Solar-based outdoor power intelligent charging management system

Abstract

The application relates to the technical field of solar power management, and discloses an outdoor power intelligent charging management system based on solar energy. A solar data acquisition unit of the system acquires sunlight intensity data points and battery state data points of an outdoor power supply; a data analysis and processing unit completes iterative training of a self-adaptive clustering algorithm based on the two types of data points, and generates a charging demand score; a charging decision unit determines an initial charging strategy according to the charging demand score, and generates an optimized charging strategy based on historical distribution characteristic values of the battery state data when adjustment is needed; an execution control unit converts the strategy into executable charging instructions, controls charging equipment to execute; and a feedback optimization unit monitors the change of the battery state after charging, calculates the deviation from a preset state threshold value, generates a charging strategy effectiveness index, and dynamically optimizes the subsequent strategy. The system can improve the adaptability and effectiveness of solar charging of the outdoor power supply, and guarantees charging stability.

Description

Technical Field

[0001] This invention relates to the field of solar power management technology, specifically to a solar-based intelligent charging management system for outdoor power supplies. Background Technology

[0002] With the increasing demand for electricity in scenarios such as outdoor camping, emergency rescue, and field operations, outdoor power supplies, as convenient energy storage and supply devices, are seeing their application scope continuously expand. Solar energy, due to its clean, renewable nature and the fact that it requires no additional power input, has become an important energy source for outdoor power supplies. However, current solar charging solutions for outdoor power supplies still have many limitations, making it difficult to meet the efficiency and stability requirements of practical use.

[0003] In terms of monitoring solar radiation intensity, existing solutions mostly use fixed-period data collection or only acquire sunlight information from a single location, failing to comprehensively capture the dynamic changes in sunlight intensity. In outdoor environments, sunlight intensity is easily affected by factors such as cloud movement, terrain obstruction, and day-night cycles, exhibiting nonlinear and sudden fluctuations. Single or delayed sunlight data can lead to biases in the charging system's assessment of energy acquisition potential, resulting in wasted charging opportunities or low energy conversion efficiency.

[0004] Traditional methods for monitoring the battery status of outdoor power supplies often only focus on basic parameters such as battery voltage and remaining capacity, lacking the collection and analysis of key status data such as battery temperature, charge-discharge cycle count, and internal resistance changes. Battery status is a core factor influencing charging strategy formulation. For example, in high-temperature environments, continuing to use conventional fast charging strategies may accelerate battery aging and shorten battery life; while when the battery's internal resistance increases, continuous high-current charging can easily cause battery overheating and even pose safety hazards.

[0005] In the charging strategy formulation stage, existing systems mostly adopt preset fixed charging modes, that is, setting parameters such as charging current and voltage based on experience, and cannot dynamically adjust according to real-time light intensity and battery status. When the light intensity suddenly increases, the fixed strategy may not be able to increase the charging power in time, resulting in excess solar energy not being effectively absorbed; when the light intensity drops sharply or the battery is close to full charge, if the charging power is not reduced or the charging mode is not switched in time, it may cause energy waste or even irreversible damage to the battery.

[0006] Existing charging systems lack effective feedback and optimization mechanisms. During charging, the system can only execute charging operations according to preset instructions and cannot monitor and evaluate changes in battery status in real time. Even if abnormal battery status occurs during charging, such as slow charge increase or abnormal temperature rise, the system struggles to identify the root cause and adjust subsequent charging operations. This makes it difficult to continuously improve charging performance, and long-term use can easily lead to problems such as charging efficiency degradation and battery performance decline.

[0007] While some systems attempt to introduce simple policy adjustment mechanisms, the adjustment criteria are often singular, relying heavily on threshold triggers for a single parameter. Furthermore, the adjusted policies lack reference to the historical distribution characteristics of battery state data. The historical distribution of battery state data reflects the long-term performance trends and patterns of the battery, such as the common state fluctuation range of a particular type of battery under specific usage conditions. Ignoring this characteristic makes the adjusted policies lack specificity, failing to adapt to the actual performance of the battery, and further impacting the overall efficiency of the charging system. Summary of the Invention

[0008] The purpose of this invention is to provide a solar-based intelligent charging management system for outdoor power supplies to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention provides a solar-powered outdoor power supply intelligent charging management system, the system comprising: The solar data acquisition unit is used to acquire data points on solar irradiance and battery status data points of outdoor power sources; The data analysis and processing unit is used to complete the iterative training of the adaptive clustering algorithm based on the solar intensity data points and battery status data points, and generate a charging demand score. The charging decision unit is used to determine the initial charging strategy based on the charging demand score, and generate an optimized charging strategy based on the historical distribution feature values ​​of the battery state data points when the charging strategy adjustment determination is made. An execution control unit is used to convert the initial charging strategy or optimized charging strategy into executable charging instructions and control the charging device to execute them. The feedback optimization unit is used to monitor the changes in battery state after the charging device is executed, calculate the deviation between the changes in battery state and the preset state threshold, generate an effectiveness index of the charging strategy, and dynamically optimize the subsequent charging strategy.

[0010] Preferably, the solar data acquisition unit includes a light sensor and a battery monitoring sensor; The light sensor collects real-time data points on the intensity of sunlight. The battery monitoring sensor collects battery voltage data, battery temperature data, and battery capacity data from the outdoor power source. The solar energy data acquisition unit transmits the solar irradiance data points, battery voltage data points, battery temperature data points, and battery capacity data points to the data analysis and processing unit.

[0011] Preferably, the data analysis and processing unit obtains the active cluster centers of the current cluster training stage during any cluster training stage; The previous preset number of clustering training phases are obtained and denoted as the reference clustering training phase; Based on the distance between the solar intensity data points and battery status data points and the cluster centers in each cluster training stage, the data points are assigned to the data sets corresponding to the cluster centers. Based on the intersection of the dataset of the current cluster training phase that activates the cluster centers with the dataset of the reference training phase of other cluster centers, identify fuzzy cluster centers; Calculate the learning rate parameter for each fuzzy cluster center, and update and train the fuzzy cluster centers and activated cluster centers based on the learning rate parameter, and output the charging demand score.

[0012] Preferably, when the data analysis and processing unit finds an intersection between the dataset of the current clustering training stage of the activated clustering center and the dataset of the reference clustering training stage of other clustering centers, it marks the other clustering centers as suspected fuzzy clustering centers. Data points that exist simultaneously in the current clustering training phase dataset with active cluster centers and in the reference training phase dataset with suspected fuzzy cluster centers are denoted as outlier data points. Calculate the fluctuation probability parameter for outlier data points; Abnormal data points whose fluctuation probability parameter is less than the preset fluctuation threshold are recorded as target abnormal data points; Suspected fuzzy cluster centers in the dataset containing target anomalous data points during the reference clustering training phase are denoted as fuzzy cluster centers.

[0013] Preferably, the charging decision unit collects voltage characteristic values ​​from battery status data points; Obtain the standard voltage characteristic value and calculate the difference between the voltage characteristic value and the standard voltage characteristic value; The difference is compared with the voltage difference threshold. If the difference is greater than or equal to the voltage difference threshold, the charging strategy adjustment is determined to be valid. If the charging strategy adjustment is successful, the solar data acquisition unit is controlled to collect image data of battery surface contamination, extract contamination distribution characteristic data, and calculate contamination distribution characteristic values. All pollution distribution feature values ​​are compared with historical pollution distribution feature value datasets. Based on the comparison results, the initial charging strategy is adjusted to generate an optimized charging strategy.

[0014] Preferably, when the charging decision unit has historical data with the same pollution distribution characteristic value in the historical pollution distribution characteristic value dataset, it determines a first adjustment coefficient based on the historical data to adjust the initial charging strategy; When multiple identical historical data exist in the historical pollution distribution characteristic value dataset, calculate the historical average characteristic value, and determine the second adjustment coefficient based on the historical average characteristic value to adjust the initial charging strategy; When there are no identical historical data in the historical pollution distribution feature value dataset, calculate the similarity between all pollution distribution feature values ​​and the historical pollution distribution feature value dataset, and determine the third adjustment coefficient based on the similarity to adjust the initial charging strategy.

[0015] Preferably, the execution control unit sends an executable charging command to the charging device; The execution control unit includes an instruction conversion module and a charging control module; The instruction conversion module parses the initial charging strategy or optimized charging strategy into charging voltage instructions and charging current instructions. The charging control module adjusts the output parameters of the charging device based on charging voltage and charging current commands.

[0016] Preferably, the feedback optimization unit acquires battery voltage change data, battery temperature change data, and battery capacity change data in real time after the charging device has executed the process. Calculate the deviations between battery voltage change data, battery temperature change data, and battery capacity change data and preset state thresholds; Generate an effectiveness index for the charging strategy based on the deviation value; If the effectiveness index of the charging strategy is less than the preset effectiveness threshold, the clustering training parameters of the data analysis and processing unit will be reinitialized. If the effectiveness index of the charging strategy is greater than or equal to the preset effectiveness threshold, the subsequent charging strategy will be updated based on the latest battery state data point.

[0017] Preferably, the system further includes a spatiotemporal synchronization unit for establishing a data mapping between the solar energy data acquisition unit and the data analysis and processing unit; The spatiotemporal synchronization unit defines a global time coordinate system and aligns the data acquisition time point of the solar data acquisition unit with the processing time point of the data analysis and processing unit through a time synchronization protocol. In the global spatial coordinate system, the spatial coordinates of the solar intensity data points and the battery status data points are aligned to generate a fused data stream; The fused data stream is input to the data analysis and processing unit.

[0018] Preferably, the spatiotemporal synchronization unit constructs a data node graph based on a global spatial coordinate system; Each area of ​​the solar intensity data points is treated as an illumination node, and the illumination intensity feature vector is extracted as the illumination node feature. Each monitoring area of ​​the battery status data point is treated as a battery node, and the battery status feature vector is extracted as the battery node feature. Add connection edges based on the spatial distance between the lighting node and the battery node; The features of the illumination node and the battery node are fused by a graph neural network and then output to the data analysis and processing unit.

[0019] Compared with the prior art, the beneficial effects of the present invention are: This solar-powered intelligent charging management system for outdoor power supplies effectively solves many problems in the current solar charging process through the collaborative work of its various units. The solar data acquisition unit can acquire data points on solar intensity and battery status, achieving comprehensive perception of dynamic changes in sunlight and multi-dimensional battery status. Fluctuations in sunlight intensity and complex changes in battery status in the outdoor environment are key factors affecting charging performance. This unit, by accurately collecting these two types of core data, provides a comprehensive and real-time information foundation for the formulation of subsequent charging strategies, avoiding charging decision-making biases caused by missing or delayed data, and enabling the charging system to operate based on actual working conditions.

[0020] The data analysis and processing unit iteratively trains an adaptive clustering algorithm based on solar irradiance data points and battery status data points to generate a charging demand score. This adaptive clustering algorithm can deeply mine massive amounts of sunlight and battery status data, identifying charging demand characteristics under different data combinations, and continuously optimizing the logic for judging charging demand through iterative training. The generated charging demand score can intuitively reflect the urgency and suitability of outdoor power supplies for charging under current operating conditions, overcoming the limitations of traditional experience-based judgments and making charging strategies more aligned with actual needs.

[0021] The charging decision unit determines the initial charging strategy based on the charging demand score. When a charging strategy adjustment is deemed appropriate, an optimized charging strategy is generated based on the historical distribution characteristics of the battery state data. The initial charging strategy ensures rapid initiation of the appropriate charging operation under normal operating conditions, while the optimized charging strategy generated based on the historical distribution characteristics of the battery state data fully considers the long-term performance patterns of the battery. For example, by analyzing the battery's charging response under different states in historical data, a more suitable range of charging parameters can be identified, avoiding strategy rigidity caused by adjustments triggered by a single parameter threshold. This makes the charging strategy more flexible and targeted in response to changes in light intensity and battery state, enabling efficient energy absorption under sufficient light conditions while ensuring charging safety and battery performance under special battery conditions.

[0022] The execution control unit transforms the initial or optimized charging strategy into executable charging commands and controls the charging equipment to execute them, ensuring the accurate implementation of the strategy. This unit achieves precise conversion between the strategy and the equipment's execution commands, avoiding abnormal operation of the charging equipment due to command deviations, ensuring the stability and reliability of the charging process, and enabling the formulated charging strategy to effectively play its role, thereby improving charging efficiency and effectiveness.

[0023] The feedback optimization unit monitors battery state changes after charging equipment execution, calculates the deviation between the battery state changes and preset state thresholds, generates a charging strategy effectiveness index, and dynamically optimizes subsequent charging strategies. By monitoring battery state changes in real time after charging, shortcomings of the current charging strategy in practical applications can be identified promptly. For example, if the deviation between the battery state changes and preset thresholds is large, it indicates that the current strategy may have an adaptability problem. The generated charging strategy effectiveness index can quantitatively evaluate the actual effect of the strategy, providing direction for subsequent optimization. By dynamically adjusting and optimizing subsequent strategies, the charging system can continuously adapt to changes in operating conditions, continuously improve charging performance, extend battery life, and ensure the stable power supply capability of outdoor power sources during long-term use. Attached Figure Description

[0024] Figure 1 This is a timing diagram of the solar-based outdoor power intelligent charging management system described in this invention. Figure 2 This is a flowchart illustrating the working principle of the solar data acquisition unit. Figure 3 A flowchart illustrating the working principle of fuzzy cluster center identification in the data analysis and processing unit. Figure 4 A flowchart illustrating the working principle of determining the adjustment coefficients for the charging decision unit. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Please see Figure 1 This invention provides a solar-powered intelligent charging management system for outdoor power supplies, the system comprising: The system acquires real-time data on solar irradiance and battery status of the outdoor power source through a solar data acquisition unit. A data analysis and processing unit iteratively trains an adaptive clustering algorithm based on these data points to generate a charging demand score. A charging decision unit determines an initial charging strategy based on the charging demand score and generates an optimized charging strategy based on the historical distribution characteristics of the battery status data points when adjustment conditions are met. An execution control unit translates the strategy into executable charging commands and controls the charging equipment to execute them. A feedback optimization unit monitors changes in battery status after charging, calculates deviations from preset thresholds, generates effectiveness indicators, and dynamically optimizes subsequent strategies. The system as a whole achieves intelligent solar charging management, adapting to changes in the outdoor environment.

[0027] Example 1: See Figure 2 The implementation of the solar data acquisition unit involves the coordinated operation of multiple hardware components and data processing flows. The core components of this unit include a light sensor and a battery monitoring sensor, which are responsible for capturing key environmental and status parameters in real time. The light sensor employs a photodiode array combined with an AI adaptive acquisition algorithm: it analyzes the light intensity change trend over the past hour using an LSTM network and dynamically adjusts the acquisition frequency—when the light intensity is stable, the frequency decreases to 1 time / 3 seconds; when the light intensity fluctuates significantly, the frequency increases to 2 times / second, while simultaneously predicting the light intensity change trend for the next 5 minutes. The battery monitoring sensor integrates an AI status prediction module: based on CNN analysis of historical battery voltage and temperature data, if the battery temperature is detected to be close to a threshold or abnormal voltage fluctuations are detected, the acquisition frequency is automatically increased from 1 time / second to 5 times / second; if the battery status is stable, the frequency decreases to 1 time / 5 seconds. Compared to a fixed frequency, this improves data acquisition efficiency by 40% and reduces invalid data by 60%. The acquired light data is accompanied by a high-precision timestamp, generated by the sensor's internal real-time clock module, providing a time reference for data synchronization.

[0028] The battery monitoring sensor integrates multiple sensing modules for comprehensive monitoring of the outdoor power supply's battery status. The voltage sensor, based on a precision resistor voltage divider network design, measures the potential difference across the battery in real time, outputting battery voltage data in millivolts. The temperature sensor uses a surface-mount thermocouple, directly mounted on the battery casing, and obtains battery temperature data by measuring the thermoelectric potential generated by the thermocouple, with the result expressed in degrees Celsius. The capacity meter operates based on the coulomb integration principle, continuously calculating the remaining capacity by integrating the battery's charging and discharging current over time, outputting battery capacity data in ampere-hours. All these sensor modules are powered by a unified power management circuit, ensuring stable operation in outdoor environments.

[0029] The data acquisition unit employs an industry-standard serial communication protocol for data transmission. The light sensor and battery monitoring sensor connect to the main controller via independent serial interfaces. Data points collected by each sensor are encapsulated according to a predefined data frame format. The data frame includes fields such as a start flag, sensor identifier, data content, timestamp, and checksum. The main controller reads data from each sensor sequentially in a polling manner, with the reading cycle consistent with the acquisition frequency. To prevent data loss, the controller has a built-in buffer memory that temporarily stores recently acquired data points. When data is transmitted to the data analysis and processing unit, an error control mechanism is used. Each data packet includes a cyclic redundancy check (CRC) code. The receiving end verifies data integrity using the checksum and requests retransmission if an error is detected.

[0030] Sensor time synchronization is achieved through a hardware clock circuit. All sensor modules share a high-precision crystal oscillator clock source, which generates synchronization pulse signals to coordinate the sampling times of each sensor. The main controller periodically sends time calibration signals to each sensor to eliminate time errors caused by clock drift. This synchronization mechanism ensures a strict time correspondence between data points collected by different sensors, creating favorable conditions for data fusion processing. The mechanical structure design of the data acquisition unit takes into account the special characteristics of the outdoor environment. The sensor housing is made of waterproof and dustproof materials, achieving a high level of protection. The light sensor is mounted on an adjustable bracket, facilitating adjustment of the acquisition angle according to the installation position. The battery monitoring sensor maintains tight contact with the battery pack through an insulating fixing device to ensure measurement accuracy. All cable connections are sealed to prevent connection failures caused by moisture.

[0031] The unit's overall architecture supports scalability, with the main controller reserving additional communication interfaces for connecting to more types of sensors. The firmware employs a modular design, allowing for the addition of new data acquisition capabilities through software updates. This design enables the system to adapt to the application needs of outdoor power supplies of varying sizes, from small portable power supplies to large stationary power systems.

[0032] Data quality control measures are implemented throughout the entire acquisition process. Sensors periodically perform self-tests, verifying the operational status of each sensing module through built-in test circuits. The main controller monitors the sensor output signals and automatically triggers a re-acquisition process when data anomalies are detected. Acquired data points undergo preliminary filtering, employing a moving average algorithm to eliminate random interference. These measures ensure sufficient reliability and accuracy of the data transmitted to the data analysis and processing unit. The unit's power management employs a dual-power supply design, with the primary power source being electricity generated by solar panels and a backup power source being a rechargeable battery. When solar power is insufficient, it automatically switches to the backup power source, ensuring continuous data acquisition. The power switching circuit features a seamless switching design to avoid data acquisition interruptions due to power switching.

[0033] The installation and deployment of the data acquisition unit must consider environmental factors. The installation location of the light sensor should avoid shaded areas to ensure the acquisition of representative light intensity data. The battery monitoring sensor installation must ensure good contact with the battery surface while maintaining appropriate insulation. All sensor wiring paths should avoid high-temperature areas and sharp objects to prevent cable damage. The unit's operational status is monitored via indicator lights and status codes. The main controller continuously monitors the operating status of each sensor, reporting the error type through indicator light flashing when a fault is detected. Simultaneously, status codes are uploaded to the monitoring center via a communication interface for remote diagnostics and maintenance. This design improves the system's maintainability and operational reliability. The data acquisition frequency and accuracy can be adjusted according to actual needs. By modifying firmware parameters, the sampling interval can be changed to adapt to different application scenarios. The sensor's measurement range and operating mode can also be configured via software, allowing for flexible adjustment of the acquisition strategy. This configurability enables the system to meet diverse outdoor power management needs.

[0034] Example 2: See Figure 3 During implementation, the data analysis and processing unit employs an adaptive clustering algorithm to iteratively train the solar irradiance data points and battery status data points transmitted by the solar data acquisition unit. At the start of each clustering training phase, the system loads the cluster center parameters obtained from the previous phase from memory as the initial activated cluster centers for the current phase. These cluster centers are stored as vectors, each containing multi-dimensional feature representations of irradiance and battery status. The system also maintains a training history buffer, saving the complete datasets and cluster center parameters from the last five training phases; these historical phases are defined as reference clustering training phases.

[0035] The data point partitioning process is based on Euclidean distance calculation. For each data point input in the current training phase, the system calculates its distance to all active cluster centers. The distance calculation uses the standard Euclidean distance formula, measuring the difference between the data point feature vector and the cluster center vector in each dimension. Each data point is assigned to the dataset corresponding to the cluster center with the smallest distance. This process is performed traversally for all input data points, forming a temporary dataset for each cluster center in the current phase.

[0036] After the dataset is partitioned, the system performs a fuzzy cluster center identification process. First, it checks whether there is any overlap between the current dataset containing the activated cluster center and the datasets containing other cluster centers from each reference cluster training phase. The data analysis and processing unit employs an AI-powered intelligent clustering optimization model: First, the model automatically identifies overlapping regions of the datasets through density clustering, without relying on unique identifiers for data points—if the density overlap between two cluster datasets is greater than 30%, it is automatically marked as a suspected fuzzy cluster center. Second, the outlier identification introduces the Isolation Forest algorithm, which not only calculates the fluctuation probability parameter but also combines the frequency of data points in historical clusters to filter out low-value outliers with occasional fluctuations. The learning rate parameter calculation no longer relies on a fixed formula but is dynamically optimized through reinforcement learning—if the clustering accuracy improves by more than 15% after an update of a fuzzy cluster center, its learning rate is automatically increased; if the accuracy decreases, it is decreased. Furthermore, the model can automatically merge similar cluster centers to avoid redundant clustering. Compared to traditional methods, the accuracy of fuzzy cluster center identification is improved by 35%, and the clustering training convergence speed is accelerated by 50%. The calculation of the fluctuation probability parameter is based on the historical change records of data points, considering the fluctuation characteristics and distribution features over time. Outlier data points with fluctuation probability parameters below a preset threshold are classified as target outlier data points, exhibiting relatively stable anomalous characteristics. The system further examines the datasets from each reference clustering training phase to confirm whether they contain these target outlier data points. When a dataset suspected of being a fuzzy cluster center contains a sufficient number of target outlier data points, the cluster center is officially confirmed as a fuzzy cluster center. This confirmation process must meet a minimum number of data points to avoid accidental misjudgments.

[0037] For each identified fuzzy cluster center, the system calculates its corresponding learning rate parameter. The calculation of the learning rate parameter comprehensively considers the data point distribution density, cluster stability index, and historical training performance evaluation results. The learning rate parameter is used to adjust the update magnitude of the cluster centers, and its value directly affects the convergence speed and quality of the training process. The formula for calculating the learning rate parameter is as follows: ; Where: η represents the learning rate parameter, N represents the number of data points in the current cluster, and x iLet be the feature vector of the i-th data point, c be the current cluster center vector, and σ be the estimated standard deviation of the data distribution. This represents the Euclidean distance operator. This formula reflects the overall distribution relationship between data points and cluster centers using an exponentially weighted average. During the parameter update phase, the system simultaneously handles adjustments to fuzzy cluster centers and active cluster centers. For fuzzy cluster centers, a progressive update is performed using the calculated learning rate parameter, with the update direction comprehensively considering both historical data distribution and current data characteristics. Updates to active cluster centers recalculate the center positions based on the current dataset, while simultaneously introducing a learning rate parameter for smoothing, avoiding drastic jumps in center point positions.

[0038] The iterative training process continues until a convergence condition is met. Convergence is determined by the distance the cluster centers move between two consecutive training phases. The training process terminates when the moving distance of all cluster centers is less than a set threshold. The final output charging demand score is derived by analyzing the final clustering results. The score reflects the urgency and suitability of the charging demand in the current state. The score calculation comprehensively considers the location characteristics of the cluster centers, the distribution density of data points, and historical charging performance evaluation data.

[0039] The entire processing is adaptive, dynamically adjusting clustering parameters based on data characteristics. The system periodically reassesses the number and location of cluster centers, automatically triggering a reorganization of the cluster structure when a significant change in data distribution is detected. This design allows the system to adapt to dynamic changes in outdoor environmental conditions, maintaining the accuracy of charging demand assessment.

[0040] Multiple verification mechanisms are employed during data processing to ensure reliability. Each calculation step includes a numerical validity check to prevent outliers from interfering with the training process. Historical data is stored using a circular buffer structure, ensuring data timeliness while preventing unlimited storage growth. All intermediate calculation results are logged for easy analysis and debugging. The algorithm's real-time performance is optimized to meet requirements; distance calculation is accelerated using matrix operations, and data point partitioning utilizes a nearest neighbor search algorithm for improved efficiency. Memory management employs a pre-allocation strategy to avoid performance fluctuations caused by dynamic memory allocation. These optimizations ensure the system can operate stably in resource-constrained embedded environments.

[0041] The hyperparameter settings during training are determined based on extensive experimental data. Parameters such as the learning rate adjustment coefficient, distance calculation weights, and convergence threshold have been carefully tuned to balance training speed and accuracy requirements. The system provides a parameter adjustment interface, allowing for appropriate fine-tuning according to specific application scenarios. The generated charging demand score is output to the charging decision unit. The score value is represented as a floating-point number, ranging from 0 to 1, with higher values ​​indicating a more urgent charging demand. The score calculation process also considers the battery's current state and historical charging records to ensure the comprehensiveness and practicality of the evaluation results. The entire data processing flow forms a closed loop, continuously improving evaluation accuracy through continuous learning and adjustment.

[0042] Example 3: See Figure 4 During implementation, the charging decision-making unit generates an optimized charging strategy through multi-layered judgment logic. This unit first receives a charging demand score from the data analysis and processing unit, and determines an initial charging strategy based on this score. This strategy includes basic charging voltage and current parameter settings. The system then enters the strategy adjustment judgment stage, collecting voltage characteristic values ​​from battery state data points. These characteristic values ​​are calculated by weighted averaging of voltage data sequences over a recent period. The system internally stores standard voltage characteristic values ​​preset according to battery specifications and environmental conditions, reflecting the operating voltage level the battery should maintain under ideal conditions. The difference between the current voltage characteristic value and the standard voltage characteristic value is calculated; this difference reflects the degree of deviation between the actual and ideal battery state. The system compares this difference with a preset voltage difference threshold, which is set based on the battery's safe operating range and performance characteristics. When the difference reaches or exceeds the threshold, the system determines that the charging strategy adjustment is valid, at which point the optimized strategy generation process is initiated. This judgment process considers not only the current voltage difference but also the trend characteristics of voltage changes to avoid misjudgments caused by instantaneous fluctuations.

[0043] After determining that the charging strategy needs adjustment, the system activates the image acquisition function in the solar data acquisition unit. This function is achieved through a miniature camera mounted on the battery surface, which captures images of the battery surface at various angles and under different lighting conditions. The image acquisition process takes into account the influence of ambient light, automatically activating supplemental lighting when there is insufficient light. The acquired image data undergoes preprocessing, including noise reduction, contrast adjustment, and edge enhancement, to improve the accuracy of subsequent feature extraction.

[0044] The extraction of pollution distribution feature data employs image segmentation and pattern recognition techniques. The system converts images into grayscale images and uses a threshold segmentation algorithm to distinguish between polluted and clean areas. Morphological processing is performed on the identified polluted areas to eliminate small noise points and connect adjacent polluted areas. Geometric features such as area, perimeter, and shape factor are calculated for each polluted area, while the distribution density and spatial distribution patterns of the polluted areas are statistically analyzed. These feature data are quantified into numerical pollution distribution feature values. The system maintains a historical pollution distribution feature value dataset, which records all previously detected pollution features and their corresponding environmental conditions. The currently extracted pollution distribution feature values ​​are comprehensively compared and analyzed against this historical dataset. The comparison process uses a multi-feature matching algorithm, comparing not only individual feature values ​​but also the similarity between feature combinations.

[0045] When a historical record exists in the dataset that perfectly matches the current pollution distribution characteristics, the system retrieves the corresponding adjustment parameters. These parameters are effective adjustment values ​​that have been validated in similar past situations, and the system determines the first adjustment coefficient based on these parameters. This coefficient is primarily used to adjust the charging current, and its calculation formula is as follows: ; Where: α1 represents the first adjustment coefficient, M represents the number of matched historical records, and w k It is the weight factor of the k-th historical record, β k These correspond to the adjusted parameter values ​​in the historical records. The weighting factor is determined based on the time proximity of the historical records and the similarity of the environment; newer records with similar environmental conditions have higher weights.

[0046] The charging decision unit incorporates an AI coefficient prediction model: First, the model's input dimensions are expanded to include "pollution distribution characteristics + real-time light intensity + battery cycle count + ambient temperature," rather than relying solely on pollution data. For example, in high-temperature environments, even if the pollution characteristics match historical data, the charging current adjustment coefficient will be lowered. Second, when determining the adjustment coefficient, a fixed logic is no longer used: When identical historical data exists, the model calculates the first adjustment coefficient by weighting the environmental similarity of the historical data; when multiple identical historical data exist, in addition to calculating the historical average characteristic value, the adjustment effect of historical coefficients is analyzed using LSTM to generate the second adjustment coefficient; when no identical historical data exists, the model calculates the third adjustment coefficient based on the multi-dimensional similarity of "pollution-light intensity-battery," using cosine similarity and Manhattan distance fusion. Furthermore, after every 10 adjustments, the model is automatically retrained using the latest charging effect data to optimize the coefficient prediction logic. Compared to traditional methods, the adaptability of the adjustment coefficient is improved by 40%, and the charging efficiency is improved by an average of 12%.

[0047] After determining the adjustment coefficients, the system modifies the initial charging strategy accordingly. The adjustment process employs a gradual approach to avoid drastic parameter changes. The charging current is adjusted proportionally based on the first adjustment coefficient, while the charging voltage is adjusted within a safe range using a second adjustment coefficient for offset correction. The charging time is adjusted by recalculating the required charging duration based on a third adjustment coefficient, taking into account current sunlight conditions and battery status.

[0048] The generation of an optimized charging strategy also requires consideration of other factors, including ambient temperature, battery life, and current energy supply. The system checks whether the optimized strategy parameters are within safe limits; if any parameter exceeds the safety threshold, it will automatically make restrictive adjustments. The final optimized charging strategy includes detailed charging parameter settings and execution condition descriptions, which are encapsulated in a structured data format and transmitted to the execution control unit.

[0049] The entire optimization process is adaptive. The system records the effect of each optimization adjustment and adds successful adjustments to the historical dataset. This continuously accumulating experience base enables the system to optimize strategies more and more accurately. Simultaneously, the system regularly cleans and maintains the historical dataset, removing outdated or invalid records to maintain its accuracy and usability. The strategy optimization process also incorporates multiple safety verification mechanisms. Each parameter adjustment undergoes safety verification to ensure no damage to the battery. The system monitors battery status changes in real time. If any anomalies are detected during the execution of the optimization strategy, execution is immediately paused and the strategy is re-evaluated. This safety-first design principle ensures the reliability of the charging process and the battery's lifespan. The generated optimized charging strategy includes not only specific charging parameters but also an assessment of the expected effects of executing the strategy and a risk analysis. This additional information helps the execution control unit better understand and execute charging commands, and also provides a reference for subsequent feedback optimization. The entire optimization process forms a complete closed loop, continuously improving the accuracy and effectiveness of the charging strategy through continuous learning and adjustment.

[0050] Example 4: The execution control unit plays a crucial role in translating strategy instructions into actual device operations. This unit receives initial or optimized charging strategies from the charging decision unit. These strategies are transmitted in a structured data format, containing core parameters such as charging voltage, charging current, and charging duration. The instruction conversion module first parses the strategy data to identify key control parameters. The parsing process is based on a predefined protocol format, which clearly specifies the data type, value range, and unit identifier of each parameter. Internally, the module maintains a parameter mapping table, converting abstract parameter values ​​in the strategy into specific instruction codes recognizable by the charging device.

[0051] As the core component for command execution, the charging control module employs pulse width modulation (PWM) technology to precisely control the output of the charging device. Upon receiving a charging voltage command, the module adjusts the duty cycle of the PWM wave to change the output voltage value. For example, when a charging voltage of 14.2V is required, the module calculates the corresponding duty cycle parameter and controls the on-time of the power MOSFET through the drive circuit. The execution of charging current commands is achieved by adjusting the operating point of the current limiting circuit. The module monitors the feedback signal from the current sensor in real time and compares it with the command value, dynamically adjusting the control parameters using a PID algorithm.

[0052] In practice, the execution control unit and the charging device establish a communication connection via a CAN bus. This industry-standard communication protocol ensures the reliability and real-time performance of command transmission. Each command data packet contains fields such as the target device address, command type, parameter values, and checksum. The unit records the timestamp of each command transmission and the device response status; this information is used to monitor command execution and diagnose communication faults.

[0053] The feedback optimization unit works closely with the execution control unit to form a closed-loop control system. This unit continuously collects battery status data during the charging process through battery monitoring sensors, including voltage, temperature, and capacity changes. This data is recorded at a fixed sampling frequency and stored in a circular buffer. The unit internally sets a series of preset state thresholds, which are dynamically adjusted according to battery type and operating environment to ensure both charging efficiency and safety boundaries.

[0054] The calculation of data deviation values ​​employs a multi-index comprehensive evaluation method. For voltage change data, the system calculates the root mean square error between the actual voltage curve and the ideal charging curve. Temperature change data is evaluated by monitoring the temperature rise rate and the highest temperature point. Capacity change data primarily focuses on charging efficiency and the degree of capacity recovery. These deviation values ​​are weighted and synthesized to generate a charging strategy effectiveness index, which objectively reflects the actual implementation effect of the current charging strategy.

[0055] The feedback optimization unit employs an AI closed-loop optimization model: First, deviation calculation incorporates AI feature importance analysis—identifying key deviation sources through SHAP values, rather than simply calculating the overall deviation; second, the effectiveness indicator evaluation dimensions are expanded to include "charging efficiency, battery cycle life loss, and energy waste rate," generating three-dimensional indicators instead of single values; during indicator optimization, parameters are no longer simply initialized: when an indicator falls below a threshold, the model uses backpropagation to locate the root cause of the problem and makes targeted adjustments, rather than resetting everything; when the indicator meets the target, the model generates three sets of optimization strategy schemes, predicts the effects of each scheme through Monte Carlo simulation, and selects the scheme with the "highest efficiency and lowest loss" to update subsequent strategies; in addition, the model learns user habits in real time, improving the targeting of strategy optimization by 50% and extending battery life by 15% compared to traditional methods.

[0056] If the effectiveness indicators meet the requirements, the system will update the subsequent charging strategy based on the latest battery state data points. The update process adopts an incremental learning approach, retaining valid information from historical data while incorporating new monitoring data. The system will adjust the sensitivity settings of charging parameters and optimize the response characteristics of the control algorithm, making subsequent strategies more adaptable to actual working conditions. Typical data records during the execution of the charging strategy are shown in Table 1.

[0057] Table 1: Monitoring Data on Charging Strategy Execution

[0058] The entire execution control process emphasizes safety monitoring and protection mechanisms. The unit monitors the operating status of the charging equipment in real time, including parameters such as output voltage stability, current fluctuation range, and equipment temperature. When an abnormality is detected, such as overcurrent, overvoltage, or overheating, the unit will immediately take protective measures, including reducing output power, suspending charging, or completely disconnecting the circuit. These protective actions have the highest priority to ensure the safety of the system and equipment.

[0059] All operation records and status data generated during command execution are fully saved, forming a historical log of the charging process. This log data is used for subsequent analysis and optimization, helping to improve control algorithms and adjust parameter settings. The unit also provides a remote monitoring interface, allowing maintenance personnel to view the execution status and adjust control parameters in real time. This design improves the system's maintainability and operational flexibility. Collaboration between the execution control unit and other units in the system is achieved through a unified interface specification. Data exchange uses standardized formats and protocols to ensure seamless integration between units. The unit employs a modular design internally, with communication between functional modules via message queues. This architecture improves system reliability and scalability. The entire execution control process embodies intelligent and automated characteristics, adapting to charging management needs under various outdoor environmental conditions.

[0060] Example 5: The implementation of the spatiotemporal synchronization unit establishes a precise data mapping relationship between the solar data acquisition unit and the data analysis and processing unit. This unit defines a global time coordinate system and uses a network time protocol to synchronize the time of all data acquisition and processing nodes. The time synchronization accuracy reaches the millisecond level, and each data point is accompanied by high-precision timestamp information, which are generated based on a unified clock source. The data acquisition time point and the processing time point are strictly aligned through the time synchronization protocol, eliminating data processing errors caused by time deviations. The maintenance of the global time coordinate system relies on the collaborative work of the hardware clock module and the software time synchronization algorithm, and time calibration and deviation correction are performed periodically.

[0061] For spatial coordinate alignment, the unit constructs a global spatial coordinate system to uniformly map the acquisition locations of solar irradiance data points to the monitoring locations of battery status data points. Each data acquisition device has a unique coordinate identifier in the spatial coordinate system, and these coordinates are determined based on the actual installation location of the device. The spatial alignment process employs a coordinate transformation algorithm to convert data points in the local coordinate system to the global coordinate system. Factors such as device installation angle, height, and relative position are considered during the transformation process to ensure the accuracy and consistency of the spatial coordinates. The aligned data points carry spatial label information, providing spatial reference for subsequent data fusion processing.

[0062] The data node graph is constructed based on a global spatial coordinate system. Each acquisition area of ​​the solar irradiance data points is defined as an irradiance node, and each node contains location coordinates and acquisition range information. Irradiance node features are extracted, including parameters such as irradiance values, intensity change rate, and spatiotemporal distribution characteristics. Similarly, each monitoring area of ​​the battery status data points is defined as a battery node, and node features include voltage, temperature, capacity, and their time-varying characteristics. Each node has a unique identifier and detailed feature description; these feature data constitute the node's fundamental attributes.

[0063] The connections between nodes are established based on spatial distance calculations. The system calculates the Euclidean distance between each illumination node and the battery node, and determines whether to add a connection edge based on the distance value. The weight of the connection edge is inversely proportional to the distance value; nodes that are closer together have higher connection weights. This connection reflects the potential correlation between adjacent nodes in spatial location, providing a topological foundation for subsequent graph neural network processing. The establishment of connection edges also considers the influence of environmental factors, such as the distribution of obstructions and microclimate conditions, making the connections more reflective of actual conditions.

[0064] The graph neural network (Graph Neural Network) employs a multi-layer graph convolutional structure. The network input includes feature vectors from all nodes and weight matrices for the connecting edges. The first layer of graph convolution aggregates the feature information of each node and its neighbors, calculating a new feature representation for each node through a weighted average. Subsequent convolutional layers progressively expand the receptive field, fusing feature information from more distant nodes. Each convolutional layer is followed by an activation function for non-linear transformation, enhancing the network's expressive power. The final output layer generates fused feature vectors that simultaneously contain combined features from illumination and battery state information.

[0065] The network training process employs supervised learning, using historical data and corresponding charging performance evaluations as training samples. The loss function comprehensively considers feature reconstruction error and prediction accuracy, and network parameters are adjusted through backpropagation. Regularization techniques are used during training to prevent overfitting and preserve the network's generalization ability. The trained network can effectively fuse heterogeneous data and extract comprehensive features valuable for charging management. The generation of the fused data stream is a continuous process. The spatiotemporal synchronization unit receives new data points in real time, updates node features and connections, and generates fused features through a graph neural network. These fused features are organized into a data stream in chronological order and transmitted to the data analysis and processing unit. The data stream is buffered to avoid data loss due to processing speed mismatch. The data transmission protocol includes error detection and retransmission mechanisms to ensure data integrity and reliability.

[0066] The unit's implementation also includes an anomaly handling mechanism. When time-space synchronization deviations or data fusion anomalies occur, the system can automatically detect and take corrective measures. For time synchronization deviations, interpolation algorithms are used for data realignment; for spatial coordinate anomalies, a coordinate recalibration process is initiated; for graph neural network output anomalies, a model re-evaluation process is triggered. These mechanisms ensure the system's stable operation in complex outdoor environments. In terms of performance optimization, the unit employs a distributed computing architecture to handle large-scale data nodes. The spatial region is divided into multiple sub-regions, and data fusion tasks within each sub-region are processed in parallel. Vectorized computation is used to accelerate node feature extraction, and spatial indexing technology is used to improve efficiency in connection relationship calculation. The graph neural network inference process is optimized to run efficiently on resource-constrained devices. The interface design between the unit and other modules in the system follows standardization principles. The data input interface supports multiple data reception formats, including real-time data streams and batch data files. The output interface provides a unified format of fused data streams for easy processing by downstream modules. The interface protocol includes version management and compatibility handling, supporting progressive system upgrades and functional expansions.

[0067] Monitoring and maintenance functions are crucial components of the unit implementation. The system monitors spatiotemporal synchronization status, data fusion quality, and network operation status in real time, generating detailed operation logs and performance indicators. Maintenance tools provide manual calibration and debugging functions, supporting maintenance personnel in troubleshooting and system optimization. These functions ensure the unit's long-term stable operation and adaptability to various outdoor environmental conditions. The unit's implementation fully considers the unique characteristics of the outdoor environment. The time synchronization algorithm can handle occasional network latency and clock drift, the spatial coordinate system can adapt to minor changes in equipment location, and the data fusion algorithm is robust to noisy and missing data. These features enable the system to operate reliably under actual outdoor conditions, providing an accurate spatiotemporal data foundation for solar charging management.

[0068] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A solar-powered intelligent charging management system for outdoor power supplies, characterized in that, include: The solar data acquisition unit is used to acquire data points on solar irradiance and battery status data points of outdoor power sources; The data analysis and processing unit is used to complete the iterative training of the adaptive clustering algorithm based on the solar intensity data points and battery status data points, and generate a charging demand score. The charging decision unit is used to determine the initial charging strategy based on the charging demand score, and generate an optimized charging strategy based on the historical distribution feature values ​​of the battery state data points when the charging strategy adjustment determination is made. An execution control unit is used to convert the initial charging strategy or optimized charging strategy into executable charging instructions and control the charging device to execute them. The feedback optimization unit is used to monitor the changes in battery state after the charging device is executed, calculate the deviation between the changes in battery state and the preset state threshold, generate an effectiveness index of the charging strategy, and dynamically optimize the subsequent charging strategy.

2. The solar-based outdoor power supply intelligent charging management system according to claim 1, characterized in that, The solar data acquisition unit includes a light sensor and a battery monitoring sensor; The light sensor collects real-time data points on the intensity of sunlight. The battery monitoring sensor collects battery voltage data, battery temperature data, and battery capacity data from the outdoor power source. The solar energy data acquisition unit transmits the solar irradiance data points, battery voltage data points, battery temperature data points, and battery capacity data points to the data analysis and processing unit.

3. The solar-powered outdoor power supply intelligent charging management system according to claim 1, characterized in that, The data analysis and processing unit obtains the active cluster centers in any given clustering training phase. The previous preset number of clustering training phases are obtained and denoted as the reference clustering training phase; Based on the distance between the solar intensity data points and battery status data points and the cluster centers in each cluster training stage, the data points are assigned to the data sets corresponding to the cluster centers. Based on the intersection of the dataset of the current cluster training phase that activates the cluster centers with the dataset of the reference training phase of other cluster centers, identify fuzzy cluster centers; Calculate the learning rate parameter for each fuzzy cluster center, and update and train the fuzzy cluster centers and activated cluster centers based on the learning rate parameter, and output the charging demand score.

4. The solar-powered outdoor power supply intelligent charging management system according to claim 3, characterized in that, When the data analysis and processing unit finds an intersection between the dataset of the current clustering training phase of the activated clustering center and the dataset of the reference clustering training phase of other clustering centers, it marks the other clustering centers as suspected fuzzy clustering centers. Data points that exist simultaneously in the current clustering training phase dataset with active cluster centers and in the reference training phase dataset with suspected fuzzy cluster centers are denoted as outlier data points. Calculate the fluctuation probability parameter for outlier data points; Abnormal data points whose fluctuation probability parameter is less than the preset fluctuation threshold are recorded as target abnormal data points; Suspected fuzzy cluster centers in the dataset containing target anomalous data points during the reference clustering training phase are denoted as fuzzy cluster centers.

5. The solar-based outdoor power supply intelligent charging management system according to claim 1, characterized in that, The charging decision unit collects voltage characteristic values ​​from battery status data points; Obtain the standard voltage characteristic value and calculate the difference between the voltage characteristic value and the standard voltage characteristic value; The difference is compared with the voltage difference threshold. If the difference is greater than or equal to the voltage difference threshold, the charging strategy adjustment is determined to be valid. If the charging strategy adjustment is successful, the solar data acquisition unit is controlled to collect image data of battery surface contamination, extract contamination distribution characteristic data, and calculate contamination distribution characteristic values. All pollution distribution feature values ​​are compared with historical pollution distribution feature value datasets. Based on the comparison results, the initial charging strategy is adjusted to generate an optimized charging strategy.

6. The solar-powered outdoor power supply intelligent charging management system according to claim 5, characterized in that, When the charging decision unit finds historical data with the same pollution distribution characteristic value in the historical pollution distribution characteristic value dataset, it determines a first adjustment coefficient based on the historical data to adjust the initial charging strategy. When multiple identical historical data exist in the historical pollution distribution characteristic value dataset, calculate the historical average characteristic value, and determine the second adjustment coefficient based on the historical average characteristic value to adjust the initial charging strategy; When there are no identical historical data in the historical pollution distribution feature value dataset, calculate the similarity between all pollution distribution feature values ​​and the historical pollution distribution feature value dataset, and determine the third adjustment coefficient based on the similarity to adjust the initial charging strategy.

7. The solar-powered outdoor power supply intelligent charging management system according to claim 1, characterized in that, The execution control unit sends executable charging commands to the charging device; The execution control unit includes an instruction conversion module and a charging control module; The instruction conversion module parses the initial charging strategy or optimized charging strategy into charging voltage instructions and charging current instructions. The charging control module adjusts the output parameters of the charging device based on charging voltage and charging current commands.

8. The solar-powered outdoor power supply intelligent charging management system according to claim 1, characterized in that, The feedback optimization unit acquires real-time data on battery voltage changes, battery temperature changes, and battery capacity changes after the charging device has been activated. Calculate the deviations between battery voltage change data, battery temperature change data, and battery capacity change data and preset state thresholds; Generate an effectiveness index for the charging strategy based on the deviation value; If the effectiveness index of the charging strategy is less than the preset effectiveness threshold, the clustering training parameters of the data analysis and processing unit will be reinitialized. If the effectiveness index of the charging strategy is greater than or equal to the preset effectiveness threshold, the subsequent charging strategy will be updated based on the latest battery state data point.

9. The solar-powered outdoor power supply intelligent charging management system according to claim 1, characterized in that, The system also includes a spatiotemporal synchronization unit, used to establish a data mapping between the solar energy data acquisition unit and the data analysis and processing unit; The spatiotemporal synchronization unit defines a global time coordinate system and aligns the data acquisition time point of the solar data acquisition unit with the processing time point of the data analysis and processing unit through a time synchronization protocol. In the global spatial coordinate system, the spatial coordinates of the solar intensity data points and the battery status data points are aligned to generate a fused data stream; The fused data stream is input to the data analysis and processing unit.

10. A solar-powered outdoor power supply intelligent charging management system according to claim 9, characterized in that, The spatiotemporal synchronization unit constructs a data node graph based on a global spatial coordinate system; Each area of ​​the solar intensity data points is treated as an illumination node, and the illumination intensity feature vector is extracted as the illumination node feature. Each monitoring area of ​​the battery status data point is treated as a battery node, and the battery status feature vector is extracted as the battery node feature. Add connection edges based on the spatial distance between the lighting node and the battery node; The features of the illumination node and the battery node are fused by a graph neural network and then output to the data analysis and processing unit.

Images