Method and device for measuring state of charge of power supply battery of meteorological equipment

Abstract

The invention discloses a power supply battery state-of-charge measurement method and device for meteorological equipment, and relates to the field of battery management. The method comprises the following steps: acquiring real-time operation parameters of a power supply battery of the meteorological equipment and environmental parameters of an environment where the meteorological equipment is located; based on the real-time operation parameters, calculating to obtain an initial battery state-of-charge value; acquiring historical operation parameters and historical environment parameters of the meteorological equipment, and constructing a geographic feature vector based on the historical operation parameters and the historical environment parameters; according to the geographic feature vector, determining a trigger condition, and judging whether the environmental parameters meet the trigger condition; if it is determined that the environmental parameters meet the triggering conditions, dynamic calibration is executed based on the geographic feature vector and the initial battery charge state value to calculate a charge state correction value; and synthesizing the state-of-charge correction value and the initial battery state-of-charge value to obtain a final battery state-of-charge value. By implementing the technical scheme provided by the invention, the measurement accuracy of the state of charge of the power supply battery is improved.

Description

Technical Field

[0001] This application relates to the field of battery management, specifically to a method and apparatus for measuring the state of charge of a power supply battery for meteorological equipment. Background Technology

[0002] Meteorological equipment is often deployed in harsh outdoor environments and relies on battery power. Therefore, accurate measurement of the battery's State of Charge (SOC) is crucial for ensuring stable operation. However, traditional measurement methods such as the ampere-hour integration method do not consider the impact of environmental factors such as temperature on battery characteristics, resulting in significant cumulative errors. This leads to inaccurate SOC estimations and poses a risk to the data security of meteorological equipment.

[0003] To address this issue, existing technologies have introduced battery models that consider the effects of temperature. This technique dynamically adjusts the model's internal parameters by monitoring the ambient temperature in real time to compensate for the impact of temperature changes on battery performance. Compared to traditional methods, this temperature-compensated approach can correct errors to a certain extent, thereby improving the accuracy of SOC estimation.

[0004] However, this improvement method still has technical flaws. The battery model it uses is usually generic, only reflecting the instantaneous impact of real-time temperature, while ignoring the long-term and irreversible aging effects caused by the specific geographical environment of the device (such as high altitude, extreme temperature differences, etc.). Therefore, as the battery's service life increases, the generic model will gradually deviate from the actual state of the battery under specific environmental conditions, eventually leading to a decrease in the accuracy of SOC measurement. Summary of the Invention

[0005] This application provides a method and apparatus for measuring the state of charge of a power supply battery for meteorological equipment, which improves the accuracy of measuring the state of charge of the power supply battery.

[0006] A first aspect of this application provides a method for measuring the state of charge (SOC) of a power supply battery for a meteorological device. The method includes: acquiring real-time operating parameters of the power supply battery of the meteorological device and environmental parameters of the environment in which the meteorological device is located; calculating a preliminary SOC value based on the real-time operating parameters; acquiring historical operating parameters and historical environmental parameters of the meteorological device, and constructing a geographic feature vector based on the historical operating parameters and the historical environmental parameters; determining a triggering condition based on the geographic feature vector, and determining whether the environmental parameters meet the triggering condition; if the environmental parameters meet the triggering condition, performing dynamic calibration based on the geographic feature vector and the preliminary SOC value to calculate a corrected SOC value; and synthesizing the corrected SOC value with the preliminary SOC value to obtain a final SOC value.

[0007] By adopting the above technical solution, the method first acquires the real-time operating parameters of the meteorological equipment's power battery and the environmental parameters of the environment in which the meteorological equipment is located. Using the real-time operating parameters, a preliminary battery state of charge (SOC) value is calculated, providing a foundation for subsequent accurate SOC estimation. Then, the method acquires the historical operating parameters and historical environmental parameters of the meteorological equipment and constructs a geographic feature vector based on the historical data, fully utilizing the characteristic information of the meteorological equipment's geographical location. Next, the method determines the triggering conditions based on the geographic feature vector, judges whether the current environmental parameters meet the triggering conditions, and adaptively decides whether the preliminary battery SOC value needs to be corrected. If the environmental parameters meet the triggering conditions, the method performs dynamic calibration based on the geographic feature vector and the preliminary battery SOC value, calculating the SOC correction value. This achieves adaptive correction of the battery SOC estimation results based on the characteristics of the meteorological equipment's geographical environment, improving estimation accuracy. Finally, the method synthesizes the SOC correction value with the preliminary battery SOC value to obtain the final battery SOC value, realizing battery SOC estimation that integrates multi-source information, balancing the accuracy and continuity of the estimation results. In summary, this method makes full use of the real-time operating status of meteorological equipment, the characteristics of the geographical environment, and historical operating data to adaptively and dynamically correct the preliminary estimation results, ultimately obtaining robust and accurate battery state of charge values, providing an important basis for the reliable power supply and maintenance of meteorological equipment.

[0008] Optionally, the step of calculating the preliminary battery state of charge (SOC) value based on the real-time operating parameters specifically includes: establishing a second-order RC equivalent circuit model of the power supply battery and determining the system state vector, wherein the system state vector includes the battery SOC, ohmic internal resistance, and polarization voltage; discretizing the second-order RC equivalent circuit model and constructing the extended Kalman filter state transition equation and observation equation, wherein the state transition equation is used to describe the change of the system state vector over time; calculating the prior state estimate and prior covariance matrix at the second time based on the posterior state estimate and state transition equation at the first time point, wherein the first time point is the time before the second time point and the second time point is the current time; using the real-time operating parameters as observations, calculating the Kalman gain, and combining the prior state estimate and the observations to correct the system state vector to obtain the posterior state estimate at the second time point; and extracting the battery SOC component from the posterior state estimate as the preliminary battery SOC value.

[0009] By adopting the above technical solution, a second-order RC equivalent circuit model of the power supply battery is first established, and the system state vector, including the battery state of charge, ohmic internal resistance, and polarization voltage, is determined, accurately describing the dynamic characteristics of the battery based on the physical mechanism. Then, the method discretizes the second-order RC equivalent circuit model and constructs the state transition equation and observation equation required for the extended Kalman filter algorithm. The state transition equation describes the evolution of the system state vector over time, laying the mathematical foundation for recursive estimation. Next, based on the posterior state estimate and state transition equation from the previous time step, the method calculates the prior state estimate and prior covariance matrix for the current time step, realizing the time update of the system state. Then, the method uses real-time operating parameters as observations, calculates the Kalman gain, and combines the prior state estimate and observations to correct the system state vector, obtaining the posterior state estimate for the current time step, realizing state correction based on measured data. Finally, the method extracts the battery state of charge component from the posterior state estimate as the preliminary battery state of charge value. In summary, this method integrates the battery circuit model and the extended Kalman filter algorithm to achieve dynamic estimation of the battery's state of charge based on physical mechanisms and mathematical models. It makes full use of the state information contained in the real-time operating parameters, has high estimation accuracy and real-time performance, and provides a reliable foundation for subsequent adaptive correction.

[0010] Optionally, constructing a geographic feature vector based on the historical operating parameters and the historical environmental parameters specifically includes: performing timestamp alignment and data cleaning on the historical operating parameters and the historical environmental parameters to obtain a standardized historical dataset; extracting multi-dimensional statistical features from the standardized historical dataset to form a preliminary feature set; dividing the preliminary feature set into multiple sub-components based on the multi-dimensional statistical features, and calculating the cumulative variance contribution rate of each sub-component; selecting principal components from the preliminary feature set whose cumulative variance contribution rate exceeds a preset threshold, and linearly combining the principal components into the geographic feature vector.

[0011] By employing the aforementioned technical solution, the method first performs timestamp alignment and data cleaning on historical operating parameters and historical environmental parameters to obtain a standardized historical dataset, laying the data foundation for subsequent feature extraction. Then, the method extracts multi-dimensional statistical features from the standardized historical dataset to form a preliminary feature set, fully mining the geographical environmental characteristics contained in the historical data. Next, based on the multi-dimensional statistical features, the method divides the preliminary feature set into multiple sub-components and calculates the cumulative variance contribution rate of each sub-component, achieving dimensionality reduction and optimization of the feature dimensions. Finally, the method selects principal components from the preliminary feature set whose cumulative variance contribution rate exceeds a preset threshold and linearly combines these principal components into a geographical feature vector. In summary, this method fully utilizes historical data accumulated during the long-term operation of meteorological equipment, extracting multi-scale features of the geographical environment where the meteorological equipment is located from both temporal and spatial dimensions. Principal component analysis achieves feature dimensionality reduction, resulting in a geographical feature vector with high information content and high discriminative power. This reduces the complexity of subsequent calculations while preserving key features of the geographical environment, providing high-quality criteria for adaptive correction.

[0012] Optionally, the step of performing dynamic calibration based on the geographic feature vector and the preliminary battery state of charge (SOC) value to calculate the SOC correction value specifically includes: in a preset multidimensional mapping model, using the geographic feature vector and the preliminary battery SOC value as a joint index, querying and obtaining the calibration parameter set and reference voltage charge response slope corresponding to the joint index; wherein, the calibration parameter set includes calibration current and calibration duration; performing charge and discharge operations on the power supply battery according to the calibration current and the calibration duration to apply a disturbance charge, and measuring the voltage change caused by the disturbance charge; calculating the actual voltage charge response slope based on the disturbance charge and the voltage change; comparing the actual voltage charge response slope with the reference voltage charge response slope to obtain the slope deviation; and calculating the SOC correction value according to the slope deviation using a preset correction function.

[0013] By employing the aforementioned technical solution, firstly, in a pre-defined multi-dimensional mapping model, the geographic feature vector and the preliminary battery state of charge (SOC) value are used as a joint index to query and obtain the corresponding calibration parameter set and the reference voltage charge response slope, achieving adaptive matching between the correction parameters and geographic features and battery state. The calibration parameter set includes calibration current and calibration duration, laying the standard for subsequent perturbation application. Then, based on the calibration current and calibration duration, the method performs charge and discharge operations on the battery, applying a perturbation charge and measuring the resulting voltage change, obtaining the battery's actual response data to the charge perturbation in the current state. Next, based on the perturbation charge and voltage change, the method calculates the actual voltage charge response slope. Then, the method compares the actual voltage charge response slope with the reference voltage charge response slope to obtain the slope deviation, quantitatively reflecting the degree of difference between the actual response and the expected value. Finally, based on the slope deviation, the method calculates the SOC correction value using a pre-defined correction function, establishing a mapping relationship between the slope deviation and the SOC correction amount. In summary, this method uses a multidimensional mapping model as a bridge to achieve adaptive matching between calibration conditions, geographical features, and battery state. By applying standardized charge perturbations, the actual measured value of the battery response is obtained, and the state of charge correction is calculated based on the deviation between the actual response and the expected value. The entire process considers both the influence of geographical features and the differences in battery state, achieving adaptive dynamic correction of the state of charge estimation and effectively improving the estimation accuracy.

[0014] Optionally, the step of calculating the state of charge correction value based on the slope deviation using a preset correction function specifically includes: dividing the slope deviation by the reference voltage charge response slope to obtain a slope deviation ratio; multiplying the slope deviation ratio by a preset slope sensitivity coefficient to obtain a state of charge deviation ratio; multiplying the state of charge deviation ratio by the battery nominal capacity to obtain an initial state of charge deviation value; and using the preset correction function to map and limit the initial state of charge deviation value to obtain the state of charge correction value.

[0015] By employing the above technical solution, the slope deviation is first divided by the reference voltage-charge response slope to obtain the slope deviation ratio, which quantitatively describes the relative difference between the actual response and the expected value. Then, the method multiplies the slope deviation ratio by a preset slope sensitivity coefficient to obtain the state-of-charge (POC) deviation ratio, establishing a linear mapping relationship between the slope deviation and the POC deviation. The slope sensitivity coefficient reflects the sensitivity of the battery's POC to changes in the voltage-charge response slope. Next, the method multiplies the POC deviation ratio by the battery's nominal capacity, converting the relative deviation into an absolute deviation to obtain the initial POC deviation value, achieving dimensional unification of the deviation based on battery capacity. Finally, the method uses a preset correction function to map and limit the initial POC deviation value, obtaining the final POC correction value. This achieves a nonlinear transformation from deviation to correction, considering both the near-linear relationship between the correction and deviation for small deviations and the saturation effect of the correction for larger deviations, thus avoiding over-correction. In summary, this method provides a way to calculate the state of charge correction value, which comprehensively considers factors such as slope deviation, battery capacity, and sensitivity coefficient. Based on quantitative modeling, it realizes the nonlinear mapping from slope deviation to the state of charge correction value, thereby enhancing the adaptability and robustness of the correction process.

[0016] Optionally, the step of synthesizing the corrected state of charge (SOC) value and the preliminary SOC value to obtain the final SOC value specifically includes: using the preliminary SOC value as a first source of evidence and the corrected SOC value as a second source of evidence; constructing a first probability allocation function for the first source of evidence and a second probability allocation function for the second source of evidence based on the uncertainties of the preliminary SOC value and the corrected SOC value; fusing the first probability allocation function and the second probability allocation function to obtain a fused probability allocation function; weighting the fused probability allocation function using a preset time decay factor to obtain a weighted fused probability allocation; and calculating the final SOC value based on the weighted fused probability allocation.

[0017] By adopting the above technical solution, the method first uses the preliminary battery state of charge (SOC) value as the first source of evidence and the corrected SOC value as the second source of evidence, forming two independent information sources. Then, based on the uncertainties of the preliminary SOC value and the corrected SOC value, the method constructs probability allocation functions for each source of evidence, quantitatively describing the likelihood of each source supporting different SOC values, considering both the credibility of the evidence itself and the relative weights between the evidence sources. Next, the method fuses the two probability allocation functions to obtain a fused probability allocation function, effectively integrating information from different sources of evidence. Then, the method uses a preset time decay factor to weight the fused probability allocation function, obtaining a weighted fused probability allocation, introducing the influence of the time dimension and reflecting the changing trend of the credibility of the evidence over time. Finally, based on the weighted fused probability allocation, the method calculates the final battery SOC value, obtaining a SOC estimation result that comprehensively considers the preliminary estimate and the corrected value. In summary, this method presents a fusion approach for state of charge information based on evidence theory. It quantitatively describes the credibility of different evidence sources through a probability allocation function, and achieves optimized fusion of preliminary and corrected values ​​using evidence combination rules. It also considers the impact of evidence timeliness, and finally obtains robust and reliable estimates of battery state of charge.

[0018] Optionally, after synthesizing the corrected state of charge value with the preliminary battery state of charge value to obtain the final battery state of charge value, the method further includes: preprocessing long-term historical data received from multiple terminal devices, wherein the long-term historical data includes historical final battery state of charge values, historical operating parameters, historical environmental parameters, and historical geographic feature vectors; dividing the preprocessed long-term historical data into a training set, a validation set, and a test set; training a global state evolution model using the training set, wherein the global state evolution model is a model based on a long short-term memory network and includes at least one input layer, at least one long short-term memory network hidden layer, and at least one fully connected output layer; calculating a state reference value using the global state evolution model, and sending the state reference value to the corresponding terminal device, wherein the state reference value includes a reference state of charge value, a reference internal resistance value, and a reference open-circuit voltage.

[0019] By adopting the above technical solution, the method first preprocesses long-term historical data received from multiple terminal devices to obtain a standardized, non-redundant training dataset. This long-term historical data includes historical final battery state of charge (SOC) values, historical operating parameters, historical environmental parameters, and historical geographic feature vectors, covering various factors influencing SOC evolution. Then, the method divides the preprocessed long-term historical data into training, validation, and test sets, providing an evaluation tool for subsequent model training. Next, the method uses the training set to train a global state evolution model, fully utilizing the massive historical data collected from different terminal devices to uncover the general laws governing battery SOC evolution over time and construct a predictive model with strong generalization ability. The global state evolution model employs a model architecture based on long short-term memory networks, and by introducing gating units and attention mechanisms, it can model long-term dependencies and dynamic trends in time-series data. Finally, the method uses the trained global state evolution model to calculate state reference values ​​and sends these values ​​to the corresponding terminal devices, providing prior knowledge and reference for online SOC estimation by the terminal devices. The state reference values ​​include reference state of charge, reference internal resistance, and reference open-circuit voltage, providing estimates of multiple state variables for the terminal device. In summary, this method presents a battery state evolution modeling and prediction approach based on global historical data. Through edge-cloud collaboration, it utilizes big data and deep learning technologies to construct a robust and generalizable global state evolution model. The model estimation results are then distributed to the terminal device, enabling the sharing and transfer of prior knowledge across different terminal devices. This provides a global perspective and long-term experience for the terminal device's state estimation, contributing to improved estimation accuracy and reliability. Furthermore, this method trains and updates the model in the cloud, reducing the computational burden on the terminal device and improving the efficiency of model iteration and optimization.

[0020] In a second aspect, embodiments of this application provide a battery state of charge measurement device for meteorological equipment. The battery state of charge measurement device for meteorological equipment includes: one or more processors and a memory; the memory is coupled to the one or more processors and is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the battery state of charge measurement device for meteorological equipment to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a power supply battery state-of-charge measuring device for meteorological equipment, cause the power supply battery state-of-charge measuring device for meteorological equipment to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a power supply battery state-of-charge measuring device for meteorological equipment, cause the power supply battery state-of-charge measuring device for meteorological equipment to perform the method described in the first aspect and any possible implementation thereof.

[0023] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages: 1. This technical solution constructs a dynamic calibration mechanism by combining real-time operating parameters, historical operating parameters, environmental parameters, and geographic feature vectors, effectively compensating for the neglect of environmental factors and long-term aging effects by traditional SOC measurement methods. A preliminary SOC value is calculated using Kalman filtering, and the state of charge is dynamically corrected, significantly improving the measurement accuracy of battery SOC and ensuring the stable operation of meteorological equipment in complex environments.

[0024] 2. A dynamic calibration method and a global state evolution model are introduced to dynamically adjust and optimize multi-dimensional parameters to address the impact of specific geographical environments (such as extreme temperature differences and long-term aging) on ​​battery performance. Simultaneously, a global evolution model based on Long Short-Term Memory (LSTM) networks is used to accurately predict the long-term operating status of terminal devices, enhancing the system's adaptability to special usage scenarios.

[0025] 3. By modeling the uncertainties of the initial and corrected SOC values, and using a weighted approach based on a fusion probability allocation function and a time decay factor, the reliability and stability of the SOC measurement results are further improved. Combining global reference values ​​with historical data from terminal equipment provides targeted calibration capabilities, ensuring the long-term efficient operation of meteorological equipment while reducing data and operational risks caused by measurement errors. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating a method for measuring the state of charge of a power supply battery for meteorological equipment, as disclosed in an embodiment of this application. Figure 2 This is another schematic flowchart of a method for measuring the state of charge of a power supply battery for meteorological equipment disclosed in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a power supply battery state-of-charge measuring device for meteorological equipment provided in an embodiment of this application.

[0027] Explanation of reference numerals in the attached drawings: 301, Central Processing Unit; 302, Read-Only Memory; 303, Random Access Memory; 304, Bus; 305, Input / Output Interface; 306, Input Section; 307, Output Section; 308, Storage Section; 309, Communication Section; 310, Driver; 311, Removable Media. Detailed Implementation

[0028] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0029] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0030] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple system devices refer to two or more system devices, and multiple screen terminals refer to two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0031] This application provides a method for measuring the state of charge (SOC) of a power supply battery for meteorological equipment, referring to... Figure 1 , Figure 1 This is a flowchart illustrating a method for measuring the state of charge (SBC) of a power supply battery for meteorological equipment, provided in an embodiment of this application. The method is applied to a device, which is a server. The server can execute a program for measuring the SBC of a power supply battery for meteorological equipment. The method includes steps S101 to S106, as follows: Step S101: Obtain the real-time operating parameters of the power supply battery of the meteorological equipment and the environmental parameters of the environment in which the meteorological equipment is located.

[0032] In step S101, real-time operating parameters refer to the state quantities of the power supply battery at the current moment, such as voltage, current, and temperature. Environmental parameters refer to the real-time environmental conditions at the deployment location of the meteorological equipment, mainly including light intensity and ambient temperature. Meteorological equipment refers to unattended equipment deployed outdoors for collecting meteorological data. The power supply battery refers to the energy storage unit that provides power to the meteorological equipment, such as a lithium battery pack.

[0033] Specifically, the server establishes a communication connection with remote meteorological equipment via a wireless communication network, such as 4G or NB-IoT. The server periodically sends data request commands to the meteorological equipment's data acquisition unit. Upon receiving the command, the meteorological equipment uses its built-in voltage, current, and temperature sensors to measure the terminal voltage of the power supply battery, the loop current (defined as positive for charging and negative for discharging), and the battery surface temperature, respectively. These data collectively constitute the real-time operating parameters. Simultaneously, the meteorological equipment uses its own light sensor and ambient temperature sensor to measure the light intensity and air temperature at the current location; these data constitute the environmental parameters. The meteorological equipment packages the collected real-time operating parameters and environmental parameters into a data frame and sends it to the server via the wireless communication network. The server receives and parses this data frame to obtain the current parameter values, preparing for subsequent calculations.

[0034] Step S102: Calculate the preliminary battery state of charge value based on real-time operating parameters.

[0035] In step S102, the preliminary battery state of charge value represents the percentage of remaining battery capacity estimated by the battery model based on real-time operating parameters.

[0036] Specifically, the server integrates and runs a battery state observer based on the Extended Kalman Filter (EKF) algorithm. This observer pre-builds a second-order RC equivalent circuit model that describes the electrochemical characteristics of the battery. The server uses the current value from the real-time operating parameters obtained in step S101 as the control input to the observer and the voltage value as the measurement output. The EKF algorithm tracks changes in the battery's internal state in real time by executing iterative loops of time updates (prediction) and measurement updates (correction). The battery's state of charge (SOC) is defined as a key component in the algorithm's state vector. In each calculation cycle, the observer outputs an optimal estimate of the current SOC; this estimate is the initial battery SOC value.

[0037] In one possible implementation, a preliminary battery state of charge value is calculated based on real-time operating parameters, specifically including steps S1021-S1025, as follows: Step S1021: Establish a second-order RC equivalent circuit model of the power supply battery and determine the system state vector, which includes the battery state of charge, ohmic internal resistance, and polarization voltage.

[0038] In step S1021, the second-order RC equivalent circuit model refers to a circuit model used to simulate the electrochemical characteristics of a battery. This model consists of an ideal voltage source, an ohmic internal resistance, and two sets of parallel resistor-capacitor networks. The system state vector represents a column vector containing all key state variables describing the battery's internal structure and is the core processing object of the extended Kalman filter algorithm. The battery state of charge represents the percentage of the battery's current remaining capacity relative to its total nominal capacity. The ohmic internal resistance refers to the equivalent resistance corresponding to the instantaneous voltage drop generated by the electrode materials, electrolyte, etc., when current flows through the battery. The polarization voltage refers to the additional voltage formed on the battery electrode surface due to electrochemical polarization and concentration polarization effects under the action of current; this voltage slowly builds up or dissipates over time.

[0039] Specifically, the server pre-constructs a second-order RC equivalent circuit model of the power supply battery. This model is based on extensive charge-discharge tests of the same battery model conducted in a laboratory environment. Key parameters in the model were identified through experimental methods such as mixed pulse power characteristic testing. These parameters include: the nonlinear functional relationship between the battery's state of charge and open-circuit voltage, the nominal value of the ohmic internal resistance, and the resistance and capacitance values ​​of two sets of RC networks representing fast and slow polarization processes, respectively. Based on this model, the server determines the system state vector that needs to be estimated online. This system state vector includes at least the battery's state of charge, the first polarization voltage, and the second polarization voltage. To improve model accuracy, the server also incorporates the ohmic internal resistance into the system state vector for online correction during battery aging. Therefore, the final determined system state vector consists of four components: the battery's state of charge, the first polarization voltage, the second polarization voltage, and the ohmic internal resistance.

[0040] Step S1022: Discretize the second-order RC equivalent circuit model and construct the state transition equation and observation equation of the extended Kalman filter. The state transition equation is used to describe the change of the system state vector over time.

[0041] In step S1022, discretization refers to the process of converting the differential equations describing the continuous-time behavior of a physical system into difference equations suitable for iterative computation by a computer at discrete time steps. Extended Kalman filtering is an efficient recursive filtering algorithm suitable for state estimation of nonlinear systems. The state transition equation is a mathematical expression that describes the evolution of the system's state vector from the previous time step to the current time step. The observation equation is a mathematical expression that establishes the relationship between the system's internal, non-measurable state vectors and external, directly measurable physical quantities.

[0042] Specifically, the server discretizes the second-order RC equivalent circuit model and its differential equations established in step S1021. Using numerical integration methods, such as the first-order forward Euler method, the server transforms the differential equations describing the continuous changes of the battery state of charge, first polarization voltage, and second polarization voltage over time into a set of discrete-time difference equations. These difference equations together constitute a nonlinear state transition equation. This state transition equation uses the current input current as input and the system state vector from the previous time step as initial conditions to calculate the predicted value of the system state vector at the current time step. Simultaneously, the server establishes an observation equation based on Kirchhoff's voltage law for the circuit. This observation equation correlates the battery state of charge, polarization voltage, ohmic internal resistance, battery open-circuit voltage, input current, and the ultimately measurable terminal voltage in the system state vector, forming a nonlinear function describing the relationship between the terminal voltage and the system state vector.

[0043] Step S1023: Based on the posterior state estimate and state transition equation at the first time, calculate the prior state estimate and prior covariance matrix at the second time. The first time is the time before the second time, and the second time is the current time.

[0044] In step S1023, the first time point refers to the time of the previous iteration cycle of the Extended Kalman Filter (EPF) algorithm. The second time point refers to the time of the current iteration cycle of the EPF algorithm. The posterior state estimate is the optimal state estimate obtained after correction by incorporating the measured values ​​at the first time point. The prior state estimate is the state estimate obtained by predicting solely through the state transition equation at the second time point, without correction by incorporating the measured values ​​at the second time point. The prior covariance matrix is ​​a square matrix used to quantify the degree of uncertainty of the prior state estimate.

[0045] Specifically, the server executes the prediction step of the extended Kalman filter algorithm. First, the server reads the posterior state estimate and its corresponding posterior error covariance matrix, calculated and stored at the first time step, from memory. The server then substitutes the posterior state estimate from the first time step, along with the input current value measured between the first and second time steps, into the state transition equation constructed in step S1022 for calculation. The result is the prior state estimate for the second time step. Simultaneously, the server calculates the Jacobian matrix of the state transition equation with respect to the system state vector, and using this Jacobian matrix, the posterior error covariance matrix from the first time step, and a preset process noise covariance matrix, calculates the prior covariance matrix for the second time step using the state covariance prediction formula. This matrix represents the propagation and accumulation of uncertainty in state prediction over time steps.

[0046] Step S1024: Using the real-time operating parameters as observations, calculate the Kalman gain, and combine the prior state estimate with the observations to correct the system state vector, thereby obtaining the posterior state estimate at the second time step.

[0047] In step S1024, the observed value refers to the physical quantity actually measured by the sensor at the second time point, specifically the terminal voltage of the power supply battery. The Kalman gain is a dynamic weighting coefficient that balances the predicted and observed values, determining the extent to which the new observation is trusted.

[0048] Specifically, the server executes the update step of the extended Kalman filter algorithm. The server extracts the battery terminal voltage at the second moment from the real-time operating parameters obtained in step S101, using it as the observation value for the current cycle. The server substitutes the prior state estimate at the second moment calculated in step S1023 into the observation equation constructed in step S1022 to calculate a theoretical predicted terminal voltage value. The server calculates the Jacobian matrix of the observation equation with respect to the system state vector, and combines it with the prior covariance matrix at the second moment and a preset measurement noise covariance matrix to calculate the Kalman gain at the current moment. Subsequently, the server calculates the difference between the actually observed terminal voltage and the theoretically predicted terminal voltage, i.e., the innovation. The server multiplies this innovation by the Kalman gain to obtain a correction, and adds this correction to the prior state estimate to obtain the posterior state estimate at the second moment after observation correction.

[0049] Step S1025: Extract the battery state of charge component from the posterior state estimate as the preliminary battery state of charge value.

[0050] In step S1025, the battery state of charge component refers to the numerical element in the multi-dimensional posterior state estimation vector that specifically represents the battery state of charge. The preliminary battery state of charge value is the best estimate of the battery state of charge at the current moment obtained after complete prediction and update cycles of the extended Kalman filter, and this value will be used as the input for subsequent processing steps.

[0051] Specifically, in step S1024, the server has calculated a complete posterior state estimation vector containing four components. These four components correspond to the battery state of charge, the first polarization voltage, the second polarization voltage, and the ohmic internal resistance, respectively, in a predetermined order. The server directly accesses the first element of this vector and extracts its value. This extracted value is the most accurate battery state of charge estimation result at the current moment. The server marks this value as the preliminary battery state of charge value and temporarily stores or outputs it for subsequent calls and processing by other parts of the system, such as the dynamic calibration decision module or the data display interface.

[0052] Step S103: Obtain historical operating parameters and historical environmental parameters of meteorological equipment, and construct a geographic feature vector based on the historical operating parameters and historical environmental parameters.

[0053] In step S103, historical operating parameters refer to the sequence of voltage, current, and temperature data recorded by the power supply battery over a period of time, such as several months or years. Historical environmental parameters refer to the sequence of environmental condition data recorded by meteorological equipment within the same time period. Geographic feature vectors refer to numerical vectors extracted through analysis of historical data that characterize the energy budget and environmental change characteristics of a specific geographical location.

[0054] Specifically, the server stores historical operating parameters and environmental parameters uploaded by all connected meteorological devices in its internal database, and establishes an independent data archive for each device. The server can execute an offline analysis task, retrieving historical data from the database for at least one full year for a specified meteorological device. The server performs statistical analysis on this massive amount of data, such as calculating statistical indicators in historical environmental parameters, including the annual average daily sunshine duration, variance of light intensity distribution, seasonal average temperature, and diurnal temperature range. Simultaneously, the server combines historical operating parameters to analyze and calculate the battery's average charging efficiency, average daily self-discharge rate, and drift between full charge and cutoff voltages in different seasons and temperature ranges. The server quantifies these statistical indicators that reflect the unique climatic conditions and energy characteristics of the station and combines them into a multi-dimensional numerical vector, which represents the geographical feature vector of the meteorological station's location.

[0055] In one possible implementation, a geographic feature vector is constructed based on historical operating parameters and historical environmental parameters, specifically including steps S1031-S1034, as follows: Step S1031: Perform timestamp alignment and data cleaning on historical operating parameters and historical environmental parameters to obtain a standardized historical dataset.

[0056] In step S1031, timestamp alignment refers to the process of unifying multiple time-series data from different sources and with varying sampling frequencies onto a common time coordinate axis through methods such as interpolation or aggregation. Data cleaning refers to the process of identifying, correcting, or removing erroneous, missing, abnormal, or duplicate data in the dataset to improve data quality. Standardized historical datasets are used to represent a preprocessed, uniformly formatted, clean historical database that can be directly used for subsequent feature extraction.

[0057] Specifically, the server first retrieves historical operating parameters of the specified meteorological stations from the database, such as battery voltage, current, and capacity, as well as historical environmental parameters, such as ambient temperature, solar radiation intensity, and humidity. Since these parameters may be recorded by different sensors at different time intervals, the server needs to set a unified time benchmark, such as one recording point every 10 minutes. For parameters with missing data at a certain time point, the server uses methods such as linear interpolation or spline interpolation to fill in the gaps; for data with excessively dense recording points within a certain time period, the server uses downsampling aggregation by calculating the average or median. After alignment, the server cleans the data. The server identifies and marks outliers by setting reasonable physical thresholds or using statistical methods, such as the 3-sigma principle, and replaces them with the mean or median of neighboring points. For long periods of missing data that cannot be repaired, the server may choose to remove that segment of data. After the above processing, all parameters are organized into a structured table with strictly aligned timestamps, free of obvious anomalies and missing data, forming a standardized historical dataset.

[0058] Step S1032: Extract multi-dimensional statistical features from the standardized historical dataset to form a preliminary feature set.

[0059] In step S1032, multidimensional statistical features refer to derived numerical indicators calculated from the original time series data, which can summarize the inherent laws and characteristics of the data from different perspectives and time scales. The preliminary feature set refers to the set formed by bringing together all the extracted multidimensional statistical features, which will serve as the input for subsequent dimensionality reduction processing.

[0060] Specifically, the server performs deep feature engineering based on the standardized historical dataset obtained in step S1031. The server executes computational tasks to extract multi-dimensional statistical features. For example, to capture the long-term characteristics of temperature, the server groups the ambient temperature data by month and calculates the average temperature for each month as the monthly mean and the temperature variance for each month as the monthly variance. To quantify the seasonal patterns of solar energy resources, the server divides the year into four seasons: spring, summer, autumn, and winter, and calculates the average solar intensity within each season, thus obtaining a quantitative index of seasonal fluctuation patterns. To reflect the correlation between battery performance and the environment, the server first defines multiple temperature ranges, such as below 0 degrees Celsius, 0 to 25 degrees Celsius, and above 25 degrees Celsius. Then, the server filters all charging process segments occurring within each temperature range from the standardized historical dataset, calculates the charging efficiency for each segment, and finally calculates the average of all charging efficiencies within each temperature range to obtain the average charging efficiency under different temperature ranges. The server will combine all such statistical features calculated, such as monthly mean, monthly variance, seasonal fluctuation pattern indicators, and segmented average charging efficiency, into a high-dimensional feature set, i.e., the preliminary feature set.

[0061] Step S1033: Based on multi-dimensional statistical features, divide the preliminary feature set into multiple sub-components and calculate the cumulative variance contribution rate of each sub-component.

[0062] In step S1033, a sub-component refers to a new set of mutually orthogonal coordinate axes or variables obtained by linearly transforming the initial feature set using principal component analysis. Each sub-component is a linear combination of the original features. The cumulative variance contribution rate refers to the percentage of the total variance of the original data that can be explained by the first N sub-components, and is used to measure the completeness of information retention.

[0063] Specifically, the server applies principal component analysis (PCA) to the preliminary feature set generated in step S1032. First, the server centers and standardizes each feature dimension in the preliminary feature set, subtracting the mean and dividing by the standard deviation to eliminate the influence of differences in dimensions and numerical ranges between different features. After processing, the server calculates the covariance matrix of the standardized feature matrix. Next, the server performs eigenvalue decomposition on the covariance matrix, obtaining a set of eigenvalues ​​and their corresponding eigenvectors. Each eigenvector defines the direction of a sub-component, and its corresponding eigenvalue represents the variance of the original data in that sub-component direction. The server sorts all eigenvalues ​​from largest to smallest and calculates the variance contribution rate of each sub-component, which is equal to the corresponding eigenvalue divided by the sum of all eigenvalues. Finally, the server accumulates the variance contribution rates starting from the largest, calculating a sequence of cumulative variance contribution rates for each sub-component, including 1, 2, and so on, up to all sub-components.

[0064] Step S1034: Select principal components whose cumulative variance contribution rate exceeds a preset threshold from the preliminary feature set, and linearly combine the principal components into a geographic feature vector.

[0065] In step S1034, the preset threshold is a manually set standard used to determine how much information is sufficient to represent the original data; it is usually set to a relatively high value, such as 0.95. Principal components are the top K sub-components selected from all sub-components whose cumulative variance contribution rate exactly meets the preset threshold. The geographic feature vector is a low-dimensional vector composed of a few principal components, capable of highly condensing the essence of the original high-dimensional geographic environment and operational condition information.

[0066] Specifically, the server compares the cumulative variance contribution rate sequence calculated in step S1033 with an internally set preset threshold, such as 0.95. Starting with the first principal component, the server checks the cumulative variance contribution rate sequentially until it finds a minimum K value, such that the cumulative variance contribution rate of the first K principal components is greater than or equal to 0.95 for the first time. These first K components are then selected as the final principal components. Next, the server uses the feature vectors corresponding to these K principal components as new basis vectors and projects the standardized preliminary feature set from step S1033 onto a low-dimensional space spanned by these K principal components through matrix multiplication. This projection process is essentially a weighted linear combination of the original features, with the weights being the elements in the feature vector. Each data point obtained after projection is a K-dimensional vector, which is the geographical feature vector that can summarize the long-term environmental and operational characteristics of the site in the simplest form.

[0067] Step S104: Determine the triggering conditions based on the geographic feature vector, and determine whether the environmental parameters meet the triggering conditions.

[0068] In step S104, the triggering condition is used to represent one or more preset environmental states. When the real-time environmental parameters meet these states, the dynamic calibration process is started.

[0069] Specifically, the server adaptively determines dynamically calibrated trigger conditions for each meteorological device based on the geographic feature vector constructed in step S103. For example, if the geographic feature vector of a site shows that its afternoon sunlight is sufficient and stable in summer, the server generates a trigger condition for that site: light intensity greater than 800 watts per square meter for 20 consecutive minutes and battery temperature between 20 and 30 degrees Celsius. This trigger condition is stored in the server and associated with the device. During real-time monitoring, the server continuously compares the real-time environmental parameters obtained in step S101 with the preset trigger conditions. If the server detects that the current light intensity is 950 watts per square meter and the battery temperature is 25 degrees Celsius, and this state has lasted for more than 20 minutes, the server determines that the current environmental parameters meet the preset trigger conditions.

[0070] Step S105: If it is determined that the environmental parameters meet the triggering conditions, then based on the geographic feature vector and the preliminary battery state of charge value, perform dynamic calibration to calculate the state of charge correction value.

[0071] In step S105, dynamic calibration refers to the process of actively performing a specific operation to measure the battery's true response when trigger conditions are met, thereby calculating the state of charge (SOC) deviation. The SOC correction value represents the amount of adjustment required to the initial SOC value of the battery, calculated through dynamic calibration.

[0072] Specifically, once the server determines that the environmental parameters meet the triggering conditions, it immediately initiates the dynamic calibration process. The server sends a precise control command to the power management unit of the target meteorological equipment. This command is generated based on the calibration strategy indicated by the geographic feature vector; for example, it instructs the weather station to disconnect the solar charging circuit and discharge the battery with a constant current of 0.5 amps for 60 seconds. During calibration, the server acquires and records detailed changes in the battery terminal voltage at a higher data request frequency (e.g., 10 times per second). After calibration, the server sends a recovery command to restore the meteorological equipment to its normal charging and discharging state. Subsequently, based on the preliminary battery state of charge value before calibration, the server calculates a theoretical voltage drop using a battery model. The server compares this theoretical voltage drop with the voltage drop actually measured during calibration. The difference directly reflects the degree of deviation in the preliminary estimate. The server uses a preset nonlinear mapping function to convert this voltage difference into a specific state of charge correction value, such as -1.2%, which indicates that the preliminary estimate is 1.2% higher than the actual value.

[0073] Please refer to Figure 2In one possible implementation, dynamic calibration is performed based on the geographic feature vector and the preliminary battery state of charge value to calculate the corrected state of charge value, specifically including steps S201-S205, as follows: Step S201: In the preset multidimensional mapping model, the geographic feature vector and the preliminary battery state of charge value are used as a joint index to query and obtain the calibration parameter set and the reference voltage charge response slope corresponding to the joint index; wherein, the calibration parameter set includes calibration current and calibration duration.

[0074] In step S201, the preset multidimensional mapping model represents a pre-built data structure, such as a multidimensional lookup table or a trained neural network, which stores ideal calibration strategies for different geographical environments and battery states. The joint index refers to concatenating or combining geographical feature vectors and preliminary battery state-of-charge values ​​to form a composite key for unique retrieval within the preset multidimensional mapping model. The calibration parameter set refers to a set of specific instruction parameters used to perform active perturbation operations. The reference voltage-charge response slope refers to the ratio of the voltage change to the corresponding charge change of an ideal healthy battery under specific geographical environments and charge states, serving as a theoretical reference benchmark.

[0075] Specifically, the server first obtains the geographic feature vector calculated in the previous stage, as well as the preliminary battery state-of-charge (SOC) value reported by the battery management system. The server combines these two input data sets, for example, concatenating the 7-dimensional geographic feature vector with the 1-dimensional SOC value into an 8-dimensional joint index vector. Subsequently, the server inputs this joint index vector into a pre-defined multi-dimensional mapping model stored in the server's memory. Based on the input joint index, the model outputs the set of data that best matches the current operating conditions through table lookup or model inference. This set of data contains two parts: one part is the calibration parameter set, specifying the calibration current magnitude (e.g., 2 amps) and the calibration duration (e.g., 90 seconds); the other part is the reference voltage charge response slope (e.g., 0.02 volts per ampere-hour).

[0076] Step S202: Perform a charge and discharge operation on the power supply battery according to the calibration current and calibration duration to apply a disturbance charge and measure the voltage change caused by the disturbance charge.

[0077] In step S202, the disturbance charge refers to the minute amount of charge injected into or extracted from the battery through a brief, precisely controlled charge-discharge operation, used to actively probe the battery's current response characteristics. The voltage change refers to the difference between the battery terminal voltage before and after the disturbance charge is applied, after the voltage has stabilized.

[0078] Specifically, the server sends a command to the battery management system (BMS) of the power supply battery. This command includes the calibration parameters obtained in step S201, namely, charging or discharging the battery with a calibration current of 2 amps for a calibration duration of 90 seconds. Before executing this command, the BMS accurately measures and records the current battery terminal voltage. Then, the BMS controls the power conversion circuit to strictly execute the charging and discharging operation according to the command. After the operation is completed, the BMS disconnects the charging and discharging circuit, allowing the battery to rest for a preset time, such as 5 minutes, to eliminate the effects of polarization, and then accurately measures and records the stabilized battery terminal voltage again. The server obtains these two voltage measurements from the BMS and calculates the difference between them to obtain the voltage change. At the same time, the server calculates the amount of disturbance charge applied in this operation according to the formula: disturbance charge = calibration current × calibration duration.

[0079] Step S203: Calculate the actual voltage-charge response slope based on the amount of disturbance charge and the amount of voltage change.

[0080] In step S203, the actual voltage-charge response slope refers to the true ratio of the change in battery voltage to the corresponding change in charge obtained by active perturbation experiment under the current actual operating conditions. This value reflects the current health status and electrochemical characteristics of the battery.

[0081] Specifically, the server uses the data obtained in step S202 for calculation. The server uses the measured voltage change as the numerator and the calculated disturbance charge as the denominator, performing a division operation. For example, if the measured voltage change is 0.0011 volts and the calculated disturbance charge is 0.05 ampere-hours, then the actual voltage-charge response slope calculated by the server is 0.0011 / 0.05 = 0.022 volts per ampere-hour. This calculation result is the actual voltage-charge response slope.

[0082] Step S204: Compare the actual voltage charge response slope with the reference voltage charge response slope to obtain the slope deviation.

[0083] In step S204, the slope deviation refers to the absolute difference between the actual measured voltage-charge response slope and the theoretical reference value, which is used to quantify the degree to which the current characteristics of the battery deviate from the ideal state.

[0084] Specifically, the server performs a subtraction operation between the actual voltage-charge response slope calculated in step S203 and the reference voltage-charge response slope retrieved from the preset multidimensional mapping model in step S201. For example, the server subtracts the reference value of 0.02 volts per ampere-hour from the actual value of 0.022 volts per ampere-hour, and the result of 0.002 volts per ampere-hour is the slope deviation.

[0085] Step S205: Divide the slope deviation by the reference voltage charge response slope to obtain the slope deviation ratio.

[0086] In step S205, the slope deviation ratio refers to the percentage or ratio of the slope deviation relative to the reference slope, providing a standardized, dimensionless measure of deviation.

[0087] Specifically, the server divides the slope deviation calculated in step S204 by the reference voltage charge response slope obtained in step S201. For example, the server divides the slope deviation of 0.002 volts per ampere by the reference slope of 0.02 volts per ampere, and the result of 0.1 is the slope deviation ratio.

[0088] Step S206: Multiply the slope deviation ratio by the preset slope sensitivity coefficient to obtain the state of charge deviation ratio.

[0089] In step S206, the preset slope sensitivity coefficient represents a pre-calibrated conversion factor used to establish a quantitative relationship between the change in the voltage-charge response slope and the battery state-of-charge estimation error. The state-of-charge deviation ratio refers to the relative error of the state-of-charge estimation calculated based on changes in the battery's electrochemical characteristics.

[0090] Specifically, the server reads a preset slope sensitivity coefficient from the system configuration. This coefficient is typically determined through extensive offline experiments and model simulations, for example, -0.5. The server multiplies the slope deviation ratio calculated in step S205 by this slope sensitivity coefficient. For example, the server multiplies a slope deviation ratio of 0.1 by -0.5, and the result -0.05 is the state-of-charge deviation ratio.

[0091] Step S207: Multiply the state-of-charge deviation ratio by the nominal battery capacity to obtain the initial state-of-charge deviation value.

[0092] In step S207, the nominal capacity of the battery refers to the total amount of charge that the battery can provide under standard conditions, as defined at the factory, and is a fixed parameter. The initial state-of-charge deviation value refers to the absolute charge deviation converted from the relative state-of-charge deviation ratio into ampere-hours.

[0093] Specifically, the server retrieves the nominal capacity of the battery from the battery specification database, for example, 100 amp-hours. The server multiplies the state-of-charge (SOC) deviation ratio calculated in step S206 by the nominal capacity of the battery. For example, the server multiplies the SOC deviation ratio of -0.05 by 100 amp-hours, and the result of -5 amp-hours is the initial SOC deviation value.

[0094] Step S208: Using a preset correction function, the initial state of charge deviation value is mapped and limited to obtain the state of charge correction value.

[0095] In step S208, the preset correction function represents a mathematical function used to smooth and constrain the correction amount, aiming to prevent excessive single correction from causing system oscillations and ensuring the stability of the calibration process. The state of charge correction value refers to the final correction amount used to adjust the initial battery state of charge value after processing by the preset correction function.

[0096] Specifically, the server takes the initial state-of-charge deviation value calculated in step S207, for example -5 amp-hours, as input and processes it into an internally configured preset correction function. This function may be a function with gain and saturation limits. For example, the function may multiply the input value by a smoothing factor of 0.9 and set a maximum correction upper limit of 5% of the battery's nominal capacity and a lower limit of -5%. In this example, -5 amp-hours × 0.9 = -4.5 amp-hours, which is within the limit of ±5 amp-hours, so the function outputs -4.5 amp-hours. This final output of -4.5 amp-hours is the state-of-charge correction value calculated in this dynamic calibration.

[0097] Step S106: Combine the corrected state of charge value with the preliminary battery state of charge value to obtain the final battery state of charge value.

[0098] In step S106, the final battery state-of-charge value refers to the more accurate percentage of remaining battery capacity obtained by combining the preliminary estimate with the correction amount obtained from dynamic calibration.

[0099] Specifically, the server employs a weighted fusion algorithm based on evidence theory to synthesize the preliminary battery state of charge (SOC) value obtained in step S102 with the SOC correction value calculated in step S105. In the short period immediately following dynamic calibration, the SOC correction value, being based on actual measurements, has extremely high confidence; therefore, the server assigns it a very high weight in the fusion algorithm. For example, the server synthesizes the preliminary SOC value of 85.0% with the SOC correction value of -1.2%, obtaining a final SOC value of 83.8% through weighted calculation. The server stores this final SOC value as the most accurate result for the current battery level and displays it on the user monitoring interface. More importantly, the server uses this calibrated final value to reset the initial SOC value of the extended Kalman filter observer in step S102, thereby eliminating accumulated errors and significantly improving the accuracy of subsequent estimations.

[0100] In one possible implementation, the corrected state of charge (SOC) value is combined with the preliminary SOC value to obtain the final SOC value, specifically including steps S1061-S1065, as follows: Step S1061: Use the preliminary battery state of charge value as the first source of evidence and the corrected state of charge value as the second source of evidence.

[0101] In step S1061, the first source of evidence refers to the initial information from the battery management system's own estimation model, which has a certain degree of uncertainty. The second source of evidence refers to the incremental information obtained through the active dynamic calibration process and used to correct the first source of evidence, which also has its own uncertainty.

[0102] Specifically, the server marks the initial battery state-of-charge (SOC) value reported by the battery management system, such as a value expressed as 60%, as the first source of evidence in the data processing logic. Simultaneously, the server marks the SOC correction value calculated in previous steps S201-S208, such as a charge value expressed as -4.8 amp-hours, as the second source of evidence. For consistent processing, the server converts the SOC correction value to the same unit as the initial battery SOC value. For example, if the battery's nominal capacity is 100 amp-hours, then -4.8 amp-hours is equivalent to a -4.8% SOC correction. In this case, the second source of evidence can be understood as an independent judgment of the initial SOC value, that is, assuming the true value should be around 60% - 4.8% = 55.2%.

[0103] Step S1062: Based on the uncertainty of the preliminary battery state of charge value and the uncertainty of the corrected state of charge value, construct a first probability allocation function for the first evidence source and a second probability allocation function for the second evidence source.

[0104] In step S1062, uncertainty refers to a quantitative assessment of the accuracy of the information provided by the evidence source, which typically stems from factors such as sensor error, model aging, and environmental changes. The probability assignment function, also known as the basic probability assignment or confidence assignment, is the process of assigning a total confidence level of 1 to various subsets within an identification framework, within the framework of evidence theory, to express the degree of support for different hypotheses.

[0105] Specifically, the server first evaluates the uncertainty of each of the two evidence sources. For the first evidence source, its uncertainty, for example, 10%, may be provided by the internal state assessment module of the battery management system, reflecting the combined effects of the cumulative error of the ampere-hour integration method and the model error of the open-circuit voltage method. For the second evidence source, its uncertainty, for example, 5%, is determined by the accuracy of the dynamic calibration process and is related to factors such as the measurement accuracy of the voltage and current sensors and the precision of the disturbance control. The server then defines an identification framework, which is the set of all possible values ​​of the battery state of charge, for example, from 0% to 100%. Next, the server constructs a first probability assignment function for the first evidence source: assigning a confidence level of 1 - 0.1 = 0.9 to the proposition "the state of charge is 60%", and assigning a confidence level of 0.1 to the entire identification framework representing complete uncertainty. Similarly, a second probability assignment function is constructed for the second evidence source: assigning a confidence level of 1 - 0.05 = 0.95 to the proposition "the state of charge is 55.2%", and assigning a confidence level of 0.05 to the entire identification framework.

[0106] Step S1063: Merge the first probability allocation function and the second probability allocation function to obtain the fused probability allocation function.

[0107] In step S1063, the fusion probability allocation function refers to the new probability allocation function obtained by merging the probability allocation functions from multiple independent evidence sources through the Dempster combination rule. This new function integrates the information of all evidence and can handle conflicts between evidence.

[0108] Specifically, the server applies Dempster's combination rule to merge the first and second probability assignment functions constructed in step S1062. The server calculates the product of all possible basic probability assignment combinations. For example, it multiplies the 0.9 confidence level from the first source of evidence supporting "a state of charge of 60%" with the 0.95 confidence level from the second source of evidence supporting "a state of charge of 55.2%". Since these two propositions conflict, this product contributes to the conflicting term. The server systematically calculates all combinations and assigns confidence levels to the intersection of each proposition. The sum of the confidence levels of all conflicting terms is used to calculate a normalization factor. Finally, the server sums the confidence products of all non-conflicting combinations and adjusts them with this normalization factor to obtain a new fusion probability assignment function that integrates information from both sources of evidence. This new function typically assigns higher confidence levels to regions commonly supported by both original pieces of evidence.

[0109] Step S1064: Use a preset time decay factor to weight the fusion probability allocation function to obtain a weighted fusion probability allocation.

[0110] In step S1064, the preset time decay factor is a coefficient that decreases over time, used to reduce the weight of historical data in the current decision, reflecting the principle of information timeliness. Weighted fusion probability allocation refers to combining the current fusion result with historical results weighted by time decay to form a dynamically updated, time-continuous probability allocation.

[0111] Specifically, the server maintains the weighted fusion probability allocation calculated in the previous time step in memory. After the new fusion probability allocation function is generated in step S1063, the server obtains the time difference between the current time and the last update time. Based on this time difference and a preset decay constant, the server calculates a time decay factor, for example, 0.8. The server multiplies all confidence values ​​in the weighted fusion probability allocation from the previous time step by this decay factor of 0.8 to obtain a decayed historical probability allocation. Then, the server performs a weighted average of this decayed historical probability allocation and the newly generated fusion probability allocation function. The weights can be allocated according to the system design, for example, old and new information each account for 50% of the weight, thus obtaining the latest weighted fusion probability allocation.

[0112] Step S1065: Calculate the final battery state of charge value based on weighted fusion probability allocation.

[0113] In step S1065, the final battery state of charge value refers to the conversion of the weighted fusion probability allocation representing uncertain beliefs into a single, explicit numerical output, which serves as the final conclusion of this estimation.

[0114] Specifically, the server performs decision processing on the weighted fusion probability allocation obtained in step S1064 to extract the most reliable single state of charge (POC) value. One approach is to calculate the pignistic probability, which involves distributing the confidence level assigned to the set equally among each individual element in the set. For example, if a set containing {56%, 57%} has a confidence level of 0.4, then 56% and 57% will each receive a pignistic probability of 0.2. The server repeats this process for all elements in the identification framework to obtain the total pignistic probability for each single POC value. Finally, the server calculates the expected value of all possible POC values ​​by multiplying each POC value by its corresponding total pignistic probability and then summing all the products. This final calculated expected value, for example, 56.1%, is the final battery POC value.

[0115] In one possible implementation, after synthesizing the corrected state of charge (SCC) value with the preliminary SCC value to obtain the final SCC value, the method further includes steps S107-S110, as follows: Step S107: Preprocess the long-term historical data received from multiple terminal devices, wherein the long-term historical data includes historical final battery state of charge values, historical operating parameters, historical environmental parameters, and historical geographic feature vectors.

[0116] In step S107, long-term historical data refers to a collection of battery operation records spanning several months or even years, collected from a large number of terminal devices. Historical final battery state of charge (SOC) values ​​refer to the sequence of final SOC values ​​calculated and stored in the preceding step S1065. Historical operating parameters refer to physical quantities that record the battery's operating state, such as current, voltage, and power. Historical environmental parameters refer to physical quantities of the environment in which the battery is located, such as ambient temperature and battery surface temperature. Historical geographic feature vectors refer to a set of features extracted from geographic location information that can affect battery performance, such as altitude and road slope. Preprocessing refers to a series of operations that clean, transform, and normalize the raw data to improve data quality and model training effectiveness.

[0117] Specifically, the server first extracts long-term historical data from all online terminal devices in the database. Next, the server cleans the data, including identifying and removing outliers caused by sensor malfunctions or communication errors, such as voltage or current values ​​exceeding physical limits. Then, the server handles missing data; for short-term data gaps, linear interpolation or nearest-neighbor imputation can be used; for long-term gaps, the data segment may need to be discarded. Subsequently, the server performs feature engineering, converting the raw geographic coordinate information into historical geographic feature vectors with actual physical meaning, such as calculating altitude and road slope using a digital elevation model. Finally, the server normalizes all numerical features, including historical final battery state of charge values, historical operating parameters, historical environmental parameters, and historical geographic feature vectors, for example, by using a min-max scaling method to map all data to the range of 0 to 1, eliminating the influence of different dimensions between features and ensuring the stability and convergence speed of model training.

[0118] Step S108: Divide the preprocessed long-term historical data into training set, validation set and test set.

[0119] In step S108, the training set refers to the subset of data used for model parameter learning and fitting. The validation set refers to the subset of data used to adjust hyperparameters and conduct preliminary performance evaluation during model training; this subset of data does not participate in updating model parameters. The test set refers to a completely independent subset of data used for the final evaluation of the model's generalization ability after model training is completed.

[0120] Specifically, the server divides the preprocessed long-term historical dataset from step S107 into chronological segments. To ensure the model can predict the future, the segmentation must avoid using future data to train past models. A strict segmentation method is to select an earlier portion of the data records as the training set, for example, the first 70% of the total data time span; then select a middle portion as the validation set, for example, the next 15%; and finally select the most recent portion as the test set, for example, the last 15%. This segmentation method simulates real-world scenarios of using historical data to predict future events, thereby effectively evaluating the model's real-world performance.

[0121] Step S109: Train a global state evolution model using the training set. The global state evolution model is a model based on a long short-term memory network and includes at least one input layer, at least one long short-term memory network hidden layer, and at least one fully connected output layer.

[0122] In step S109, the global state evolution model refers to a deep learning model trained on a cloud server using aggregated data from all terminal devices, capable of describing the general laws governing the state evolution of batteries under various operating conditions. Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network that addresses the vanishing or exploding gradient problems common in traditional recurrent neural networks when processing long sequences by introducing a gating mechanism. They are particularly suitable for learning long-term dependencies in time-series data. The input layer receives external input data. The hidden layers of the LTM network extract and memorize complex dynamic features from the time-series data. The fully connected output layer maps the features extracted by the hidden layers to the final prediction target value.

[0123] Specifically, the server first defines a global state evolution model based on a pre-defined architecture. The number of neurons in the input layer matches the number of preprocessed features; for example, if there are five features—state of charge, current, voltage, temperature, and altitude—the input layer has five nodes. The input layer is then connected to one or more long short-term memory (LSTM) hidden layers to capture the complex nonlinear relationship of battery state evolution over time. The output of the hidden layers is fed into a fully connected output layer with three neurons, corresponding to the reference state of charge, reference internal resistance, and reference open-circuit voltage to be predicted. The server inputs the training set data partitioned in step S108 into the model and continuously adjusts the weight parameters in the network using a backpropagation algorithm and an optimizer such as Adam to minimize the mean squared error between the model's predicted values ​​and the true label values ​​in the training set. Simultaneously, after each training cycle, the server evaluates the model performance using a validation set, adjusts hyperparameters such as the learning rate based on the performance on the validation set, and employs an early stopping strategy to prevent overfitting.

[0124] Step S110: Calculate the state reference value using the global state evolution model and send the state reference value to the corresponding terminal device. The state reference value includes the reference state of charge value, the reference internal resistance value, and the reference open circuit voltage.

[0125] In step S110, the state reference value refers to the ideal state parameters of the device at a future moment, predicted by the global state evolution model based on the historical operating data of a specific terminal device. These reference values ​​represent the "standard" or "expected" battery behavior derived from massive amounts of data. The reference state of charge, reference internal resistance, and reference open-circuit voltage are the three specific components constituting the state reference value.

[0126] Specifically, when a state reference is needed for a specific terminal device, the server receives a sequence of its recent operational data from that device. This sequence contains the same features as the training data. The server performs the same preprocessing procedure on this data as in step S107. Then, the server inputs the preprocessed data sequence into the global state evolution model trained in step S109 for forward propagation calculation. Based on the input historical sequence, the model outputs a prediction for the next time point. This output is a vector containing three values, representing the reference state of charge, reference internal resistance, and reference open-circuit voltage. The server packages these three calculated state reference values ​​and sends them back to the requesting terminal device via a network communication protocol. After receiving these reference values, the terminal device can compare them with its own local estimates to calibrate its local model, diagnose abnormal operating conditions, or assess the health of the battery.

[0127] The following describes a battery state-of-charge measurement device for meteorological equipment from the perspective of hardware processing. Please refer to [link to relevant documentation]. Figure 3 This is a schematic diagram of the structure of a power supply battery state-of-charge measuring device for meteorological equipment in an embodiment of this application.

[0128] It should be noted that, Figure 3 The structure of a battery state-of-charge measuring device for meteorological equipment shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of the present invention.

[0129] like Figure 3As shown, a battery state-of-charge measurement device for meteorological equipment includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage section 308 into a random access memory (RAM) 303, such as performing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for device operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0130] The following components are connected to I / O interface 305: input section 306 including audio input devices, push-button switches, etc.; output section 307 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 308 including a hard disk, etc.; and communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 309 performs communication processing via a network such as the Internet. Drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0131] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs the various functions defined in the present invention.

[0132] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0133] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0134] Specifically, the battery state-of-charge measurement device for meteorological equipment in this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the battery state-of-charge measurement method for meteorological equipment provided in the above embodiment.

[0135] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the state-of-charge (POC) measurement device for a power supply battery of a meteorological device described in the above embodiments; or it may exist independently and not assembled into the state-of-charge (POC) measurement device for a power supply battery of a meteorological device. The storage medium carries one or more computer programs, which, when executed by a processor of the state-of-charge (POC) measurement device for a power supply battery of a meteorological device, cause the device to implement the method for measuring the POC of a power supply battery for a meteorological device based on encrypted data transmission via the Internet of Things (IoT) provided in the above embodiments.

Claims

1. A method for measuring the state of charge of a power supply battery for meteorological equipment, characterized in that, The method includes: The real-time operating parameters of the power supply battery of the meteorological equipment and the environmental parameters of the environment in which the meteorological equipment is located are obtained. Based on the aforementioned real-time operating parameters, a preliminary battery state of charge value is calculated. The historical operating parameters and historical environmental parameters of the meteorological equipment are obtained, and a geographic feature vector is constructed based on the historical operating parameters and the historical environmental parameters; Based on the geographic feature vector, determine the triggering conditions and determine whether the environmental parameters meet the triggering conditions; If it is determined that the environmental parameters meet the triggering conditions, then based on the geographic feature vector and the preliminary battery state of charge value, dynamic calibration is performed to calculate the state of charge correction value; The corrected state of charge value is combined with the preliminary battery state of charge value to obtain the final battery state of charge value.

2. The method according to claim 1, characterized in that, The calculation of the preliminary battery state of charge value based on the real-time operating parameters specifically includes: A second-order RC equivalent circuit model of the power supply battery is established, and the system state vector is determined. The system state vector includes the battery state of charge, ohmic internal resistance, and polarization voltage. The second-order RC equivalent circuit model is discretized to construct the state transition equation and observation equation of the extended Kalman filter, wherein the state transition equation is used to describe the change of the system state vector over time. Based on the posterior state estimate and state transition equation at the first time point, the prior state estimate and prior covariance matrix at the second time point are calculated. The first time point is the time before the second time point, and the second time point is the current time point. Using the real-time operating parameters as observations, the Kalman gain is calculated, and the system state vector is corrected by combining the prior state estimate with the observations to obtain the posterior state estimate at the second time step. The battery state of charge component is extracted from the posterior state estimate as the preliminary battery state of charge value.

3. The method according to claim 1, characterized in that, The construction of the geographic feature vector based on the historical operating parameters and the historical environmental parameters specifically includes: The historical operating parameters and historical environmental parameters are timestamped and cleaned to obtain a standardized historical dataset. Multi-dimensional statistical features are extracted from the standardized historical dataset to form a preliminary feature set; Based on the aforementioned multidimensional statistical features, the preliminary feature set is divided into multiple sub-components, and the cumulative variance contribution rate of each sub-component is calculated. Principal components whose cumulative variance contribution rate exceeds a preset threshold are selected from the preliminary feature set, and the principal components are linearly combined into the geographic feature vector.

4. The method according to claim 1, characterized in that, The step of performing dynamic calibration based on the geographic feature vector and the preliminary battery state of charge value to calculate the corrected state of charge value specifically includes: In the preset multidimensional mapping model, the geographic feature vector and the preliminary battery state of charge value are used as a joint index to query and obtain the calibration parameter set and the reference voltage charge response slope corresponding to the joint index; wherein, the calibration parameter set includes calibration current and calibration duration; Based on the calibration current and the calibration duration, a charge-discharge operation is performed on the power supply battery to apply a perturbation charge, and the voltage change caused by the perturbation charge is measured. Based on the amount of disturbance charge and the amount of voltage change, the actual voltage charge response slope is calculated; The slope deviation is obtained by comparing the actual voltage-charge response slope with the reference voltage-charge response slope. Based on the slope deviation, the state of charge correction value is calculated using a preset correction function.

5. The method according to claim 4, characterized in that, The step of calculating the state of charge correction value based on the slope deviation using a preset correction function specifically includes: Divide the slope deviation by the reference voltage charge response slope to obtain the slope deviation ratio; Multiply the slope deviation ratio by a preset slope sensitivity coefficient to obtain the state of charge deviation ratio; Multiply the state of charge deviation ratio by the nominal battery capacity to obtain the initial state of charge deviation value; The initial state of charge deviation value is mapped and limited using the preset correction function to obtain the state of charge correction value.

6. The method according to claim 1, characterized in that, The step of combining the corrected state of charge value with the preliminary battery state of charge value to obtain the final battery state of charge value specifically includes: The preliminary battery state of charge value is used as the first source of evidence, and the corrected state of charge value is used as the second source of evidence. Based on the uncertainty of the preliminary battery state of charge value and the uncertainty of the corrected state of charge value, a first probability allocation function is constructed for the first evidence source, and a second probability allocation function is constructed for the second evidence source; The first probability allocation function and the second probability allocation function are fused to obtain the fused probability allocation function; The fusion probability allocation function is weighted using a preset time decay factor to obtain a weighted fusion probability allocation; The final battery state of charge value is calculated based on the weighted fusion probability allocation.

7. The method according to claim 1, characterized in that, After synthesizing the corrected state of charge value with the preliminary battery state of charge value to obtain the final battery state of charge value, the method further includes: Preprocessing is performed on long-term historical data received from multiple terminal devices, wherein the long-term historical data includes historical final battery state of charge values, historical operating parameters, historical environmental parameters, and historical geographic feature vectors; The preprocessed long-term historical data is divided into training set, validation set and test set; The training set is used to train a large global state evolution model, wherein the large global state evolution model is a model based on a long short-term memory network and includes at least one input layer, at least one long short-term memory network hidden layer and at least one fully connected output layer. Using the global state evolution model, a state reference value is calculated and sent to the corresponding terminal device. The state reference value includes a reference state of charge value, a reference internal resistance value, and a reference open-circuit voltage.

8. A device for measuring the state of charge of a power supply battery for meteorological equipment, characterized in that, The battery state of charge measurement device for meteorological equipment includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the battery state of charge measurement device for meteorological equipment to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the power supply battery state-of-charge measuring device for meteorological equipment, the power supply battery state-of-charge measuring device for meteorological equipment performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the power supply battery state-of-charge measuring device for meteorological equipment, the power supply battery state-of-charge measuring device for meteorological equipment performs the method as described in any one of claims 1-7.

Images