Energy storage data processing method and system based on digital twinning

Abstract

The invention discloses an energy storage data processing algorithm and system based on digital twinning. The algorithm comprises the following steps: step 1, collecting key operation data of an energy storage system by using a sensor and a data acquisition system; step 2, constructing a digital twin model based on the key operation data and the physical space in the step 1; 3, based on the model constructed in the step 2, performing data analysis and processing on the equivalent circuit model to generate a data analysis result; 4, based on a result generated in the step 3, generating an optimization suggestion of operation and maintenance, providing a user interface, and displaying an analysis result and the optimization suggestion; 5, designing a feedback mechanism based on the suggestions generated in the step 4, and feeding back a system operation result and a user operation decision to the digital twin model for continuous optimization and fine adjustment of the model; through continuous learning and environment change adaptation, the performance and adaptability of the algorithm are continuously improved. The operation performance and the management efficiency of the energy storage system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to energy management and energy storage system optimization technology, and in particular to an energy storage data processing algorithm and system based on digital twin technology applied in power systems, aiming to improve the operating efficiency and reliability of energy storage systems, and is particularly suitable for the efficient utilization of renewable energy and the stable operation of power grids. Background Art

[0002] In recent years, with the rapid development of technologies such as the Internet of Things (IoT), cloud computing, and machine learning and artificial intelligence, the global energy industry is gradually moving towards a more digital and interconnected future. In this evolving energy system, large-scale access to renewable energy sources such as wind and solar energy has become a trend. These energy sources are highly unstable and unpredictable due to their dependence on natural conditions, which poses challenges to the stable operation of the power grid. Energy storage technology has therefore become one of the key technologies to solve this problem. It can store energy when there is excess energy and release it when needed to balance supply and demand. However, existing energy storage system management methods usually rely on traditional data processing technologies, which often cannot fully predict and respond to real-time state changes of the system, resulting in low energy utilization efficiency and failure to maximize the economic and environmental benefits of the energy storage system. In addition, traditional methods face the problems of slow processing speed and low accuracy when processing large-scale data, which makes it difficult to meet the needs of modern power grids. Summary of the invention

[0003] The purpose of the present invention is to solve the problems of insufficient real-time performance and accuracy in existing energy storage data processing technologies, and to provide an energy storage data processing algorithm based on digital twin technology. The algorithm aims to achieve real-time and accurate monitoring and prediction of the state of the energy storage system through efficient data synchronization, precise model mapping and advanced analysis technology, thereby optimizing the management and operation efficiency of the energy storage system. Specifically, the purposes of the present invention include:

[0004] Improve the real-time performance of data processing: Through real-time data collection and instant feedback mechanisms, ensure that the operating data of the energy storage system can be quickly processed and analyzed so as to respond to changes in system status in a timely manner.

[0005] Enhance the accuracy of data processing: Use high-precision models built using digital twin technology to improve the accuracy of data analysis, ensure more accurate status assessment and prediction of energy storage systems, and reduce errors.

[0006] Optimize the operation strategy of the energy storage system: Based on real-time and accurate data analysis results, provide scientific decision support for the operation of the energy storage system, optimize the charging and discharging strategy, extend the service life of the energy storage equipment, and reduce operation and maintenance costs.

[0007] Improve the adaptability and flexibility of the system: Enable the energy storage system to dynamically adjust its operating strategy according to changes in the energy market and grid demand, and improve the system's ability to adapt to external changes.

[0008] The present invention adopts the following technical solution.

[0009] The first aspect of the present invention provides an energy storage data processing algorithm based on digital twins, comprising:

[0010] Step 1: Use sensors and data acquisition systems to collect key operating data of the energy storage system;

[0011] Step 2: Build a digital twin model based on the key operation data and physical space in step 1;

[0012] Step 3, based on the model constructed in step 2, perform data analysis and processing of the equivalent circuit model to generate data analysis results;

[0013] Step 4: Based on the results generated in step 3, generate optimization suggestions for operation and maintenance, and provide a user interface to display the analysis results and optimization suggestions;

[0014] Step 5: Based on the suggestions generated in step 4, design a feedback mechanism to feed back the system operation results and user operation decisions into the digital twin model for continuous optimization and fine-tuning of the model; continuously improve the performance and adaptability of the algorithm through continuous learning and adaptation to environmental changes.

[0015] Preferably, in step 1, the key operating data of the energy storage system includes: voltage, current, temperature and charge and discharge status information of the battery.

[0016] Preferably, in step 2, model construction includes the fusion of physical model and data-driven model, and machine learning technology is used to optimize and calibrate model parameters.

[0017] Preferably, in step 3, a forgetting factor recursive least squares method is used to identify and update the parameters of the 2-RC equivalent circuit model;

[0018] The least squares form of the 2-RC equivalent circuit model is:

[0019]

[0020] Where u is the forgetting factor, ranging from 0.9 to 0.999; K Ls (k) is the algorithm gain; P Ls (k) is the error covariance matrix; φ(k) is the data matrix at time k; θ(k-1) is the parameter matrix at time k-1; y(k) is the measurement value output at time k.

[0021] Preferably, in step 3, the updated 2-RC equivalent circuit model parameters are provided to the HIF algorithm to find a suitable strategy for the selected performance limit L so that the cost function J satisfies the following equation:

[0022]

[0023] Where k is the sampling time interval, and its value range is 0≤k≤N-1, and x0 and They represent the initial value and initial setting value of the voltage on the capacitor in the 2-RC equivalent circuit respectively; P0 represents the initial error covariance matrix; S k represents weight; Q k and R k They represent the noise covariance matrix of the state equation and the noise covariance matrix of the measurement equation respectively; L is the performance limit, and L≠0; x k is the actual value of the voltage on the capacitor in the 2-RC equivalent circuit at time k; is the estimated value of the voltage on the capacitor in the 2-RC equivalent circuit at time k; N is the maximum number of time steps; w k is the process noise at time k, and its value range is v k is the measurement noise at time k, and its value range is

[0024] Preferably, in step 3, by solving the cost function, a recursive equation satisfying the condition is obtained:

[0025]

[0026] Among them, K k is the gain matrix; P k is the relationship obtained recursively using the symmetric matrix P0; A k is the state transfer matrix; C k is the observation matrix; Q k is the state equation noise covariance matrix; is the estimated value of the state quantity at time k; I is the unit matrix; L is the performance limit, and L≠0; S k is the weight; R k -1 is the inverse matrix of the observation noise covariance matrix; z k is the observed value.

[0027] Preferably, in step 3, the HIF algorithm is combined with a particle filter to optimize information processing and decision support in the filtering process;

[0028] The particle filter algorithm estimates the state of charge by:

[0029] (1) Initialization: Generate a particle set based on the prior probability P(x0) The particle weight is 1 / N s ;

[0030] (2) Update particles:

[0031]

[0032] in, is the state of the ith particle at time k; q is the importance density function; z k is the observed value at time k; is a normal distribution;

[0033] (3) Update weights:

[0034]

[0035] Among them, w i k is the weight of the i-th particle at time k; is the observation model; is the importance density function; It is the state transition model;

[0036] (4) Standardized weight:

[0037]

[0038] Among them, ω k i is the normalized weight of the ith particle at time k;

[0039] (5) Resample the proposed effective particles to calculate the number of effective particles N eff ; compare it to the boundary value to determine if resampling is needed:

[0040]

[0041] (6) Calculation of state parameters:

[0042] (7) Loop: k=k+1.

[0043] Preferably, in step 4, generating an optimization suggestion for maintenance operation includes:

[0044] Based on the data in step 3, identify key indicators and potential problems of system performance, and propose improvement measures such as adjusting system parameters, upgrading hardware and / or optimizing operating processes for the identified potential problems.

[0045] Preferably, in step 5, a closed-loop feedback system is established to feed back user operations and system operation results into the digital twin model, and the model performance is continuously optimized through real-time monitoring and data analysis.

[0046] A second aspect of the present invention provides an energy storage data processing system, which adopts the above-mentioned energy storage data processing method based on digital twins, including:

[0047] Data collection module, digital twin model construction module, data analysis and processing module, decision support module and feedback and iterative optimization module;

[0048] Among them, the data collection module is used to collect key operating data from the energy storage system in real time;

[0049] The digital twin model building module is used to build a digital twin model based on the physical system;

[0050] The data analysis and processing module is used to process data to achieve real-time monitoring and prediction of the energy storage system status;

[0051] The decision support module is used to generate optimization suggestions for operation and maintenance based on data analysis results;

[0052] The feedback and iterative optimization module is used for continuous optimization and fine-tuning of the model.

[0053] The beneficial effects of the present invention are as follows:

[0054] In order to deal with these problems in the background technology, the present invention proposes a storage data processing algorithm based on digital twins. As one of the key technologies to promote optimization, improve efficiency and enhance system resilience, digital twin (DT) can simulate, simulate or "twin" the life cycle of a physical entity, where the physical entity can be an object or a process. The application of digital twin technology in smart energy systems is mainly because it can cope with a series of complex problems such as demand forecasting, asset management and renewable energy integration, thereby improving the sustainability and reliability of energy infrastructure. It is worth mentioning that the adoption of digital twins is seen as a catalyst to promote operational performance improvement, extend asset life, and ensure compliance with strict environmental standards. These virtual models are not only used to verify and monitor real-time functions, but also to simulate and optimize processes before physical implementation, which not only reduces the reliance on expensive prototypes, but also accelerates the development process.

[0055] Through the application of these technologies, digital twins not only improve the technical level of energy storage systems, but also provide strong support for the modernization and sustainable development of the entire energy industry.

[0056] Improve real-time performance and accuracy: Through real-time data collection and instant processing, the present invention significantly improves the real-time performance of energy storage system status monitoring, enabling the system to quickly respond to changes in various operating conditions. The application of digital twin models enhances the accuracy of data analysis, reduces errors in prediction and evaluation, and provides more accurate system status information.

[0057] Optimize operation strategy and improve system efficiency: Through accurate data analysis and model prediction, the present invention helps optimize the charging and discharging strategy of the energy storage system, improve energy utilization efficiency, and extend equipment life. The system's optimized operation strategy can also reduce energy waste, reduce operation and maintenance costs, and improve economic benefits.

[0058] Enhance the adaptability and flexibility of the system: This invention enables the energy storage system to dynamically adjust the operating mode according to real-time data and market demand, enhancing the system's ability to adapt to external changes. The improved flexibility of the system helps to better integrate into the complex and changing energy market and power grid environment, and enhance the market competitiveness of the system.

[0059] Support decision-making and reduce human errors: The decision support module provides data-based operational suggestions to help operators make more scientific and reasonable decisions and reduce human operational errors. The intuitive display and suggestions of the user interface can simplify the operation process and improve operational efficiency and safety.

[0060] Continuous optimization and self-learning capabilities: The feedback and iterative optimization module enables the digital twin model to continuously optimize based on new data and feedback, improving the accuracy and generalization of the model. The system's self-learning ability ensures continuous performance improvement in long-term operation and adapts to technological development and market changes.

[0061] In general, the present invention significantly improves the operating performance and management efficiency of the energy storage system through efficient data processing and advanced model application, and has important practical value and broad market prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the classification of digital twins in smart energy systems;

[0063] Figure 2 It is a schematic diagram of the main branches of the energy storage system;

[0064] Figure 3 It is a schematic diagram of digital twin in smart grid;

[0065] Figure 4 It is a structural diagram of energy storage equipment (battery) based on digital twin;

[0066] Figure 5 It is a schematic diagram of the 2-RC equivalent circuit model;

[0067] Figure 6 It is the flow chart of the joint estimation algorithm. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only embodiments of a part of the present invention, rather than all embodiments. Based on the spirit of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the protection scope of the present invention.

[0069] like Figure 1-Figure 6 As shown, an embodiment of the present invention provides a method for processing energy storage data based on digital twins, comprising the following steps:

[0070] Step 1, data collection, collects key operating data from the physical system of the energy storage system in real time, uses sensors and data acquisition systems to monitor and record the key operating data of the energy storage system in real time, and ensures the integrity and real-time nature of the key operating data.

[0071] In a preferred but non-limiting embodiment of the present invention, Figure 2 As shown in the figure, energy storage systems include electrochemical energy storage, mechanical energy storage, thermal energy storage, chemical energy storage and magnetic energy storage; among them, electrochemical energy storage includes batteries, supercapacitors and capacitors; mechanical energy storage includes compressed air energy storage, pumped storage and flywheel energy storage; thermal energy storage includes latent heat storage and sensible heat storage; chemical energy storage includes fuel cells, hydrogen batteries and biofuels. Key operating data include battery voltage, current, temperature, charge and discharge status and other information.

[0072] Step 2: Build a digital twin model based on the key operating data and physical system in step 1. The physical system of the dual system includes batteries, circuit boards, motors, and connection modules.

[0073] Among them, the dual system refers to the physical model of the energy storage system in the digital twin model and its corresponding data-driven model.

[0074] In a preferred but non-limiting embodiment of the present invention, the designed battery storage system has 16 series and 1 parallel connections, with a total voltage of 57.6V, which can provide 204.6Wh of energy and a maximum power of 581.6W. The BMS board uses the BQ76PL455EVM from Texas Instruments, which can collect the individual voltage, total voltage and temperature of 16 batteries. The Hall current sensor uses ACS712-ELC-30A, which is used to collect current because the circuit board cannot detect the current. The engine model is JGB37-550. The connection module connects the battery, circuit board, sensor and host computer.

[0075] like Figure 3-4 As shown in the figure, a digital twin model is built based on the physical system, and the operating data of the energy storage system is deeply analyzed using the built digital twin model. The analysis content includes but is not limited to performance evaluation, condition monitoring, fault diagnosis and predictive maintenance. Advanced algorithms such as time series analysis, anomaly detection and optimization algorithms are applied to process the data to achieve real-time monitoring and prediction of the energy storage system status.

[0076] Based on the collected real-time data, a digital twin model of the energy storage system is constructed. The digital twin model of the energy storage system is a virtual representation of the energy storage system that can accurately simulate its physical and chemical behavior. The construction of the digital twin model of the energy storage system includes the fusion of the physical model and the data-driven model of the energy storage system, and the use of machine learning technology to optimize and calibrate the parameters of the digital twin model to improve the prediction accuracy and generalization ability of the model.

[0077] A. Building a digital twin model includes the following steps:

[0078] 1. Data collection: First, a large amount of data needs to be collected from the physical system of the energy storage system. This includes using sensors, monitoring equipment, etc. to collect real-time operation data. In a preferred but non-limiting embodiment of the present invention, the real-time operation data is data such as temperature, pressure, and speed.

[0079] 2. Model design: Based on the collected data, design a computational model that can reflect the behavior of the physical system, including physical models, statistical models, or hybrid models.

[0080] 3System modeling: Use specialized software to create dynamic simulations of systems. These models are able to simulate the behavior of physical systems under different conditions.

[0081] B. Model fusion refers to the integration of data and models from different sources to improve the accuracy and robustness of the model. In the digital twin model, model fusion involves the following steps:

[0082] 1. Multi-source data integration: Integrate real-time data from physical systems with other relevant data such as historical data and environmental data.

[0083] 2. Multi-model integration: Combine different models (such as physical models and data-driven models of energy storage systems) to leverage their respective strengths and improve prediction accuracy and system reliability.

[0084] C. Machine learning techniques can be used to optimize and calibrate the parameters of the digital twin model to ensure that the output of the model is as consistent as possible with the performance of the actual system.

[0085] In the embodiments of the present invention, the machine learning techniques used include but are not limited to:

[0086] 1Supervised learning: such as regression analysis, support vector machine (SVM), etc., is used to train models based on input and output data.

[0087] 2 Reinforcement learning: used in the model’s decision-making process to learn the optimal strategy through interaction with the environment.

[0088] 3 Deep learning: Using neural networks, especially when the data volume is large and the complexity is high, it can effectively perform feature extraction and pattern recognition.

[0089] In a preferred but non-limiting embodiment of the present invention, Figure 1 As shown, digital twin applications in smart energy include smart grid, power transmission, building energy consumption, energy storage and others; smart grid applications include fault analysis and anomaly monitoring, renewable energy, microgrid security and 5G and wireless sensor networks; power transmission includes electric vehicles, aviation and multi-purpose transportation models; building energy consumption applications include residential and commercial; energy storage applications include batteries, distributed storage and supercapacitors, thermal energy and gas; others include cyber-physical power systems, general power systems and online power consumption analysis.

[0090] Enabling technologies for digital twins in smart energy include modeling and simulation, IoT-based communications, and lifecycle tools.

[0091] Step 3: Based on the model constructed in step 2, perform data analysis and processing of the equivalent circuit model:

[0092] In a preferred but non-limiting embodiment of the present invention, after comparing multiple battery models, a 2-RC equivalent circuit model is selected, and the calculation accuracy, operation speed and model complexity of the model all meet the design requirements.

[0093] like Figure 5 As shown, in the 2-RC equivalent circuit model, U oc represents the open circuit voltage, R Ω Represents the ohmic internal resistance of the battery. The R1 and C1 circuits are used to simulate the process of rapid rise of the discharge voltage; the R2 and C2 circuits are used to simulate the process of slow stabilization of the discharge voltage; R1 and R2 are the polarization internal resistance of the battery, and C1 and C2 are the polarization capacitance of the battery. In a preferred but non-limiting embodiment of the present invention, in this model, two RC circuits can better simulate the steady-state and transient characteristics of the battery.

[0094] The 2-RC equivalent circuit model is often used in battery management systems to simulate the charge and discharge behavior and internal impedance characteristics of the battery. The model includes two resistors and two capacitors, which represent different electrochemical processes of the battery. By analyzing the measured data in the data matrix, the resistance and capacitance values ​​in the parameter matrix can be estimated. These parameters directly affect the accuracy and reliability of the 2-RC model, and thus affect the model's ability to predict battery behavior.

[0095] In order to solve the data saturation problem of the traditional recursive least squares method, the forgetting factor recursive least squares algorithm is obtained by adding a forgetting factor to the algorithm. The forgetting factor can adjust the new and old data so that the identification result can quickly converge to a value close to the true value when the input changes. In the online identification of the forgetting factor recursive least squares method, the terminal current I is regarded as the system input and the measured voltage U is regarded as the system output. For this single-input single-output system, the differential equations and correlation coefficients containing some parameters are obtained, and then the battery model parameters are solved. When the forgetting factor recursive least squares method is used for parameter identification, the 2-RC equivalent circuit model needs to be converted into a discrete least squares form, which refers to the recursive least squares algorithm containing the forgetting factor.

[0096] The data matrix and parameter matrix of the 2-RC equivalent circuit model are:

[0097]

[0098] Among them, φ(k) is the data matrix of the 2-RC equivalent circuit model; UL(k-1) is the terminal voltage at time k-1; UL(k-2) is the terminal voltage at time k-2; I(k) is the current at time k; θ(k) is the parameter matrix of the 2-RC equivalent circuit model; α1, α2, α3, α4, and α5 are other parameters of the model, which are related to the characteristics of the capacitance and resistance of the circuit; k is the sampling time interval; U oc (k-1) is the open circuit voltage at time (k-1).

[0099] Forgetting factor recursive least squares algorithm flow, the discrete least squares form of the 2-RC equivalent circuit model is:

[0100]

[0101] Where u is the forgetting factor, which is a weighting factor in the error measurement function. The purpose of introducing it is to assign different weights to the original data and the new data so that the algorithm can quickly respond to changes in the input process characteristics. It is usually assigned to 0.9 to 0.999. The smaller the assignment, the stronger the adjustment ability, but there is a possibility of jumping; the larger the task, the slower the operation speed; when specified as 1, the forgetting factor recursive least squares algorithm degenerates into a recursive least squares method; KLs (k) is the algorithm gain; P Ls (k) is the error covariance matrix; φ(k) is the data matrix of the 2-RC equivalent circuit model at time k; θ(k-1) is the parameter matrix of the 2-RC equivalent circuit model at time k-1; y(k) is the measurement value output by the 2-RC equivalent circuit model at time k.

[0102] The parameter matrix and data matrix in equation (1) and equation (2) have no unified symbols. They are unified with reference to the symbols in equation (2). The purpose of constructing the matrix in equation (1) is to convert the equivalent circuit model into the forgetting factor recursive least squares algorithm expression.

[0103] like Figure 6 As shown, based on the recursive least squares expression of the forgetting factor in formula (2), the result of the model parameter update is provided to the state estimation process of the battery charging state joint estimation. Next, the battery charging state joint estimation algorithm is introduced. The battery charging state joint estimation algorithm adopts the HIF (H-infinity filter) algorithm. The game theory method is used in the HIF algorithm.

[0104] Methods for estimating battery SOC using the HIF algorithm include:

[0105] 1. Initialization

[0106] Parameter initialization:

[0107] oSet the initial resistor and capacitor values: R0, R1, R2, C1, C2.

[0108] oInitialize the initial state of charge SoC of the battery.

[0109] 2. Data Collection

[0110] Collect battery voltage, current and temperature data.

[0111] Record timestamps for dynamic analysis.

[0112] 3. HIF estimation

[0113] Prediction Steps:

[0114] oUse the battery model to predict the voltage and SoC at the next moment.

[0115] oCompute the forecast error covariance matrix.

[0116] Update steps:

[0117] oCalculate the error between the actual measured value and the predicted value.

[0118] oUse high-order mutual information to adjust the prediction value and update the SoC.

[0119] 4. HIF-PF Joint Estimation

[0120] Particle filter initialization:

[0121] oGenerate multiple particles, each particle represents a possible SoC state.

[0122] oAssign initial weights to each particle.

[0123] Particle filter iteration:

[0124] o Make predictions and updates for each particle.

[0125] oCalculate the weight of each particle and adjust it based on the measured value.

[0126] Resampling:

[0127] oResample particles according to their weights, keeping particles with high weights.

[0128] 5. Correction of initial value deviation

[0129] Compare the predicted SoC with the actual measured SoC.

[0130] When the absolute difference between the two is less than 0.0001, the calibration is considered complete.

[0131] 6. Results Analysis

[0132] Analyze the estimated SoC over time.

[0133] Evaluate the accuracy and stability of the model.

[0134] 7. Optimization and adjustment

[0135] Adjust model parameters based on the results.

[0136] Conduct multiple experiments to verify the robustness of the model.

[0137] The embodiment provided by the present invention dynamically adjusts the parameters of the filter according to the change of input data to adapt to different environments and data characteristics. By monitoring the data change in real time, the gain and noise covariance matrix of the filter are automatically optimized, thereby realizing adaptive filter parameter adjustment.

[0138] If all sampling time intervals k: Then, the goal of the HIF algorithm is to find a suitable strategy for the selected performance bound L (L≠0) so that the cost function J satisfies the following equation:

[0139]

[0140] Where k is the sampling time interval, and its value range is 0≤k≤N-1, and x0 and They represent the initial value and initial setting value of the voltage on the capacitor in the 2-RC equivalent circuit respectively; P0 represents the initial error covariance matrix; S k represents weight; Q k and R k They represent the noise covariance matrix of the state equation and the noise covariance matrix of the measurement equation respectively; L is the performance limit, and L≠0; x k is the actual value of the voltage on the capacitor in the 2-RC equivalent circuit at time k; is the estimated value of the voltage on the capacitor in the 2-RC equivalent circuit at time k; N is the maximum number of time steps; w k is the process noise at time k, and its value range is v k is the measurement noise at time k, and its value range is

[0141] The diagonal elements of the initial error covariance matrix represent the variance of the initial estimates of the system state variables, that is, the uncertainty of these state estimates. The non-diagonal elements of the initial error covariance matrix represent the covariance of the initial estimation errors between different state variables, that is, the correlation of the initial estimation uncertainties between these variables.

[0142] Weight S k The confirmation method is:

[0143] 1. Problem definition and model building: First, it is necessary to define the dynamic model of the system, including the state equation and measurement equation. These models should accurately reflect the physical behavior of the system and its changes under different operating conditions.

[0144] 2. Preliminary selection of weights: Based on the system's performance requirements and design specifications, the weight values ​​are preliminarily selected. This usually requires the designer to have a deep understanding of the system's sensitivity and noise characteristics.

[0145] 3. System Analysis: Using the selected weights, perform stability and performance analysis of the system. This can be done through simulation and sensitivity analysis to evaluate the impact of different weight configurations on system performance.

[0146] 4. Optimize and adjust: Based on the analysis results, adjust the weight values ​​to optimize system performance. This may require multiple iterations, with each iteration refining the weight values ​​to achieve a better balance between performance and robustness.

[0147] 5. Verification and implementation: Finally, verify the performance of the designed filter through real data or higher fidelity simulations to ensure that the performance requirements can be met under various operating conditions.

[0148] The state equation noise covariance matrix Q k The covariance of the process noise introduced in the state equation is described. Process noise refers to random disturbances that cannot be predicted when the model predicts the system state, which may be caused by a variety of factors, such as model errors, external disturbances, etc. in a preferred but non-limiting embodiment of the present invention.

[0149] Measurement equation noise covariance matrix R k The covariance of the measurement noise is described, that is, the variance of the noise of each measurement variable and the covariance between them. In a preferred but non-limiting embodiment of the present invention, these noises may come from sensor errors, environmental interference or data transmission errors.

[0150] By solving the above cost function problem, we can finally get the recursive equation that satisfies the conditions:

[0151]

[0152] Among them, K k is the gain matrix; P k is the relationship obtained recursively using the symmetric matrix P0; A k is the state transfer matrix; C k is the observation matrix; Q k is the state equation noise covariance matrix; is the estimated value of the state quantity at time k; I is the unit matrix; L is the performance limit, and L≠0; S k is the weight; is the inverse matrix of the observation noise covariance matrix; z k is the observed value.

[0153] Matrix P k It is recursively calculated at each time step through the Kalman filter update formula. This process involves using the state transfer matrix, observation matrix, process noise covariance matrix, and measurement noise covariance matrix to update P k , thus reflecting the evolution of the uncertainty of the system state over time.

[0154] The HIF algorithm can be combined with the particle filter to optimize information processing and decision support in the filtering process:

[0155] 1. Data fusion: The HIF algorithm can be used to more effectively fuse data from different sensors in the prediction and update steps of the particle filter, improving the accuracy and robustness of state estimation.

[0156] 2. Algorithm optimization: The HIF algorithm can optimize the particle generation and weight update strategy, and improve the performance of the particle filter in dealing with complex dynamic environments.

[0157] 3. Performance improvement: Combined with the HIF algorithm, the particle filter can improve computational efficiency and filtering accuracy while maintaining nonlinear processing capabilities, especially in resource-constrained application scenarios.

[0158] The particle filter has no restrictions on process noise and measurement noise, but there is a particle degradation problem, that is, the loss of particle diversity as the number of iterations increases. The best way to solve this problem is to choose a good importance probability density function and use a resampling method. The steps of the particle filter (PF) algorithm to estimate the state of charge are as follows:

[0159] 1) Initialization: Generate a particle set based on the prior probability P(x0) The particle weight is 1 / N s .

[0160] In a preferred but non-limiting embodiment of the present invention, in order to enhance the diversity of particles, a random sampling method is used to generate initial particles to ensure that a wide range of state spaces are covered, and a perturbation mechanism is introduced to increase the differences between particles.

[0161] 2) Update particles:

[0162]

[0163] in, is the state of the ith particle at time k; q is the importance density function; z k is the observed value at time k; is a normal distribution.

[0164] In a preferred but non-limiting embodiment of the present invention, in order to maintain the diversity of particles, the particle distribution is regularly evaluated to prevent particles from being concentrated in a local area, and entropy or other diversity indicators are used for monitoring, and new particles are introduced when necessary.

[0165] 3) Update weights:

[0166]

[0167] in, is the weight of the i-th particle at time k; is the observation model; is the importance density function; It is a state transition model.

[0168] 4) Standardized weight:

[0169]

[0170] in, is the normalized weight of the ith particle at time k.

[0171] 5) Resample the proposed valid particles to calculate the number of valid particles N eff . It is then compared to the boundary values ​​to determine if resampling is necessary:

[0172]

[0173] Among them, N eff is the effective particle number; is the normalized weight of the ith particle at time k.

[0174] In a preferred but non-limiting embodiment of the present invention, the resampling strategy is optimized and the resampling conditions are set: 1. The resampling trigger condition is set, such as the number of valid particles is lower than the set threshold; 2. The reference of system state resampling is used.

[0175] The resampling methods are: 1. Use systematic resampling or residual resampling methods to reduce particle degradation; 2. Introduce randomness in the resampling process to maintain particle diversity.

[0176] 6) Calculate state parameters:

[0177] In a preferred but non-limiting embodiment of the present invention, the computational burden is optimized by: 1. Optimizing resampling; 2. Improving resampling efficiency by using parallel computing technology.

[0178] 7) Loop: k = k + 1

[0179] In the embodiment provided by the present invention, the weight adjustment mechanism of HIF and PF is as follows:

[0180] 1. Data collection and preprocessing

[0181] Collect multi-source data: obtain battery voltage, current, temperature and other data.

[0182] Data cleaning: remove noise and outliers to ensure data quality.

[0183] 2. Application of data fusion technology

[0184] Feature extraction: Extract key features from multi-source data, such as voltage change rate, temperature gradient, etc.

[0185] Fusion algorithm selection: Select an appropriate data fusion algorithm (such as Kalman filtering, Bayesian fusion) to integrate multi-source information.

[0186] 3. Weight adjustment mechanism design

[0187] Initial weight setting: Set the initial weight based on historical data and expert experience.

[0188] Dynamic adjustment strategy:

[0189] Real-time monitoring: monitor battery status and environmental changes.

[0190] Feedback mechanism: weights are adjusted based on prediction errors and actual measurements.

[0191] Adaptive adjustment: Use machine learning algorithms (such as reinforcement learning) to achieve adaptive adjustment of weights.

[0192] 4. Experiment and Verification

[0193] Experimental design: Conduct experiments under different environmental conditions (e.g., temperature fluctuations, battery aging).

[0194] Performance evaluation: Compare the algorithm performance before and after weight adjustment, focusing on accuracy and real-time performance.

[0195] 5. Result analysis and optimization

[0196] Result analysis: Analyze the experimental data and evaluate the effect of weight adjustment.

[0197] Optimization strategy: Further optimize the weight adjustment mechanism based on the analysis results.

[0198] Based on the online recognition method of the forgetting factor recursive least squares algorithm, combined with the fact that HIF can correct the initial value deviation and the particle filter algorithm has a high estimation accuracy, the present invention proposes a joint HIF-PF online estimation algorithm. First, the forgetting factor recursive least squares algorithm is used for online recognition to obtain the R at k moments in real time. Ω , R1, R2, C1 and C2, and then perform joint HIF-PF estimation. Taking advantage of the fact that HIF can correct the initial value deviation, the battery SoC is first estimated using HIF, and the estimated state value at each moment is compared with the state value calculated by the algorithm. When the absolute value of the two values ​​is less than 0.0001, it is considered that the estimated value is closer to the actual value and the corresponding charging state deviation is small, and then jump to the PF algorithm for calculation.

[0199] Although there is a HIF-PF algorithm in the prior art, the algorithm has high complexity, insufficient real-time performance, and the accuracy still needs to be improved under certain specific working conditions. The algorithm is not adaptable to environmental changes, such as temperature fluctuations and aging effects. The present invention introduces a new data fusion technology to optimize the weight adjustment mechanism of HIF and PF to improve the adaptability and accuracy of the algorithm in a dynamic environment; and the present invention develops a new battery degradation model, which is combined with the HIF-PF algorithm to more accurately predict and compensate for battery performance degradation. The accuracy and stability of SOC estimation are improved, especially under conditions of battery aging and environmental changes, the algorithm's operating complexity is reduced, and real-time performance is improved.

[0200] The present invention adopts an efficient particle weight calculation method to improve filtering performance. The improved algorithm includes more accurate likelihood function evaluation or the use of advanced mathematical tools to optimize the weight update process. Traditional particle filters are inefficient when processing high-dimensional data. The present invention effectively improves the processing capability in high-dimensional state space by optimizing the algorithm structure. The present invention can more effectively process systems with rapid dynamic changes and solves the application limitations of traditional methods in such systems. By enhancing particle diversity and optimizing resampling strategies, the present invention improves the accuracy of filtering and the robustness of the algorithm.

[0201] Optimizing weight calculation and particle management strategies reduces unnecessary computational burden and improves the algorithm's running speed. The improved example filter can be applied to a wider range of fields, such as autonomous driving, robot navigation and other demanding real-time systems. While improving efficiency and accuracy, it reduces the demand for hardware resources and reduces operating costs.

[0202] In a preferred but non-limiting embodiment of the present invention, the battery virtual entity is 3D modeled using 3dsMax software. The entire system is decomposed into 6 parts for drawing, namely: the fuel tank body, tracks and gears, the battery and its circuit board, the BMS board, the supporting copper column and the road surface.

[0203] 1) Export all models as .fbx format files;

[0204] 2) Then import it into Unity3D for model setting and adjustment;

[0205] 3) Then start building the UI: select Canvas to generate the canvas;

[0206] 4) Then import the flipped image in the previous step to generate a simple data interface;

[0207] 5) Use Xchart components to draw charts, call database data and realize dynamic display;

[0208] 6) Using the xchart module in Unity3D, you can connect to the local database sqlserver in real time.

[0209] Realize current and voltage monitoring of the battery system and real-time online display of charging status.

[0210] The physical system, digital twin model and digital twin data are connected through the connection module. The two data include: collected current, voltage and charge state calculated by the algorithm. The data is stored and accessed through the database. Sensor data is collected in the physical system and transmitted to the local database receiver-transmitter through universal asynchronous transmission (Uart), and these communications are preprocessed into twin data. Then, the twin data interacts with the twin space and makes the display visualization.

[0211] The voltage, current and charging status data are stored in the SQL Server database and managed by SQL Server Management Studio (Contoso). The database can be updated regularly and communicate and interact with Matlab and Unity3D in real time.

[0212] Python on the host is used to send command frames to the BQ76PL455A-Q1. The data is then processed and written to the Sql server database. The BQ76PL455 class is created by Python and can automatically initialize the device.

[0213] The constructed digital twin model is used to conduct in-depth analysis of the operating data of the energy storage system. The analysis content includes but is not limited to performance evaluation, condition monitoring, fault diagnosis and predictive maintenance. Advanced algorithms such as time series analysis, anomaly detection and optimization algorithms are applied to process the data to achieve real-time monitoring and prediction of the energy storage system status.

[0214] Step 4, decision support: Generate optimization suggestions for operation and maintenance based on the data analysis results generated in step 3. These suggestions are designed to improve the efficiency and reliability of the energy storage system and reduce operation and maintenance costs. Provide a user interface to display analysis results and optimization suggestions to support operators in making quick decisions.

[0215] In a preferred but non-limiting embodiment of the present invention, the interpretation of data analysis results includes:

[0216] oIdentify key indicators of system performance and potential issues by analyzing the data from step 3.

[0217] oIdentify specific areas for optimization, such as energy efficiency, reliability or maintenance costs.

[0218] Optimization recommendations for build operations and maintenance include:

[0219] oPropose specific improvement measures for the identified potential problems, such as adjusting system parameters, upgrading hardware and / or optimizing operating procedures.

[0220] oProvide visualization of analysis results to help users understand the context and expected effects of the recommendations.

[0221] User interface design includes:

[0222] oCreate a user-friendly interface to present analysis results and recommendations.

[0223] oSupport users to make quick decisions and provide operation guidelines and implementation steps.

[0224] It is worth mentioning that the present invention has no special provisions for user interface design. As long as the user interface can display analysis results and optimization suggestions, it falls within the protection scope of the present invention.

[0225] Step 5: Based on the suggestions generated in step 4, feedback and iterative optimization are performed, and a feedback mechanism is designed to feed back the system operation results and user operation decisions to the digital twin model for continuous optimization and fine-tuning of the model. The performance and adaptability of the algorithm are continuously improved through continuous learning and adaptation to environmental changes.

[0226] In a preferred but non-limiting embodiment of the present invention, the feedback mechanism, continuous optimization and fine-tuning method include:

[0227] 1. Feedback mechanism design:

[0228] oEstablish a closed-loop feedback system to feed user operations and system operation results back to the digital twin model.

[0229] oContinuously optimize model performance through real-time monitoring and data analysis.

[0230] 2. Optimize process refinement:

[0231] oClarify the technical details and implementation methods of each optimization step.

[0232] oProvide detailed operation manuals to ensure users can smoothly perform optimization measures.

[0233] 3. Continuous Improvement:

[0234] oRegularly evaluate optimization results and continuously adjust strategies based on environmental changes and user feedback.

[0235] oVerify the effectiveness and adaptability of optimization measures through experiments and simulations.

[0236] In the embodiment of the present invention, the digital twin technology can simulate and analyze the system status and behavior in real time by creating a virtual copy of the energy storage system, while the HIF-PF algorithm is used to process and optimize the fault detection and predictive maintenance of the system. This combination not only improves the monitoring and response capabilities of the system, but also enhances the reliability and efficiency of the system through accurate simulation and prediction.

[0237] Digital twin technology can provide more comprehensive system data and real-time feedback of environmental variables, which may not be fully obtained when the HIF-PF algorithm runs alone. In this way, digital twin technology not only enhances the richness and accuracy of data, but also enables the HIF-PF algorithm to perform more effective fault prediction and processing in more complex practical application scenarios.

[0238] The present invention adopts a more advanced mathematical model and algorithm structure, allowing the algorithm to dynamically adjust the filter parameters according to the changes in the input data, thereby improving the filtering accuracy and adaptability; by optimizing the calculation steps of the algorithm, the unnecessary amount of calculation is reduced, and by effective data preprocessing or introducing parallel computing technology, the operation speed of the algorithm is significantly improved. The traditional H-infinity filter mainly focuses on performance optimization in the worst case, and may not be flexible enough when processing data with high dynamic changes. The HIF algorithm of the present invention can better cope with the rapidly changing data environment and provide more accurate filtering results by introducing an adaptive mechanism.

[0239] Traditional H-infinity filters mainly focus on performance optimization in the worst case, and may not be flexible enough when processing data with high dynamic changes. The HIF algorithm of the present invention can better cope with the rapidly changing data environment and provide more accurate filtering results by introducing an adaptive mechanism. The application of the adaptive adjustment mechanism enables the HIF algorithm to adjust its filtering parameters according to the characteristics of real-time data, which is particularly important in practical applications of processing nonlinear or non-Gaussian noise, such as in complex industrial control systems or high-speed communication systems. The optimized calculation steps allow the algorithm to run with lower computing resources, which is suitable for embedded systems or real-time systems, such as automotive electronics or mobile communication devices.

[0240] By adaptively adjusting and optimizing the HIF algorithm, the accuracy of the filter is improved and the response time of the system is reduced, which is critical for application scenarios that require fast and accurate responses, such as autonomous driving or emergency response systems. The improvement in computing efficiency also reduces energy consumption, which is particularly beneficial for applications running on battery-powered devices and helps reduce the overall operation and maintenance costs of the system.

[0241] The embodiment of the present invention further provides an energy storage data processing system based on a digital twin energy storage data processing algorithm, comprising:

[0242] Data collection module, digital twin model construction module, data analysis and processing module, decision support module and feedback and iterative optimization module;

[0243] Among them, the data collection module is used to collect key operating data from the energy storage system in real time;

[0244] The digital twin model building module is used to build a digital twin model based on the physical system;

[0245] The data analysis and processing module is used to process data to achieve real-time monitoring and prediction of the energy storage system status;

[0246] The decision support module is used to generate optimization suggestions for operation and maintenance based on data analysis results;

[0247] The feedback and iterative optimization module is used for continuous optimization and fine-tuning of the model.

[0248] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0249] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0250] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0251] The computer program instructions for performing the operation of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed completely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be customized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0252] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for processing energy storage data based on digital twins, characterized in that: include: Step 1: Use sensors and data acquisition systems to collect key operating data of the energy storage system; Step 2: Build a digital twin model based on the key operation data and physical space in step 1; Step 3, based on the model constructed in step 2, perform data analysis and processing of the equivalent circuit model to generate data analysis results; Step 4: Based on the results generated in step 3, generate optimization suggestions for operation and maintenance, and provide a user interface to display the analysis results and optimization suggestions; Step 5: Based on the suggestions generated in step 4, design a feedback mechanism to feed back the system operation results and user operation decisions into the digital twin model for continuous optimization and fine-tuning of the model; continuously improve the performance and adaptability of the algorithm through continuous learning and adaptation to environmental changes.

2. The energy storage data processing method based on digital twin according to claim 1 is characterized in that: In step 1, the key operating data of the energy storage system include: battery voltage, current, temperature, and charge and discharge status information.

3. The energy storage data processing method based on digital twin according to claim 1 is characterized in that: In step 2, model construction includes the integration of physical models and data-driven models, and the use of machine learning technology to optimize and calibrate model parameters.

4. The energy storage data processing method based on digital twin according to claim 1 is characterized in that: In step 3, the forgetting factor recursive least square method is used to identify and update the parameters of the 2-RC equivalent circuit model; The least squares form of the 2-RC equivalent circuit model is: Where u is the forgetting factor, ranging from 0.9 to 0.999; K Ls (k) is the algorithm gain; P Ls (k) is the error covariance matrix; φ(k) is the data matrix at time k; θ(k-1) is the parameter matrix at time k-1; y(k) is the measurement value output at time k.

5. The energy storage data processing method based on digital twin according to claim 4 is characterized in that: In step 3, the updated 2-RC equivalent circuit model parameters are provided to the HIF algorithm to find a suitable strategy for the selected performance limit L so that the cost function J satisfies the following equation: Where k is the sampling time interval, and its value range is 0≤k≤N-1, and x0 and They represent the initial value and initial setting value of the voltage on the capacitor in the 2-RC equivalent circuit respectively; P0 represents the initial error covariance matrix; S k represents weight; Q k and R k They represent the noise covariance matrix of the state equation and the noise covariance matrix of the measurement equation respectively; L is the performance limit, and L≠0; x k is the actual value of the voltage on the capacitor in the 2-RC equivalent circuit at time k; is the estimated value of the voltage on the capacitor in the 2-RC equivalent circuit at time k; N is the maximum number of time steps; w k is the process noise at time k, and its value range is v k is the measurement noise at time k, and its value range is 6. The energy storage data processing method based on digital twin according to claim 5 is characterized in that: In step 3, by solving the cost function, we get the recursive equation that satisfies the conditions: Among them, K k is the gain matrix; P k is the relationship obtained recursively using the symmetric matrix P0; A k is the state transfer matrix; C k is the observation matrix; Q k is the state equation noise covariance matrix; is the estimated value of the state quantity at time k; I is the unit matrix; L is the performance limit, and L≠0; S k is the weight; is the inverse matrix of the observation noise covariance matrix; z k is the observed value.

7. The energy storage data processing method based on digital twin according to claim 6 is characterized in that: In step 3, the HIF algorithm is combined with the particle filter to optimize information processing and decision support in the filtering process; The particle filter algorithm estimates the state of charge by: (1) Initialization: Generate a particle set based on the prior probability P(x0) The particle weight is 1 / N s ; (2) Update particles: in, is the state of the ith particle at time k; q is the importance density function; z k is the observed value at time k; is a normal distribution; (3) Update weights: in, is the weight of the i-th particle at time k; is the observation model; is the importance density function; It is the state transition model; (4) Standardized weight: in, is the normalized weight of the ith particle at time k; (5) Resample the proposed effective particles to calculate the number of effective particles N eff ; compare it to the boundary value to determine if resampling is needed: (6) Calculation of state parameters: (7) Loop: k=k+1.

8. The energy storage data processing method based on digital twin according to claim 1, characterized in that: In step 4, optimization recommendations for operations and maintenance are generated, including: Based on the data in step 3, identify key indicators and potential problems of system performance, and propose improvement measures such as adjusting system parameters, upgrading hardware and / or optimizing operating processes for the identified potential problems.

9. The energy storage data processing method based on digital twin according to claim 1, characterized in that: In step 5, a closed-loop feedback system is established to feed back user operations and system operation results into the digital twin model, and the model performance is continuously optimized through real-time monitoring and data analysis.

10. An energy storage data processing system, using the energy storage data processing method based on digital twins according to any one of claims 1 to 9, characterized in that: include: Data collection module, digital twin model construction module, data analysis and processing module, decision support module and feedback and iterative optimization module; Among them, the data collection module is used to collect key operating data from the energy storage system in real time; The digital twin model building module is used to build a digital twin model based on the physical system; The data analysis and processing module is used to process data to achieve real-time monitoring and prediction of the energy storage system status; The decision support module is used to generate optimization suggestions for operation and maintenance based on data analysis results; The feedback and iterative optimization module is used for continuous optimization and fine-tuning of the model.

Images