enhance the flexibility and effectiveness of a smart EMS in managing greenhouse energy?

• Which concrete results or measurable indicators demonstrate the superior energy efficiency and operational improvements delivered by the proposed multi-objective optimization framework designed specifically for greenhouse energy management?

To address these questions, the present study fills an important gap in the current body of knowledge by introducing a multi-objective optimization framework specifically designed for greenhouse energy management. The approach integrates solar energy utilization with a smart EMS. The sections that follow describe the proposed methodology, report simulation outcomes, and provide a thorough evaluation of the system’s advantages relative to standard optimization methods.

Motivation and main contributions 

The primary motivations behind this paper stem from several critical aspects: 

• Multi-Objective Optimization for GH Energy Management: The study proposes a novel multi-objective optimization framework designed to simultaneously reduce grid energy consumption and maximize the battery state of charge (SOC) over a given time horizon. This method promotes more thorough and effective management of energy resources in greenhouse operations. 

• Integration of Solar Energy: A key contribution of this work is the incorporation of solar energy as the main renewable input within the developed EMS. By leveraging the reliability and environmental friendliness of solar power, the proposed system aims to reduce reliance on traditional electricity grids and thereby enhance the sustainability profile of greenhouse operations. 

• Smart EMS: This paper presents and evaluates a smart energy management system (EMS) tailored for optimizing energy use in greenhouses. The system employs sophisticated techniques, such as machine learning and predictive modeling, to dynamically regulate energy consumption in response to changing environmental factors. This distinguishes it from conventional optimization strategies and paves the way for more adaptive and intelligent energy management solutions. 

• Identification and Optimization of Hyperparameters: The research identifies and fine-tunes essential hyperparameters that are vital to the performance of the proposed EMS. Key parameters—including plateau time, prediction time, and optimization time throughout the day—are optimized using the ellipse optimization method. This tuning process significantly boosts the system’s flexibility and operational efficiency across diverse conditions. 

In addition, the main contributions of this paper encompass the following: 

• Simplifying the objective function and limiting decision variables to those that are independent. 

• Defining the relevant hyperparameters and identifying their optimal values. 

• Incorporating uncertainty in the forecasted values of photovoltaic power (Ppv) and load power. 

• Taking into account environmental influences, including partial shading effects. 

• Continuously updating the advanced predictions of Ppv throughout the day. 

Overall, this paper addresses multiple real-world dimensions of the energy management challenge to deliver a solution that is practical, high-performing, and feasible for actual implementation. Through a comprehensive strategy that combines multi-objective optimization, solar energy utilization, advanced smart EMS with cutting-edge algorithms, and careful tuning of critical hyperparameters, the work seeks to advance integrated, sustainable approaches to greenhouse energy management while achieving maximum energy efficiency.

Efficient energy management in greenhouse (GH) operations is fundamental to contemporary agriculture and long-term environmental sustainability. As the world’s population steadily increases, food demand continues to grow, underscoring the essential role of greenhouse farming in global food security [22]. Nevertheless, the traditional high-energy practices employed in GHs create substantial economic and ecological difficulties. 

Proper regulation of energy use is essential for reducing operational expenses and lessening the environmental burden associated with conventional electricity sources. GH managers face the ongoing challenge of balancing the requirements for ideal plant development with the need to curb energy-related costs. The growing need to adopt more sustainable methods has spurred considerable research into integrating renewable energy sources (RESs) into greenhouse systems [23]. 

This study addresses that need by introducing a smart energy management system (EMS) specifically designed for greenhouse settings. Central to this approach is the utilization of solar energy, valued for its consistent availability. Through a multi-objective optimization framework, the paper aims to reduce dependence on external power grids, improve battery storage performance, and advance broader environmental sustainability in greenhouse operations. The value of this research lies in its capacity to transform energy-use patterns in GHs, supporting more cost-effective and environmentally sound agricultural practices. 

In the sections that follow, we outline an optimized strategy to tackle these challenges. 

Throughout this article, the terms “solar energy” and “PV” are employed. “Solar energy” denotes energy obtained from the sun in a broad sense, whereas “PV” refers specifically to the photovoltaic technology that transforms sunlight into electrical power. In the context of our work, these terms are used interchangeably to encompass both the general notion of solar energy and the particular application of PV systems. 

Objective function for multi-objective optimization 

To deliver an effective energy optimization solution for the greenhouse, the core objective of this study is to minimize the cost of grid-supplied power while simultaneously maximizing the battery state of charge (SOC) at the end of the day. This dual focus reflects the combined aims of reducing reliance on external electricity supplies and enhancing battery storage utilization within the GH environment. The battery SOC is calculated according to Eq. 1 [24]. 

SOC_batt, the battery state of charge, is expressed as the ratio of the current battery energy, Ebatt, to the battery’s total capacity, C_batt. This formulation ensures equilibrium between stored energy and the battery’s overall capacity. Battery energy and power are further governed by constraints expressed in Eq. 2: 

Here, E_min and E_max indicate the minimum and maximum allowable battery energy levels, respectively. Similarly, P_min and P_max represent the minimum and maximum permissible battery power limits, respectively. 

These constraints ensure that both the battery's energy level and power output remain within the acceptable operational boundaries defined by its design specifications. The formulation accounts for the battery’s initial energy state and the inherent limitations of the storage system.

 Energy Management System (EMS)  

EMSs operate at different hierarchical levels depending on the scale of application, ranging from residential buildings and commercial facilities to industrial plants and entire districts. As illustrated in Table 1, these systems can interconnect models representing EMSs across multiple tiers and varying degrees of sophistication [25]. 

An EMS constitutes an integrated framework responsible for monitoring, regulating, and optimizing both energy consumption and generation across diverse settings. It integrates hardware components, software tools, and operational procedures to deploy energy assets while preserving required performance standards efficiently. The central purpose is to reduce energy expenditures, minimize adverse environmental effects, and enhance overall system effectiveness [26, 27]. EMSs are broadly categorized into traditional and smart types, with increasing preference for intelligent solutions as electrification expands. Conventional EMSs depend on fixed rules and operate without access to live data, whereas the shift toward smart EMSs emphasizes adaptive strategies. These include demand forecasting, dynamic electricity tariffs, optimization-driven techniques, and real-time measurements to enable fully automated decision-making [28]. Such advanced practices are already commonplace in large-scale utility operations, where generation schedules are adjusted according to load patterns and market prices. This same principle can be readily scaled down to smaller systems, such as microgrids, and extended to a wide range of applications. The fundamental aim of smart EMSs is to maintain ongoing surveillance of energy demand and system responses, thereby continuously refining overall performance [29]. Through this iterative optimization loop, multiple benefits can be realized, including energy conservation, emission reductions, and cost minimization. Figure 2 presents an overview of different EMS methodologies, with the traditional approach shown on the left. 

The figure shows the load power profile (dotted line) alongside the PV power input. The load typically peaks in the afternoon and evening, coinciding with periods of residential occupancy. During peak solar irradiance—especially on clear days—the PV contribution is maximized during periods of elevated demand. The shaded areas highlight the behavior of the conventional heuristic-based management strategy. In the early morning, grid power supplies the base load. When demand is low, and solar generation becomes available, PV energy is used directly to meet needs while excess is directed to charge the storage system. At a designated threshold, surplus energy is exported to the grid. Later in the evening, the stored energy is discharged to offset demand. The smart strategy, shown on the right, additionally factors in electricity pricing. For example, during low-price intervals, grid power may be intentionally used to charge storage, with discharge and potential grid export occurring when prices rise. 

In the present study, the focus is placed on optimizing an EMS configured for a microgrid architecture. This microgrid integrates PV arrays, battery energy storage, several residential loads, and a connection to the main utility grid. It is assumed that the underlying power electronic converters and interfaces are already in place, allowing the EMS to operate at a supervisory level with a system-wide perspective. Optimization is achieved through a decision-making algorithm that leverages predictive information and forecasts to schedule charging and discharging actions intelligently, determine optimal timing for grid interaction, and maximize overall energy efficiency. The approach respects all relevant operational constraints and strives to deliver the most effective coordinated strategy for utilizing the microgrid’s diverse components. 

The high-level control logic forms the backbone of the EMS, coordinating the interplay among all elements to ensure smooth, efficient, and optimized microgrid performance [31]. To further strengthen the optimization process, this work adopts a smart EMS formulation to achieve superior efficiency. The core objective remains minimizing grid electricity costs while simultaneously maximizing the battery’s final state of charge (SOC), as expressed in Eq. 3.

Where P_pv, P_batt, and P_load represent the power delivered by the PV array, the battery, and the load, respectively, and wFinal is a weighting coefficient. In addition, several constraints must be satisfied, one of the most important being the power balance condition expressed as: 

The decision variables must be mutually independent, meaning the value assigned to one variable must not influence the value of another. Keeping the number of decision variables as small as possible helps reduce problem complexity. Including dependent variables compromises the accuracy of the EMS algorithm, whereas incorporating unnecessary dependencies increases both computational load and execution time. Notably, E_batt and P_batt are interdependent variables. Meanwhile, P_grid can be directly derived from Eq. 4. Given that fixed profiles for Pload and PPV are assumed at each optimization step, the objective function f_obj can be formulated accordingly. 

The value of 𝑃𝑏𝑎𝑡𝑡 that minimizes 𝑓𝑜𝑏𝑗 will also minimize the following expression. 

Since energy corresponds to the time integral of power (i.e., the accumulated power over a defined interval), the battery energy can be expressed accordingly. 

By combining Eq. 1 and 7, the modified objective function fnew can be rewritten as follows. 

Thus, the P_batt value that minimizes f_obj also minimizes f_new as defined in Eq. 8. After eliminating the constant term, the expression simplifies to: 

The P_batt minimizing f_obj will likewise minimize F_New as given in Eq. 9. Consequently, the optimization problem can be stated in the following form: 

The optimization procedure determines the decision variables over the prediction horizon Tpred, assuming they remain constant throughout the plateau duration Tplateau. The entire optimization routine is repeated at regular intervals defined by the optimization time Topt. The key hyperparameters governing this process are Tpred, Tplateau, and Topt. 

Owing to the high dimensionality of the problem, certain numerical techniques—such as Bayesian optimization—involve considerable computational expense when solving the optimization problem Eq. 10 to compute the decision variable Pbatt at every time step. To address this, the following MATLAB functions are applied to optimize the decision variables, yielding virtually equivalent results: 

- “fmincon”: employed to locate the minimum of a constrained nonlinear multivariable function. 

- “fminimax”: used to resolve minimax constrained optimization problems. 

- “solve”: applied to tackle general optimization or equation-solving tasks. 

Tpred is treated as a variable-length variable and set equal to the remaining hours of the day, since grid electricity prices peak during the final hours of the day. It follows that Tpred < Topt < Tplateau.

The Bayesian optimization technique is applied to determine the optimal values of Tplateau and Topt using the “bayesopt” function in MATLAB. This method represents a robust numerical approach to optimization, enabling efficient exploration of the search space by prioritizing the most promising regions while accounting for uncertainty and incorporating prior information.

To evaluate the performance of the proposed intelligent strategy, the conventional EMS algorithm is illustrated in Figure 3. Similarly, Figure 4 depicts the operational workflow of the smart EMS. Over the course of 24 hours, the optimization procedure is executed at intervals of Topt. At each optimization instant (ti), the objective is to minimize the grid power cost across the subsequent prediction interval Tpred. The resulting optimal battery power value is then applied at ti and held constant until the next optimization step, during which the battery state of charge (SOC) is updated. For every optimization cycle, the SOC value at the current instant ti is known and used as the starting point.

To demonstrate the outcomes of the simulations, we present and compare two distinct operational modes: the smart energy management approach and the conventional energy management approach. Figure 5 shows the implementation of the EMS strategies for greenhouses within the Simulink environment. As previously noted, by incorporating relevant energy management metrics and load characteristics, the greenhouse can be effectively modeled as a microgrid. A simple microgrid configuration has been developed in Simulink for this purpose. This model allows observation of system behavior under both the traditional rule-based EMS and the optimization-driven smart EMS, facilitating a direct performance comparison between the two methods. The model includes a variable load representing residential households, a constant base load that remains unchanged over time, and a photovoltaic array serving as the renewable input source, capable of adopting different irradiance profiles (corresponding to clear-sky and overcast conditions), as shown in Figure 6.

Figure 7 presents the load profile used in the case study. In contrast, Figure 8 shows how grid electricity costs vary with different settings of the prediction horizon Tpred and the optimization interval Topt. These results highlight the critical importance of tuning these hyperparameters. It is clear that prolonging the prediction horizon significantly increases the computational demand. Nevertheless, since electricity prices typically peak in the final hours of the day, it is essential to include this high-cost period in Tpred to maintain prediction accuracy. Consequently, the prediction horizon Tpred is treated as a dynamic parameter, set equal to the remaining duration of the day. As a result, Tpred reaches its maximum early in the morning and gradually decreases throughout the day.

Figure 9 provides a detailed overview of key parameters in the microgrid system, including microgrid voltage, PV power output, battery state of charge, grid power exchange, and load demand.

Figure 9 illustrates the battery state of charge (SOC) and the associated grid power costs throughout the day across various load profiles when employing the smart EMS. By applying two distinct traditional intelligent strategies, a clear comparison can be drawn with the smart EMS's performance. For ease of evaluation, Table 2 summarizes the key numerical simulation outcomes for both the conventional and smart EMS approaches. The results reveal a striking improvement: the smart EMS reduces the objective function value by more than 50%. This substantial decrease demonstrates the smart EMS's enhanced ability to lower grid electricity costs while simultaneously maximizing the battery SOC by the end of the day. The comparative analysis confirms the effectiveness of the smart EMS in optimizing energy use within greenhouse (GH) settings. It emphasizes the system’s capacity to dynamically adjust to fluctuating load demands, changing environmental factors, and varying operational needs, thereby improving overall energy efficiency and resource management in the greenhouse context. These outcomes reinforce the importance and promise of the smart EMS methodology in tackling the multifaceted challenges of energy management in agricultural applications. They lay a strong foundation for continued research and real-world deployment of sustainable energy strategies in greenhouse environments, advancing more resource-efficient and environmentally responsible farming practices.

 To better reflect realistic operating conditions, a ± 10% tolerance has been applied to both the load power and irradiance profiles, as depicted in Figure 10. The corresponding numerical outcomes are presented in Table 3, with Tplateau and Topt both set to 5 minutes. Despite these intentional variations in load power and solar irradiance, the objective function value remains notably stable. This stability demonstrates the robust, reliable performance of the proposed EMS under realistic disturbances.

When the simulation is repeated with a partial shading event lasting 30 minutes beginning at 11 a.m.—as shown in Figure 11—the results are summarized in Table 4. The findings indicate that partial shading causes only a modest increase of 0.35553% in the objective function. This minimal rise underscores the proposed EMS's strong adaptability and resilience in responding effectively to natural phenomena, such as partial shading.

As anticipated, the drop in PV power output during shading leads to greater reliance on grid power and increased battery discharge to maintain power balance. Figure 12 provides a side-by-side comparison of PV power, battery power, grid power, load power, and battery SOC, both with and without partial shading, specifically for Load 2.

Initial forecasts of irradiance and temperature are used to estimate photovoltaic power (Ppv). Recognizing that forecasts are not always accurate, three distinct scenarios are examined as follows: 

• Case 1: Ppv prediction is nearly accurate. The initial Ppv forecast is used throughout the optimization. 

• Case 2: Ppv prediction is inaccurate. The initial (incorrect) Ppv forecast is used for optimization. 

• Case 3: Ppv prediction is inaccurate. An updated Ppv forecast is employed during optimization. 

It is assumed that the PV power forecast is revised at 11 a.m., with the updated value reduced to 70% of the original prediction. The numerical simulation results for these cases across various optimization time steps are compared in Table 5. The corresponding battery power profiles for Cases 1, 2, and 3 are shown in Figures 13–15. These results highlight the critical role of refreshing the Ppv forecast in shaping the objective function value. Timely updates are essential for aligning the model with actual conditions and achieving superior performance. Moreover, the findings demonstrate the substantial influence of the chosen optimization time step on both the objective function value and computational runtime.

Longer optimization time steps lead to shorter execution times. However, as shown in Table 5, no uniform or predictable relationship exists between the optimization time step and the objective function (fobj). While Case 1 shows that increasing the time step reduces fobj, this pattern does not hold for Cases 2 and 3. Notably, the lowest fobj value occurs at an optimization time step of 180 minutes, reinforcing the importance of carefully tuning the optimization interval.

 The optimization procedure determines the decision variables over the prediction horizon Tpred by employing MATLAB functions such as “solve,” assuming that these variables remain unchanged throughout the plateau duration Tplateau. The entire optimization routine is executed repeatedly at intervals defined by the optimization time Topt. Tpred is treated as a dynamic parameter corresponding to the remaining hours of the day. The hyperparameters Tplateau and Topt are fine-tuned using the MATLAB “bayesopt” function. 

Simulation outcomes reveal a marked improvement, with the smart EMS yielding a reduction of more than 50% in the objective function compared to the conventional EMS. This considerable gain confirms the effectiveness of the proposed method in reducing grid electricity costs while simultaneously maximizing the battery state of charge (SOC) at the end of the day. Moreover, the smart EMS exhibits strong robustness, delivering stable performance despite fluctuations in load power and solar irradiance, demonstrating its resilience against varying environmental conditions. 

Furthermore, when partial shading conditions are introduced, the objective function experiences only a minor rise of 0.35553%. This limited impact illustrates the smart EMS’s high adaptability to natural disturbances, a critical attribute for maintaining dependable energy management under unpredictable conditions. 

In addition, the simulation findings highlight the substantial influence of refreshing PV power forecasts on the objective function value. Aligning predictions with actual conditions is essential for achieving optimal outcomes. Likewise, the careful selection of the optimization time step is shown to affect both the resulting objective function value and the computational runtime, underscoring the importance of precise hyperparameter tuning. 

Table 6 presents a comparative analysis between the present study and three closely related publications in the domain of greenhouse energy optimization and management. This side-by-side evaluation highlights the distinct strengths of the current work, particularly the substantial reduction in the objective function achieved through the smart EMS and its superior adaptability to environmental variations. 

Overall, the simulation results strongly support the effectiveness of the proposed smart EMS in enhancing energy utilization within greenhouse settings. Its proven robustness, flexibility in the face of environmental changes, and sensitivity to parameter tuning collectively affirm the practical value of the smart EMS for real-world deployment, paving the way for more efficient and environmentally sustainable energy practices in agriculture. 

In summary, this study addresses a critical aspect of energy management in agricultural greenhouses (GHs) by presenting an integrated solution that combines renewable energy sources (RESs) with an advanced smart energy management system (EMS). As global food demand continues to rise, the transition to more sustainable, resource-efficient greenhouse farming has become increasingly urgent. The developed multi-objective optimization framework—focused on simultaneously minimizing grid energy usage and maximizing battery state of charge (SOC)—marks a meaningful step forward. By prioritizing solar energy as the primary renewable energy source and implementing a smart EMS, the research effectively addresses both the economic burden of high operational costs and the ecological consequences of reliance on conventional power sources. 

Beyond its theoretical contributions, this work delivers practical advantages for greenhouse operators and supports broader environmental goals. The key hyperparameters, refined through the ellipse optimization technique, significantly enhance the smart EMS’s adaptability and overall performance. Comparisons against conventional optimization methods further validate the superior efficiency of the proposed smart EMS, reinforcing its capacity to transform energy management practices in greenhouse agriculture. 

As the agriculture sector confronts the pressing need for sustainability, the findings of this study encourage the wider adoption of innovative, environmentally friendly technologies, laying the foundation for more resilient and efficient greenhouse operations in the future. 

Potential directions for future work include exploring more sophisticated optimization algorithms to enhance the smart EMS’s capabilities further. Expanding the system to incorporate additional renewable sources—such as wind or geothermal energy—also warrants investigation. Moreover, conducting field trials in real-world greenhouses to verify the proposed approach's performance under actual greenhouse conditions would provide valuable insights. Such efforts would build upon existing initiatives to refine energy management strategies and promote greater sustainability across the agricultural sector.

The natural language artificial intelligence, ChatGPT (Version 3.5, OpenAI), was employed in this process. This AI tool, renowned for its proficiency in language tasks, has been instrumental in meticulously polishing the manuscript's language.

None

This project was funded by the Mitacs Accelerate Program, in partnership with the Petroleum Technology Research Centre (PTRC) in Regina, SK, Canada, project ID: IT29866.

None

Samuel Boateng & Kwesi Mensah contributed to this work.

Authors and affiliations

Department of Smart Power Networks, Faculty of Engineering, University of Cape Coast, Cape Coast, Ghana
Samuel Boateng & Kwesi Mensah

Corresponding author

Correspondence to Samuel Boateng

Open Access The author(s) retain copyright. This article is licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. It may be shared and adapted for non-commercial purposes with appropriate attribution, an indication of changes, and distribution of adaptations under the same license. Third-party material may be subject to separate terms identified in its credit line. View the license at https://creativecommons.org/licenses/by-nc-sa/4.0/.