3.1. Power Management Algorithm

In the study of power management for energy-harvesting systems, a nuanced approach is essential for optimizing battery life and energy efficiency. The core objective of this algorithm is twofold: first, to harness energy from the most potent source available and second, to enhance battery longevity through specialized charging strategies. These objectives align with the overarching goals of the study, which includes the evaluation of such algorithms [20].

Implementing these SoC boundaries necessitates the use of a battery with marginally higher capacity, but this is a tradeoff for extended battery life.

For a more granular understanding, the algorithm’s process flow is delineated in flow charts, shown in Figure 3. The implemented algorithm, as encoded into the microcontroller, begins by assessing if the SoC is below a minimum threshold. If so, all energy harvesters are activated, and the microcontroller enters a sleep state, awaiting an interrupt signal, such as a timer set to wake the system every 15 min.

If the SoC is at an acceptable level, the system analyzes the power from each EH source. Priority is given to measuring the power from the source currently in use for its efficiency, avoiding the need for the MPPT circuit to adjust. This circuit, utilizing the fixed open circuit voltage (FOCV) approach, can take up to 16 s in extreme cases to locate the most efficient power point.

To counter potential inaccuracies in power readings during the MPPT’s sampling phase, a protective mechanism is in place. Should the initial power reading be zero, indicating an error, a follow-up reading is taken after a short interval [21]. The algorithm then selects the higher of these two values to ensure a precise power measurement.

After analyzing the output from each energy harvester, the algorithm checks if the SoC is overly high. If it is within acceptable limits, it activates the harvester with the greatest output [22]. Subsequently, the system dispatches a LoRa message detailing the harvested power, battery voltage, and the energy source in use, before entering sleep mode.

An enhanced algorithm is also proposed, which is more versatile and can accommodate multiple harvesters. This advanced algorithm has not yet been implemented in physical hardware. In this version, the algorithm iteratively reads power from each source, adjusting the reading order based on the active harvester, and ensuring that all sources are evaluated.

This work presents a novel integrated energy management algorithm specifically designed for multi-input systems that harness solar, wind, and rain energies. This method seeks to maximize the utilization of various energy sources in order to improve the dependability of the system and prolong the lifespan of the battery. It addresses the real-world requirements of energy management by using recent optimization theories. The process begins by configuring system characteristics, such as battery state and energy availability, to provide a foundation for operation. It continually analyzes the real-time power outputs from different sources and adjusts energy inputs depending on predetermined criteria to optimum efficiency. The algorithm’s main function is to assess and choose the most suitable energy combination by considering power stability and output. This, in turn, enhances system performance and extends battery lifespan. Simultaneously, it regulates battery charging and discharging by employing tactics that are tailored to the present battery condition and chosen energy sources, while also implementing measures to prevent overcharging and excessive discharge. The algorithm’s response to changes in system conditions and external elements is ensured by frequently updating the system statuses. This technique not only emphasizes the algorithm’s capacity to adjust and endure, but also greatly improves the overall efficiency of the system. Future study will aim to incorporate sophisticated predictive models and machine learning approaches to enhance the reliability and efficiency of the algorithm in different circumstances, hence, bolstering the sustainable functioning of low-power devices in various settings.

3.2. System Configuration and Schematic Overview

The EH unit, which is comprised of solar panels and wind generators, is essential for the conversion of solar and wind energy into electrical power in the integrated energy management system that is illustrated, shown in Figure 4. This unit is capable of operating in a variety of climatic conditions. During periods of insufficient sunlight, the wind turbines supplement power, while the solar panels efficiently capture sunlight. The environmental parameters are continually monitored by the data acquisition and control unit, which is outfitted with transducers. The microcontroller-driven unit intelligently determines the most suitable energy source by analyzing real-time data and battery status, thereby efficiently administering the charging and discharging processes of the battery. The battery management device, which is essential for the storage of the harnessed energy, utilizes sophisticated management algorithms to prevent overcharging and complete discharge, thereby significantly extending the battery’s lifespan. Finally, the output and communication units not only convert the stored energy into the necessary forms for other systems, but also enable data exchange and remote control through a communication module that is connected to a central monitoring system. This system architecture not only improves the efficacy and stability of energy utilization, but also considerably increases the system’s overall sustainability and reliability. A robust solution for the sustainable operation of low-power wide-area network devices is provided by such a comprehensive and optimized design, particularly in scenarios where environmental conditions are highly variable or resources are limited.

3.3. Principle of Operation and Efficiency of DC–DC Converters

Here, R_s represents the internal equivalent series resistance and R_load is the load resistance.

3.4. Dynamic Response and Stability

Here, τ represents the system’s time constant, reflecting the response speed; K is the gain, indicating the degree of impact of input changes on the output.

4.1. Simulation of the EH System

The simulation, crafted in Simulink, mirrors the hardware configuration, creating a realistic representation of the EH scenario [24]. It integrates various inputs, including solar irradiance, wind velocity, and rainfall intensity. A user-friendly interface permits interaction, allowing users to manipulate energy harvester inputs and storage options. In the solar cell module of the simulation, solar irradiance data is utilized to compute electrical power, leveraging a cell efficiency look-up table. The wind power module follows a similar approach, using wind speed to determine electrical output and referencing a turbine efficiency table. The rain harvester module, albeit less detailed due to data limitations, estimates kinetic energy from rainfall and its conversion to electrical power.

The simulation’s power management algorithm is streamlined, choosing the energy source with the highest power output at each interval. This approach, more efficient than its hardware counterpart, bypasses the delay inherent in reading power from multiple sources.

The simulation incorporates two distinct MPPT strategies: the perturb and observe (P&O) technique and the FOCV method. Special emphasis is placed on FOCV, aligning with its practical application in the real-world system. Each algorithm is engineered to achieve maximum efficiency within specific operational parameters.

DC–DC converters, pivotal in EH, are represented with varying efficiency levels determined by their design and power output. The simulation employs efficiency look-up tables derived from data on specific EH integrated circuits, shown in Figure 5.

4.2. Hardware Configuration

The physical aspect of the study involves an EH system enclosed in a 3D printed case, outfitted with external solar panels and wind turbines, shown in Figure 6.

At the heart of this system is a Microchip SAM R34 microcontroller, responsible for computation and control. Optimization techniques were applied to minimize power consumption, thereby boosting the system’s overall energy efficiency.

Power monitoring is executed via a combined voltage and current sensor. This setup enables accurate tracking of the system’s energy consumption and output. The selection of solar cells and wind turbines is based on their suitability for EH, diversifying the energy sources utilized in the system.

The management of MPPT and battery charging is entrusted to a specialized integrated circuit, chosen for its congruence with the system’s design requirements. A lithium-ion battery, chosen for its capacity and dimensions, serves as the energy storage medium.

For communication, the system employs LoRaWAN, utilizing a public network for data relay. This arrangement facilitates remote monitoring and data analysis, essential for assessing the system’s operational efficacy.

In this paper, we meticulously detail the setup for experiments involving a simulation and hardware system in EH applications, highlighting the sources of weather data and the parameters utilized.

5.1. Experimental Setup

The experimental setup for our EH system is comprehensively detailed in this section, shown in Figure 7. A key aspect of our setup is the precise specification of the hardware used. Tables 1 and 2 provide a detailed list of all critical hardware components, their specifications, and their roles in the experiment.

The core element of this study’s simulation is meteorological data, encompassing measurements of solar irradiance, rain intensity, and wind speed. These metrics are pivotal in determining the energy yield of solar cells, rain harvesting mechanisms, and wind turbines. To facilitate this analysis, historical weather data from a variety of locations, predominantly in China, have been incorporated. This data is sourced from extensive meteorological databases, each station within these databases providing unique measurement frequencies and levels of accuracy [25].

To address instances where data gaps exist, typically due to instrumental errors or other anomalies, a replication strategy is implemented. This approach involves substituting missing data with information from prior days, thereby ensuring both consistency and temporal integrity within the dataset. Notably, while data pertaining to solar and rain measurements are largely complete, wind data at times presents sporadic gaps, though these instances are rare and limited in scope.

Visual representation of the weather data is a key aspect of this research, elucidating patterns in solar energy, as well as the more variable trends observed in wind speed and rain intensity.

5.2. Hardware System Placement

The physical setup of the EH system is strategically located on the balcony of a high-rise apartment, an environment conducive to unimpeded energy collection [26]. This positioning is chosen deliberately to maximize exposure to essential weather elements, thereby, optimizing the system’s operational efficiency.

5.3. Simulation Assumptions and Parameters

Parameters for the simulation are meticulously chosen to correspond with the physical characteristics of the actual hardware system and mirror standard values identified in related literature. These parameters cover a range of aspects such as microcontroller voltage, the maximum capacity for EH, and efficiency ratings for different components of the system. Self-discharge rates for both the battery and capacitor are derived from empirical studies, with the capacitor specifically modeled as a lithium ion type to simulate similar voltage characteristics as a battery, shown in Table 3.

6.1. Simulation Results

Experiments were conducted to evaluate various energy harvesters, all adjusted to yield similar average power levels. This study found that, in most cases, solar energy was the most efficient and dependable, requiring the least space for generating energy. However, when using a 350 mAh lithium-ion battery, the wind turbine was more space-efficient.

The area needed for the wind turbine stayed more or less unchanged across different battery sizes, unlike the solar cell, whose area requirement varied greatly. This variability is due to the seasonal changes in solar energy, with larger areas needed during low-energy periods like winter.

The average power output from each type of harvester was lower than figures often reported in the literature. Tests with various energy harvester combinations indicated that no single type could power the system all year round. Solar power was effective in summer but insufficient in winter, whereas wind power offered a steadier supply throughout the year. Combining wind and rain harvesters proved most effective in reducing system downtime, with the solar and wind combination next inline. A system integrating all three energy sources (Figure 8) was considered the most reliable for continuous power generation.

The SoC of the battery is plainly demonstrated in Figure 8 through the use of box diagrams, which demonstrate the fluctuation of the SoC in response to various EH configurations (solar, wind, rain, and their combinations). Figure 8 provides a detailed analysis of the fluctuations in battery SoC under a constant power output of 1 mW across various energy configurations, both single and combined. The analysis highlights distinct differences in performance:

Single energy configurations: Solar power displays a broad range of SoC, with a median approximately at 0.3, indicating its relative instability and dependence on sufficient sunlight. Wind energy, while experiencing narrower fluctuations, has a slightly lower median SoC of about 0.25, reflecting its vulnerability to wind speed variations that may result in extended periods of inadequate energy supply. Rain energy, although the least fluctuating, shows the lowest median SoC around 0.15, due to its lower availability and energy conversion efficiency compared to solar and wind.

Combined energy configurations: The combination of solar and wind energy demonstrates a synergistic effect, showcasing a higher median SoC of around 0.45 and reduced fluctuations, which points to the mitigated instability that either source would face individually. Combinations of solar and rain, as well as wind and rain, though improving SoC medians and reducing fluctuations, are not as effective as the combination of solar and wind. Notably, the integration of all three energy sources—solar, wind, and rain—yields the highest median SoC at about 0.65 with the least fluctuations, significantly enhancing system stability and reliability. This comprehensive evaluation underscores the benefits of diversified energy configurations in improving the robustness and effectiveness of energy systems, demonstrating that combining multiple energy sources can substantially stabilize and enhance the overall energy management framework.

A further comparative study focused on adjusting the harvesters to a uniform area, a common practice in situations with spatial limitations. In this scenario, when both solar and wind harvesters were scaled down to 2 cm², neither experienced any deadtime individually. Yet, wind power encountered increased downtime, a result of its lesser power density relative to solar. At this size, integrating solar and wind energy sources proved effective in maximizing system uptime and minimizing downtime.

6.2. Comparison with Simulation

A critical aspect of this study involved comparing the energy harvested by the hardware system with the simulated projections. For this purpose, the simulation parameters were meticulously updated to mirror the real system’s configuration as closely as possible. This included adjustments to the solar cell efficiency, microcontroller power consumption, and the system’s voltage limits, which were used in lieu of SoC limits for a more practical and direct comparison.

Notably, the analysis revealed that the real system’s solar panel was unable to surpass ~55 mW, a figure significantly lower than the over 250 mW peak observed in the simulation. To account for this disparity, a power limit of 55 mW was introduced in the simulation, aligning it more closely with the hardware system’s performance. The physical constraints of the system’s environment, such as the shadowing effects on the solar cell, were also considered, as they contributed to the lower power output in the late afternoon, a factor not accounted for in the simulation.

Further, the comparison of the wind harvester between the real system and the simulation indicated a substantial difference in power generation, with the real wind turbine producing considerably less power. This discrepancy led to the conclusion that wind power had minimal impact on the battery’s SoC, making the simulation more representative of the real system when only considering solar harvesting.

6.2.1. Statistical Analysis

A statistical comparison of the system’s battery voltage between the hardware implementation and the simulation was conducted (Figure 9). When wind power was excluded from the simulation, a higher correlation with the real system’s data was observed. This correlation was quantified using the Pearson correlation coefficient, and the significance of the differences was assessed using analysis of variance (ANOVA), as indicated in the detailed tables and figures.

Despite the small p value indicating significant differences, the Pearson correlation coefficient suggested a strong correlation between the simulated and real-system voltages. This was particularly evident when wind power was excluded from the simulation, as shown in the provided box plot and statistical tables.

Figure 9 provides additional information regarding the influence of different energy configurations on battery SoC, both with and without the implementation of an optimization algorithm: Without algorithm: SoC demonstrates substantial fluctuations in all configurations, with generally low medians. This suggests that energy management efficiency is low without optimization, resulting in frequent low battery charge states. With algorithm: The algorithm substantially increases the median SoC and reduces fluctuations across the same configurations, thereby demonstrating effective management of energy inputs and battery charging/discharging. This, in turn, enhances battery efficiency and overall system stability. Altogether comprehensive contrast, it is evident that the efficiency and stability of the system are considerably improved by the use of an algorithm and diversified energy combinations when the data from Figures 8 and 9 is compared. In particular, the algorithm’s implementation enhances the average SoC and mitigates fluctuations across all configurations, underscoring the indispensable function of intelligent algorithms in the optimization of energy distribution and management, particularly in multisource energy systems. This optimization not only enhances energy utilization efficiency but also reduces maintenance costs and extends the life of batteries. In conclusion, this analysis underscores the challenges inaccurately simulating real-world environmental conditions and system constraints. While the simulation provided valuable insights, it became evident that certain practical aspects, particularly concerning the wind harvester, were not fully captured. Nonetheless, the strong correlation in solar power harvesting between the simulation and the hardware system highlights the simulation’s value in predicting system performance under specified conditions.

7.1. Evaluation of EH Methods

The study initially investigated different EH methods such as solar, wind, rain, and RF to evaluate their effectiveness in powering a LoRa node consistently. It was found that solar panels and wind turbines, due to their superior average power densities, were the most suitable for this task. In contrast, harvesting energy from rain using piezoelectric elements and from RF sources proved to be inefficient. Their low power densities failed to meet the LoRa node’s typical energy demand of around 1 mW.

The research showcased that by combining a 2 cm² solar cell and a similarly sized wind turbine, coupled with a compact lithium battery (350 mAh, 1,260 mW hr), a LoRa node requiring an average power of 1 mW could achieve 99% operational availability. This result highlights the benefits of using both solar and wind EH techniques together to provide a dependable power source for low-power devices such as LoRa nodes.

7.2. Smart Algorithm for Energy Source Selection

The second part of the research involved the development and implementation of a power management algorithm, both in simulation and hardware. This algorithm’s objective was to intelligently select the most available energy source at any given time. The implementation of this algorithm led to a decrease in system deadtime, where deadtime is defined as periods when the battery is completely discharged. Consequently, the system exhibited higher overall availability. However, the research also highlighted a tradeoff introduced by the algorithm. While it extended battery life by maintaining the SoC within certain limits, this also resulted in increased downtime, defined as periods when the SoC is below 20%, during which the system conserves power by not transmitting messages. The research concludes that the use of a capacitor alone or complete elimination of an energy storage element is not feasible for maintaining sufficient system availability. Therefore, a small battery or a large capacitor remains essential for ensuring the system’s effective operation.