High-power charging pile charging current automatic adjusting method, medium and system

Abstract

The invention provides a method, medium and system for automatically adjusting charging current of a high-power charging pile, and belongs to the technical field of charging piles. Key parameters of a battery are collected in real time, a charging power offset matrix is established to quantify power deviation, and a current adjustment adjustable matrix is constructed to define a safe adjustment range; generating a harmonic interference probability matrix to evaluate the interference influence of the power grid, establishing an upper-layer game model with charging efficiency maximization as a target and a lower-layer game model with battery safety optimization as a target, processing multi-dimensional monitoring data by adopting a time sequence adjustment model, and outputting a current adjustment suggested value; the neuron activation threshold value is adjusted by adopting the corresponding gain coefficient and the damping parameter according to different working conditions through the charging current adaptive function, intelligent adaptive output of the charging current is finally achieved, and the technical problem that the charging pile cannot achieve intelligent adaptive adjustment of the charging current in a complex power grid environment is solved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of charging piles, and in particular, relates to a high-power charging pile charging current automatic adjustment method, medium and system. BACKGROUND

[0002] With the rapid development of the electric vehicle industry, as a key infrastructure, the core technology of high-power charging piles lies in accurate control of charging current according to battery state and power grid conditions. Traditional charging pile systems mainly use fixed parameter PID control algorithms or simple fuzzy control strategies to adjust charging current. These methods can meet basic needs in stable power grid environments and standard charging conditions, and are widely used in commercial charging stations, residential area charging facilities and highway service areas and other scenarios. However, in the current complex power grid environment, due to the coupling effects of multiple factors such as power grid voltage fluctuations, harmonic interference and load mutations, traditional control methods are difficult to accurately perceive changes in battery state and fluctuations in power grid quality, and cannot realize dynamic optimization and adjustment of charging parameters, resulting in low charging efficiency and increased safety risks. In actual applications, fixed control strategies often ignore the nonlinear changes of battery charging characteristics and the time-varying characteristics of power grid environments, lack deep modeling capabilities for multi-parameter coupling relationships, and cannot maximize charging efficiency while ensuring charging safety. That is, there is a technical problem in the prior art that charging piles cannot realize intelligent adaptive adjustment of charging current in a complex power grid environment. SUMMARY

[0003] Therefore, the application provides a high-power charging pile charging current automatic adjustment method, medium and system, which can solve the technical problem that charging piles cannot realize intelligent adaptive adjustment of charging current in a complex power grid environment in the prior art.

[0004] The application is implemented in the following manner: a first aspect of the application provides a large-power charging pile charging current automatic adjustment method, which comprises constructing a battery state monitoring matrix, establishing a basic data structure for battery state evaluation by real-time acquisition of battery voltage, battery current, battery temperature, battery internal resistance and other parameters; a charging power offset matrix is established, the deviation between the charging pile output power and the target power is calculated according to the power grid voltage fluctuation and the load change, and the data basis for power offset compensation is formed; a current adjustment adjustable matrix is constructed, a dynamic adjustment range matrix of the charging current is established based on the battery charging characteristic curve and the safety threshold constraint; a harmonic interference probability matrix is generated, the harmonic content of the power grid is detected by a frequency domain analysis method, the interference probability of different frequency harmonics on the charging process is calculated, and a harmonic influence evaluation model is established; an upper game model with the maximum charging efficiency as the target and a lower game model with the optimal battery safety as the target are established, and the two models are coordinated and optimized through a charging current coupling function; a pre-trained liquid time sequence adjustment model is used to process the battery state monitoring matrix and the charging power offset matrix, and an optimized charging current adjustment suggestion value is output; and the final charging current output value is calculated through a charging current adaptive function.

[0005] The number of rows of the battery state monitoring matrix corresponds to the type of monitoring parameters, the number of columns corresponds to the time sequence, and the time sequence data of the key state parameters of the battery are stored and organized, and each matrix element represents the measurement value of the battery voltage, the battery current, the battery temperature and the battery internal resistance at the moment, and multi-parameter fusion analysis is realized through matrix operation.

[0006] The charging power offset matrix is used to quantify the difference between the actual output power of the charging pile and the ideal output power, and to provide data support for power adjustment. Each row of the charging power offset matrix corresponds to a different power level, and each column corresponds to a different compensation strategy.

[0007] The current adjustment adjustable matrix is used to define the safe adjustment range of the charging current. The current adjustment adjustable matrix element represents the charging current adjustment boundary under different working conditions. The current adjustment adjustable matrix element is determined according to the battery type, the charging stage and the environmental conditions, so as to ensure that the charging process is carried out within the safety boundary.

[0008] The harmonic interference probability matrix is used to evaluate the potential influence of power grid harmonics on charging equipment. By statistically analyzing historical harmonic data and real-time monitoring results, the occurrence probability and influence degree of harmonic interference are predicted.

[0009] The objective function of the upper game model is the maximum charging efficiency function, the input includes the charging pile output power, the charging current adjustment suggestion value, the charging time and the battery voltage, and the output is the charging efficiency optimization value. The constraint conditions include that the charging power does not exceed the rated power of the equipment, the charging current is within the safety range, and the battery temperature is lower than the safety threshold.

[0010] wherein the objective function of the lower-level game model is a battery safety optimization function, the inputs include battery temperature, charging current adjustment suggestion value, battery internal resistance and voltage difference value, the output is a battery safety evaluation value, and the constraint conditions include that the battery voltage does not exceed the cut-off voltage, the charging current change rate is limited, and the battery internal resistance change is within a normal range.

[0011] wherein the charging current coupling function is used to coordinate the optimization process of the upper-level game model and the lower-level game model, the inputs include the charging efficiency optimization value and the battery safety evaluation value, and the output is a coordinated charging current setting reference value.

[0012] wherein the structure of the liquid time series regulation model is a recurrent neural network architecture based on a liquid state machine, including an input layer, a liquid reservoir layer, a readout layer and a feedback connection layer, the liquid reservoir layer is composed of a large number of randomly connected neurons, and the connection weights and time constants between neurons are dynamically adjusted according to the frequency characteristics of the input signal. The liquid reservoir layer uses a leaky integrator neuron model inside, which has memory decay and nonlinear activation characteristics. The readout layer extracts the state information of the liquid reservoir layer by linear combination, and the feedback connection layer feeds back the output information to the liquid reservoir layer to enhance the time series modeling capability. The neuron connection weights of the liquid time series regulation model are dynamically adjusted according to the battery temperature, voltage difference and charging time, the linear weights of the readout layer are trained using ridge regression method, and the optimization goal is to minimize the mean square error between the predicted charging current and the actual optimal charging current.

[0013] wherein the training data set establishment step of the liquid time series regulation model includes collecting charging data of different types of batteries under various working conditions, the data covers normal charging, fast charging, trickle charging and other charging modes, and the original data is preprocessed including denoising, normalization and feature extraction. The training step of the liquid time series regulation model includes initializing the connection weights and time constants of the liquid reservoir neurons, training the dynamic characteristics of the liquid reservoir layer using unsupervised learning method, exciting the liquid reservoir layer to produce rich dynamic response patterns by inputting historical charging data sequence, and adjusting the model hyperparameters including the size of the liquid reservoir layer, the spectral radius and the input scaling factor using cross-validation method.

[0014] The charging current adaptive function is used to calculate the final charging current output value according to the output of the liquid time sequence adjustment model and the current system state, the input includes the charging current adjustment suggestion value, the battery voltage, the power grid quality index and the environmental temperature, and the output is the charging current output value after adaptive adjustment. When the charging current adjustment suggestion value is in different ranges, different gain coefficients and damping parameters are used to adjust the neuron activation threshold of the liquid time sequence adjustment model. When the charging current adjustment suggestion value is in the safe charging range, a smaller gain coefficient and moderate damping parameter are used to adjust the neuron activation threshold to maintain stable output. When the charging current adjustment suggestion value is close to the safety boundary, a larger gain coefficient and stronger damping parameter are used to increase the neuron activation threshold to enhance the conservatism of the system. When the charging current adjustment suggestion value exceeds the safe range, the maximum gain coefficient and the strongest damping parameter are used to significantly increase the neuron activation threshold to ensure system safety.

[0015] The second aspect of the present application provides a computer readable storage medium, the computer readable storage medium stores program instructions, the program instructions are used to execute the above-mentioned large power charging pile charging current automatic adjustment method when running in the computer.

[0016] The third aspect of the present application provides a large power charging pile charging current automatic adjustment system, which contains the above-mentioned computer readable storage medium, the system is any one of computer, server and single chip microcomputer, the computer readable storage medium is arranged in the system, and the system is provided with a microprocessor for executing the program instructions stored in the computer readable storage medium.

[0017] The present application establishes a multi-dimensional data fusion charging state perception system by constructing a battery state monitoring matrix, a charging power offset matrix, a current adjustment adjustable matrix and a harmonic interference probability matrix, and realizes comprehensive real-time monitoring of the battery state and the power grid environment. By establishing an upper and lower game model and a liquid time sequence adjustment model, the present application constructs a dual-objective coordinated optimization mechanism of maximizing charging efficiency and optimizing battery safety, overcomes the local optimization problem caused by the traditional single control target, and realizes the global optimal configuration of the charging parameters. The liquid time sequence adjustment model is based on the recursive neural network architecture of the liquid state machine, has strong nonlinear modeling capability and time sequence memory characteristics, can accurately capture the dynamic change law of the battery charging process and the time-varying characteristics of the power grid environment, and realizes intelligent adaptive adjustment of the charging current by adaptively adjusting the neuron connection weight and the activation threshold. In summary, the present application solves the technical problem that the charging pile cannot realize intelligent adaptive adjustment of the charging current in the complex power grid environment mentioned in the background art. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is a flowchart of the method of the present application.

[0019] Figure 2 A structural schematic diagram of a liquid state timing adjustment model.

[0020] Figure 3 A charging current suggestion value over time graph output by the liquid state timing adjustment model in Example 2.

[0021] Figure 4 A charging current control curve comparison graph of different types of batteries in Example 2.

[0022] Figure 5 A performance index change trend graph of the system running for 6 months in Example 2. DETAILED DESCRIPTION

[0023] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0024] As Figure 1 shown is a flowchart of a large-power charging pile charging current automatic adjustment method provided by the first aspect of the present application, and the method comprises the following steps:

[0025] S01, a battery state monitoring matrix is constructed, and by collecting parameters such as battery voltage, battery current, battery temperature and battery internal resistance in real time, a basic data structure for battery state evaluation is established, wherein the number of rows of the battery state monitoring matrix corresponds to the type of monitoring parameters, and the number of columns corresponds to the time sequence;

[0026] S02, a charging power offset matrix is established, and according to the grid voltage fluctuation and load change, the deviation between the charging pile output power and the target power is calculated to form a data basis for power offset compensation;

[0027] S03, a current adjustment adjustable matrix is constructed, and based on the battery charging characteristic curve and the safety threshold constraint, a dynamic adjustment range matrix of the charging current is established, and the elements of the current adjustment adjustable matrix represent the charging current adjustment boundary under different working conditions;

[0028] S04, a harmonic interference probability matrix is generated, and by using a frequency domain analysis method, the harmonic content of the grid is detected, the interference probability of different frequency harmonics on the charging process is calculated, and a harmonic influence evaluation model is established;

[0029] S05, an upper game model with the maximum charging efficiency as the target and a lower game model with the optimal battery safety as the target are established, and the two models are coordinated and optimized through a charging current coupling function;

[0030] S06, processing the battery state monitoring matrix and the charging power offset matrix by using a pre-trained liquid time sequence adjustment model, and outputting an optimized charging current adjustment suggestion value, wherein the neuron connection weight of the liquid time sequence adjustment model is dynamically adjusted according to the battery temperature, the voltage difference value and the charging time length;

[0031] S07, optionally, further comprising: calculating a final charging current output value by using a charging current adaptive function, and adjusting the neuron activation threshold of the liquid time sequence adjustment model by using different gain coefficients and damping parameters when the charging current adjustment suggestion value is in different ranges.

[0032] The battery state monitoring matrix is used for storing and organizing time sequence data of key state parameters of the battery, each matrix element representing the measurement value of the battery voltage, the battery current, the battery temperature and the battery internal resistance at a time, and multi-parameter fusion analysis is realized by matrix operation. The charging power offset matrix is used for quantifying the difference between the actual output power of the charging pile and the ideal output power, and providing data support for power adjustment, each row of the charging power offset matrix corresponding to a different power level, and each column corresponding to a different compensation strategy. The current adjustment adjustable matrix is used for defining the safe adjustment range of the charging current, and the elements of the current adjustment adjustable matrix are determined according to the battery type, the charging stage and the environmental conditions, so as to ensure that the charging process is carried out within the safe boundary. The harmonic interference probability matrix is used for evaluating the potential influence of grid harmonics on the charging equipment, and by statistically analyzing historical harmonic data and real-time monitoring results, the occurrence probability and influence degree of harmonic interference are predicted.

[0033] The objective function of the upper game model is a charging efficiency maximization function, the inputs include the charging pile output power, the charging current adjustment suggestion value, the charging time length and the battery voltage, and the output is a charging efficiency optimization value; the constraint conditions include that the charging power does not exceed the rated power of the equipment, the charging current is within the safe range, and the battery temperature is lower than the safety threshold; the objective function of the lower game model is a battery safety optimization function, the inputs include the battery temperature, the charging current adjustment suggestion value, the battery internal resistance and the voltage difference value, and the output is a battery safety evaluation value; the constraint conditions include that the battery voltage does not exceed the cutoff voltage, the charging current change rate is limited, and the battery internal resistance change is within the normal range; the charging current coupling function is used for coordinating the optimization process of the upper game model and the lower game model, the inputs include the charging efficiency optimization value and the battery safety evaluation value, and the output is a coordinated charging current setting reference value.

[0034] The liquid timing adjustment model is a recurrent neural network architecture based on a liquid state machine, including an input layer, a liquid reservoir, a readout layer, and a feedback connection layer. The liquid reservoir is composed of a large number of randomly connected neurons, the connection weights and time constants between the neurons are dynamically adjusted according to the frequency characteristics of the input signal, the internal liquid reservoir adopts a leaky integrator neuron model with memory decay and nonlinear activation characteristics, the readout layer extracts the state information of the liquid reservoir through linear combination, and the feedback connection layer feeds back the output information to the liquid reservoir to enhance the timing modeling capability. The training data set of the liquid timing adjustment model includes collecting charging data of different types of batteries under various working conditions, including the battery voltage, the battery current, the battery temperature, the battery internal resistance, and the corresponding optimal charging current set value. The data covers normal charging, fast charging, trickle charging, and other charging modes, and contains charging data under different environmental temperature and humidity conditions. The original data is preprocessed including denoising, normalization, and feature extraction, the corresponding relationship between the input feature vector and the target output vector is established, and the training set, the validation set, and the test set are divided according to the time sequence. The training steps of the liquid timing adjustment model include initializing the connection weights and time constants of the neurons in the liquid reservoir, training the dynamic characteristics of the liquid reservoir using unsupervised learning method, stimulating the liquid reservoir to generate rich dynamic response patterns by inputting historical charging data sequence, training the linear weights of the readout layer using ridge regression method, optimizing the objective function to minimize the mean square error between the predicted charging current and the actual optimal charging current, adjusting the model hyperparameters including the liquid reservoir size, spectral radius, and input scaling factor using cross-validation method, and fine-tuning the weights of the feedback connection layer using backpropagation algorithm to improve the timing prediction accuracy.

[0035] The charging current adaptive function is used to calculate the final charging current output value based on the output of the liquid timing adjustment model and the current system state. The input includes the charging current adjustment suggestion value, the battery voltage, the grid quality indicator, and the environmental temperature. The output is the adaptive charging current output value. When the charging current adjustment suggestion value is within the safe charging range, a smaller gain coefficient and moderate damping parameter are used to adjust the neuron activation threshold to maintain stable output. When the charging current adjustment suggestion value is close to the safety boundary, a larger gain coefficient and stronger damping parameter are used to increase the neuron activation threshold to enhance the conservatism of the system. When the charging current adjustment suggestion value exceeds the safe range, the maximum gain coefficient and the strongest damping parameter are used to significantly raise the neuron activation threshold to ensure system safety.

[0036] The specific implementation of the above steps is described in detail below.

[0037] The specific implementation of step S01 is to establish a battery state monitoring matrix through a multi-sensor data acquisition system, which uses a time series data organization method to sample and store key battery parameters at fixed time intervals. First, a high-precision voltage sensor is used to collect battery voltage data in real time with a sampling interval of 100 milliseconds, with a sampling accuracy of 1 millivolt and a voltage measurement range of 0 to 1000 volts. Then, the battery charging current data is collected by a Hall current sensor with a sampling frequency synchronized with the voltage, with a current measurement accuracy of 1 milliampere and a measurement range of 0 to 500 amperes. At the same time, a multi-point temperature sensor array is deployed to monitor temperature changes at different positions of the battery, with a temperature sampling accuracy of 0.1 degrees Celsius and a monitoring range of -20 degrees Celsius to 85 degrees Celsius. The battery internal resistance measurement uses the alternating current impedance method, which measures the change in battery internal resistance by injecting a small amplitude high frequency alternating current signal, with a measurement frequency of 1 kilohertz and an internal resistance measurement accuracy of 0.1 milliohm. The number of rows of the monitoring matrix is set to 4 corresponding to voltage, current, temperature and internal resistance parameters, and the number of columns is determined according to the monitoring time length, with 10 data points generated per second. The matrix uses a sliding window mechanism to realize dynamic updating of data, with a window length of 300 seconds to ensure that the system can capture short-term fluctuations and long-term trend changes in battery state.

[0038] The specific implementation of step S02 is to establish a charging power offset matrix based on a power deviation analysis algorithm, which uses real-time power monitoring and target power comparison to quantify the power offset. The system first measures the actual output power of the charging pile in real time through the power metering module, with a power measurement accuracy of 0.2% of the total power and a measurement frequency of 20 times per second. The target power is dynamically set according to the battery charging demand and the grid load, taking into account factors such as battery state of charge, charging mode selection and grid power supply capacity. The power offset is calculated by the difference between the actual power and the target power, with the number of rows of the offset matrix corresponding to different power levels, and the power levels divided into low power mode 0 to 30 kilowatts, medium power mode 30 to 120 kilowatts, and high power mode 120 to 350 kilowatts. The number of columns corresponds to different compensation strategies, including voltage compensation, current limitation, frequency adjustment and load balancing. The matrix element value represents the effectiveness weight of different compensation strategies under each power level, with a weight range of 0 to 1, and the weight value determined by historical data statistical analysis and machine learning methods. The system uses the exponential smoothing method to filter the power offset data, with a smoothing coefficient of 0.3 to effectively reduce the influence of measurement noise on offset calculation.

[0039] The specific implementation of step S03 is to construct a current regulation adjustable matrix using the constraint optimization theory, which establishes the feasible region of current regulation based on the battery charging characteristic curve and multiple safety constraint conditions. The system first determines the basic charging characteristics of the battery according to the battery type identification algorithm, supporting automatic identification of mainstream battery types such as lithium iron phosphate, ternary lithium, and lithium titanate. Corresponding charging characteristic mathematical models are established for different battery types, and the model parameters include maximum charging current, cutoff voltage, temperature coefficient, and aging factor, etc. Safety threshold constraints include battery temperature not exceeding 60 degrees Celsius, single cell voltage not exceeding 110% of the rated voltage, and charging current change rate not exceeding 5% of the current value per second. The number of rows of the adjustable matrix corresponds to different intervals of the battery state of charge, which is divided into three stages: 0 to 20%, 20% to 80%, and 80% to 100%. The number of columns corresponds to different combinations of environmental conditions, including temperature range, humidity level, and ventilation conditions, etc. The matrix elements represent the upper and lower limits of current regulation under the corresponding working conditions, and the regulation range is corrected by a safety margin coefficient, which is dynamically adjusted according to the battery aging degree and use history, ranging from 0.8 to 1.0. The system uses fuzzy logic control method to handle the uncertainty of boundary conditions, ensuring smooth transition of current regulation within the safety boundary.

[0040] The specific implementation of step S04 is to generate a harmonic interference probability matrix using frequency domain analysis technology, which is based on the fast Fourier transform algorithm for spectral analysis of grid voltage and current signals. The system is configured with a high sampling rate data acquisition module, with a sampling frequency of 10.24 kHz, which can accurately capture the 50 Hz fundamental wave and its harmonic components within 50 times. The harmonic detection algorithm uses a sliding window fast Fourier transform method, with a window length of 2048 sampling points and a window overlap rate of 50%, ensuring that the time resolution and frequency resolution of the spectral analysis achieve the best balance. The system focuses on monitoring low-order harmonics such as 2nd, 3rd, 5th, 7th, 11th, 13th, and high-order harmonic groups, and the harmonic content is calculated based on the total harmonic distortion index, which should be less than 5% during normal operation. The establishment of the probability matrix uses statistical methods to establish a prediction model of harmonic interference by analyzing the probability distribution characteristics of historical harmonic data. The number of rows of the matrix corresponds to different harmonic frequency components, and the number of columns corresponds to different interference intensity levels, which are divided into slight interference 0 to 2%, moderate interference 2% to 5%, and severe interference 5% and above. The matrix element value represents the probability of each frequency harmonic under different intensity levels, and the probability calculation uses the Bayesian inference method combined with real-time monitoring data and historical statistical data. The system also considers the interaction and superposition effect between harmonics, and establishes a quantitative model of the influence of harmonic interference on the charging process through nonlinear regression analysis.

[0041] The specific implementation of step S05 is to establish a double-layer optimization model using hierarchical game theory, which realizes the multi-objective coordinated optimization of charging efficiency and battery safety through the Stackelberg game framework. The upper game model aims to maximize charging efficiency, and the gradient ascent algorithm is used to solve the optimal solution. The input parameters of the objective function include charging pile output power, charging current adjustment suggestion value, charging time and battery voltage. The calculation of charging efficiency considers factors such as energy conversion loss, heat loss and time cost, and the efficiency target value is set to 92% or more. The constraint condition is processed by the Lagrange multiplier method to ensure that the charging power does not exceed 95% of the rated power of the equipment, the charging current is within the safe range, and the battery temperature is below the safety threshold of 55 degrees Celsius. The lower game model aims to optimize battery safety, and the particle swarm optimization algorithm is used to find the optimal strategy. The input parameters of the objective function include battery temperature, charging current adjustment suggestion value, battery internal resistance and voltage difference. The safety evaluation is based on a multi-element risk assessment model, which considers factors such as thermal runaway risk, overvoltage risk and aging acceleration risk. The constraint conditions include that the battery voltage does not exceed 98% of the cutoff voltage, the charging current change rate does not exceed 10% of the rated current per minute, and the battery internal resistance change is within 5% of the normal range. The charging current coupling function uses a weighted comprehensive evaluation method to coordinate the optimization results of the two-layer game model, and the weight coefficient is dynamically adjusted according to the charging stage and battery state. In the early stage, the safety is given priority to with a weight of 0.7, and in the later stage, the efficiency is given priority to with a weight of 0.6.

[0042] The specific implementation of step S06 is to use a liquid time sequence adjustment model to process multi-dimensional state data and output an optimized charging current adjustment suggestion value. This model is based on the liquid state machine theory and reservoir calculation principles to realize nonlinear mapping and prediction of time sequence data. The model receives a battery state monitoring matrix and a charging power offset matrix as input, and standardizes the multi-dimensional data through the input layer. The standardization method uses zero-mean unit-variance normalization. The liquid reservoir contains 500 leaky integrator neurons, and the connection between neurons uses a small-world network topology with a connection probability of 10%. The connection weights follow a normal distribution with a mean of 0 and a standard deviation of 0.5. The time constant of the neurons is adaptively adjusted according to the frequency characteristics of the input signal, with a range of 10 milliseconds to 100 milliseconds. The adjustment strategy is based on the matching degree of signal frequency and neuron characteristic frequency. The activation function of the reservoir uses the hyperbolic tangent function, and the activation threshold is dynamically adjusted according to the battery temperature, voltage difference and charging time. The readout layer uses linear regression to extract charging current adjustment information from the reservoir state, and the regression weights are trained by the ridge regression algorithm with a regularization parameter of 0.01. The feedback connection layer feeds back the output information to the reservoir input end at a ratio of 0.1, enhancing the model's ability to model time sequence dependencies. The output of the model is a charging current adjustment suggestion value ranging from 0 to 500 amperes, with an output accuracy of 0.1 amperes.

[0043] The specific implementation of step S07 is the intelligent adjustment of the final charging current output value through a charging current adaptive function, which adopts a piecewise linear control strategy to adjust the neuron activation parameters according to the range in which the recommended value is located. The function inputs include the charging current adjustment recommendation value output by the liquid timing adjustment model, real-time battery voltage, power grid quality indicators, and environmental temperature parameters. The system first determines whether the recommended value is within the safe charging range, which is defined as between 20% and 80% of the rated current. When the recommended value is within the safe range, the gain coefficient is set to 1.0, the damping parameter is set to 0.5, and the neuron activation threshold remains at the standard value of 0.5. When the recommended value approaches the safe boundary, i.e., within the range of 80% to 95% of the rated current, the gain coefficient is adjusted to 1.2, the damping parameter is increased to 0.7, and the activation threshold is increased to 0.7 to enhance the conservatism of the system. When the recommended value exceeds the safe range, i.e., more than 95% of the rated current, the gain coefficient is set to a maximum value of 1.5, the damping parameter is set to a maximum value of 0.9, and the activation threshold is significantly increased to 0.9 to ensure system safety. The adaptive function also considers the influence of the power grid quality indicators, and when the power grid voltage fluctuation exceeds the rated value by 5% or the frequency deviation exceeds 0.5 Hz, all parameters are adjusted by 10% in the conservative direction. The environmental temperature compensation mechanism adjusts the current output according to temperature changes, and for every 10 degrees Celsius increase in temperature, the maximum output current decreases by 5%. The final output charging current value is processed through a smoothing filter to avoid sudden changes, with a filter time constant of 2 seconds.

[0044] The key technical ideas of the present application mainly reflect in four aspects. First, the multi-dimensional state matrix fusion technology, by establishing the battery state monitoring matrix, the charging power offset matrix, the current adjustment adjustable matrix and the harmonic interference probability matrix, realizes the unified modeling and collaborative analysis of the multi-source heterogeneous data of the charging system. Compared with the traditional method which only considers single or a few parameters control strategy, this technology can fully capture various factors and their mutual relationships that affect the charging process, significantly improving the accuracy and robustness of control decisions. Traditional control methods are often based on fixed control rules or simple feedback control, which are difficult to cope with complex and variable charging environments. The multi-dimensional matrix fusion technology realizes multi-parameter coupling analysis through matrix operations, which can identify the nonlinear relationship and time-varying characteristics between parameters.

[0045] Secondly, the double-layer game optimization technology is used to build an upper game model with the goal of maximizing charging efficiency and a lower game model with the goal of optimizing battery safety, achieving dynamic balance optimization of efficiency and safety. Compared with traditional single-objective optimization methods, the double-layer game model can maximize charging efficiency while ensuring battery safety, avoiding the problem of mutual restriction between efficiency and safety in traditional methods. Traditional charging control strategies usually use conservative safety strategies or aggressive efficiency strategies, making it difficult to find the optimal balance point between the two. The double-layer game technology establishes a hierarchical decision-making mechanism through the Stackelberg game framework. The upper model pursues efficiency optimization, and the lower model guarantees the safety bottom line. The two models are optimized through a coupling function to ensure that the optimal control strategy is obtained in different charging stages and working conditions.

[0046] Thirdly, the liquid timing adjustment technology is based on a neural network model constructed based on the liquid state machine theory, which has strong timing modeling and nonlinear mapping capabilities. Compared with traditional feedforward neural networks or simple recurrent neural networks, the liquid timing model achieves efficient processing of timing data and capture of long-term dependencies through reservoir computing mechanisms. Traditional neural networks have problems such as gradient vanishing and training difficulty when processing timing data during the charging process. The random connection structure and dynamic weight adjustment mechanism of the liquid reservoir can naturally process timing information without the need for complex training processes to achieve good prediction performance. This technology is particularly suitable for processing non-stationary timing data during the charging process and can adapt to long-term trends such as battery aging and environmental changes.

[0047] Fourthly, the adaptive threshold adjustment technology is used to dynamically adjust the neuron activation threshold and control parameters according to different ranges of charging current recommendation values, achieving intelligent adjustment of system response characteristics. Compared with traditional fixed threshold control methods, adaptive threshold technology can dynamically adjust the conservatism of control strategies according to changes in system state, maximizing system performance while ensuring safety. Traditional control systems usually use fixed conservative strategies to ensure safety, resulting in limited system performance in safe working conditions. Adaptive threshold technology achieves adaptive optimization of control strategies through piecewise linear control strategies and dynamic parameter adjustment, using aggressive strategies to improve efficiency within the safety range and conservative strategies to ensure safety at the risk boundary.

[0048] The synergy of these four key technical ideas forms a complete intelligent charging control system. The multi-dimensional matrix fusion provides comprehensive and accurate state information for the system, the double-layer game optimization establishes a scientific and reasonable decision-making mechanism, the liquid time sequence adjustment realizes high-precision control and prediction, and the adaptive threshold adjustment ensures the safe and reliable operation of the system. Compared with existing technologies, this collaborative system can achieve dynamic balance between charging efficiency and battery safety in complex and variable charging environments, significantly improving the intelligent level and adaptability of the charging system, and providing a new technical path for the development of high-power charging technology.

[0049] It needs to be noted that the detailed structure of the liquid time sequence adjustment model is based on reservoir calculation theory and liquid state machine principles, as shown in Figure 2 The input layer is responsible for receiving and preprocessing multi-dimensional time sequence data, including battery state monitoring matrix and charging power offset matrix, and uses a sliding time window mechanism to realize time sequence organization. The window length is set to 50 time steps, and the step size is 100 milliseconds. The liquid reservoir is the core calculation unit of the model, consisting of 500 leaky integrator neurons. The neurons are connected randomly with a connection density of 10%, and the connection weights are initialized as Gaussian distribution. The neurons inside the reservoir have different time constants ranging from 10 milliseconds to 100 milliseconds, forming a multi-scale time dynamics characteristic. The readout layer extracts reservoir state information through linear combination, and the weight matrix dimension is the output dimension multiplied by the number of reservoir neurons. The feedback connection layer feeds back the output information to the reservoir input end at a rate of 0.1, forming a closed-loop dynamic system to enhance the time sequence modeling capability.

[0050] The establishment of the training data set includes five detailed steps. In the data collection stage, charging data of different types of batteries under various working conditions are obtained, covering mainstream battery types such as lithium iron phosphate, ternary lithium, and lithium titanate. The working conditions include normal charging, fast charging, trickle charging, etc., and the environmental conditions cover temperatures from -20 degrees Celsius to 60 degrees Celsius and humidity from 20% to 90%. In the data preprocessing stage, the original data is checked and cleaned to remove outliers and missing values, median filtering is used to remove high-frequency noise, and zero-mean unit variance method is used for data standardization. In the feature extraction stage, statistical features, frequency domain features, and time domain features are extracted from the original time sequence data, including mean, variance, peak value, spectral energy distribution, etc. In the data division stage, the data set is divided into training set, validation set and test set according to the continuity of time sequence, the proportion is 6:2:2, and the time independence between different data sets is ensured. In the label generation stage, the optimal charging current set value corresponding to the expert experience and historical optimal charging strategy is generated as the training target.

[0051] The training process adopts a phased training strategy. The reservoir initialization phase sets the spectral radius of the neuron connection weight to 0.95, and the input scaling factor to 0.1, ensuring that the reservoir has edge-stable dynamics. The unsupervised pre-training phase uses historical charging data to stimulate the reservoir to generate rich dynamic response patterns, and evaluates the diversity and expression ability of the reservoir state through principal component analysis. The supervised learning phase uses ridge regression to train the readout layer weights, and the regularization parameter is determined through cross-validation. The objective function is the mean square error between the predicted current and the actual optimal current. The feedback weight optimization phase uses gradient descent method to fine-tune the feedback connection weight, with a learning rate of 0.001 and 1000 training rounds. The model validation phase evaluates the model performance on an independent test set, including prediction accuracy, response time and stability, etc.

[0052] The liquid time series regulation model is particularly suitable for solving the technical problems of the present application, the main reason is that the charging process has strong time dependence and nonlinear characteristics. The state change in the battery charging process shows obvious memory effect, the current state not only depends on the instant input, but also is closely related to the historical state. The traditional feedforward neural network lacks time series memory ability and cannot effectively capture the dynamic characteristics of the charging process. Although the recurrent neural network has time series modeling ability, it has gradient disappearance problem when dealing with long sequence, and the training process is complex. The liquid time series model naturally realizes the time series memory through the random recursive connection of the reservoir, without complex gradient backpropagation training, only needs to train a simple linear readout layer, avoiding the gradient disappearance problem.

[0053] Compared with the closest prior art fuzzy control method and traditional PID control method, the liquid time series regulation model has significant advantages. The fuzzy control method relies on expert experience to develop control rules, which is difficult to handle complex nonlinear relationships and time-varying characteristics, and the control accuracy is limited. The PID control method is based on linear control theory, which cannot adapt to the strong nonlinear characteristics of the charging process, and the parameter adjustment is difficult and has poor adaptability. The liquid time series model learns the complex mapping relationship through a nonlinear dynamic system, which can automatically adapt to different battery types and working condition changes without manual parameter adjustment. The reservoir computing mechanism enables the model to have strong generalization ability and robustness, and can handle working condition combinations that have not been seen in training data. The dynamic weight adjustment mechanism enables the model to automatically adjust the control strategy as the battery ages and the environment changes, achieving long-term stable control performance. The model has excellent real-time performance, low inference computation complexity, and is suitable for online real-time control applications.

[0054] It should be noted that the present application also solves the technical problems of the lack of effective identification and compensation ability of power grid harmonic interference in traditional charging pile systems. In the actual charging process, various frequency harmonics existing in the power grid will interfere with the charging equipment and affect the stability of the charging current and the charging efficiency. The present application constructs a harmonic interference probability matrix, uses a frequency domain analysis method to detect the harmonic content of the power grid in real time, calculates the interference probability of different frequency harmonics on the charging process, establishes a harmonic influence evaluation model, and provides a harmonic compensation basis for the charging control strategy. The liquid time sequence adjustment model can learn and identify the time-varying characteristics of harmonic interference, and realize intelligent compensation and suppression of harmonic interference by dynamically adjusting the neuron connection weight and activation threshold.

[0055] In addition, the present application also solves the technical problem that the existing charging control system cannot realize the coordinated optimization of charging efficiency and battery safety. The traditional control method often uses a single optimization target, either pursuing maximum charging efficiency while ignoring battery safety, or excessively conservatively protecting the battery at the expense of charging efficiency, and it is difficult to find the best balance point between the two. The present application establishes a double-layer game model architecture, the upper model focuses on charging efficiency optimization, and the lower model focuses on battery safety guarantee, and the two models realize information interaction and strategy coordination through a charging current coupling function, realize the maximization of charging efficiency under the premise of guaranteeing battery safety, and avoid the one-sidedness and limitations of traditional methods.

[0056] The second aspect of the present application provides a computer readable storage medium, the computer readable storage medium stores program instructions, the program instructions are used to execute the above-mentioned large power charging pile charging current automatic adjustment method when running in the computer.

[0057] The third aspect of the present application provides a large power charging pile charging current automatic adjustment system, which contains the above-mentioned computer readable storage medium, the system is any one of a computer, a server and a single chip microcomputer, the computer readable storage medium is arranged in the system, and the system is provided with a microprocessor for executing the program instructions stored in the computer readable storage medium.

[0058] Specifically, the principle of this invention is as follows: The key to solving the technical problem of charging piles' inability to achieve intelligent adaptive adjustment of charging current in complex power grid environments lies in constructing an intelligent control architecture based on multi-dimensional data fusion and two-layer game optimization. First, through the collaborative construction of the battery state monitoring matrix, charging power offset matrix, adjustable current adjustment matrix, and harmonic interference probability matrix, this invention establishes a comprehensive perception system covering the internal state of the battery, the external environment of the power grid, safety constraints, and interference impact assessment, providing a complete data foundation for intelligent decision-making. Second, the upper-layer game model aims to maximize charging efficiency, while the lower-layer game model aims to optimize battery safety. The two models achieve coordinated optimization through a charging current coupling function. This two-layer game architecture ensures both the high efficiency of the charging process and the safety of system operation, avoiding the conflict between efficiency and safety that may occur with traditional single-objective optimization. The liquid-state time-series adjustment model serves as the core controller, employing a recurrent neural network architecture based on a liquid state machine. Its liquid reservoir consists of a large number of randomly connected neurons, possessing rich dynamic response modes and powerful nonlinear mapping capabilities, enabling accurate modeling of the complex time-varying characteristics of the battery charging process. In the model, the neuron connection weights are dynamically adjusted based on battery temperature, voltage difference, and charging time. The activation threshold is optimized in real time according to the current system state through an adaptive charging current function, achieving adaptive adjustment of control parameters. This design enables the charging pile to automatically adjust its charging strategy based on real-time monitoring of battery status and grid environment, maintaining optimal charging performance even under complex and changing operating conditions.

[0059] The following provides a specific embodiment 1 of the present invention, and the specific implementation of each step in this embodiment 1 is described in detail below.

[0060] The specific implementation of step S01 involves constructing a battery state monitoring matrix, which is used to store time-series data of key battery parameters, as shown below:

[0061] ;

[0062] In the formula, Battery status monitoring matrix; For the first The battery voltage at any given moment, in volts; For the first The battery current at any given moment, measured in amperes; For the first Battery temperature at any given moment, in degrees Celsius; For the first The internal resistance of the battery at a given moment, in milliohms; For the first Each sampling time, ; is the total number of sampling points. The parameter acquisition method is as follows: Real-time acquisition by high-precision voltage sensor, sampling frequency is 10 Hz, measurement accuracy is 1 mV; Acquired by Hall current sensor, sampling frequency is synchronized with voltage, measurement accuracy is 1 mA; Acquired by multi-point temperature sensor array, using thermocouple temperature measurement method, measurement accuracy is 0.1℃; Measured by AC impedance method, calculated by injecting a small amplitude AC signal with a frequency of 1 kHz.

[0063] The specific implementation of step S02 is to establish a charging power offset matrix. First, calculate the deviation between the actual output power of the charging pile and the target power, which is specifically represented as follows:

[0064] ;

[0065] In the formula, is the power deviation value at time , unit: kW; is the actual output power of the charging pile, unit: kW; is the target output power, unit: kW; is the measurement noise term, unit: kW. The charging power offset matrix is represented as:

[0066] ;

[0067] In the formula, is the charging power offset matrix; is the weight coefficient of the th power level under the th compensation strategy, respectively corresponding to low, medium and high power modes, respectively corresponding to voltage compensation, current limitation, frequency adjustment, load balancing strategy. The parameter acquisition method is as follows: Measured in real time by power metering module, using three-phase power measurement method, measurement accuracy is 0.2% of total power; Calculated and determined according to the state of charge of the battery and the charging demand; The standard deviation range is 0.1% to 0.5% of the measured value; Determined by historical data statistical analysis and least squares fitting.

[0068] The specific implementation of step S03 is to construct a current regulation adjustable matrix. Based on the constraint optimization theory, the feasible region of current regulation is established, and the calculation formula of current regulation range is:

[0069] ;

[0070] wherein, is the state of charge and environmental conditions , in amperes; is the reference charging current, in amperes; is the safety factor, ranging from 0.8 to 1.0; is the aging correction factor, ranging from 0.85 to 1.0; is the temperature compensation term, in amperes; is the ambient temperature, in degrees Celsius. The current regulation adjustable matrix is expressed as:

[0071] ;

[0072] wherein, is the current regulation adjustable matrix; corresponding to the state of charge intervals 0-20%, 20-80%, and 80-100%, respectively; corresponding to different combinations of environmental conditions. The parameter acquisition method is: determined according to the battery charging characteristic curve; calculated based on safety threshold constraints; obtained through battery capacity attenuation testing; linearly interpolated using temperature coefficients.

[0073] The specific implementation of step S04 is to generate a harmonic interference probability matrix. The power grid signal is analyzed in the frequency domain through fast Fourier transform, and the calculation formula of total harmonic distortion is:

[0074] ;

[0075] wherein, is the total harmonic distortion rate, in percent; is the amplitude of the th harmonic, in volts or amperes; is the fundamental amplitude, in volts or amperes; is the highest harmonic number for analysis, usually 50; is the measurement error term, ranging from 0.1% to 0.3%. The harmonic interference probability matrix is expressed as:

[0076] ;

[0077] wherein, is the harmonic interference probability matrix; is the th harmonic in the interference intensity level The probability of occurrence of the following, is a harmonic number index, respectively corresponding to slight, moderate, severe interference levels. Among them, the parameter acquisition method is: Through fast Fourier transform calculation, the sampling frequency is 10.24 kHz, and the window length is 2048 points; Based on statistical analysis of historical harmonic data, Bayesian inference method is used for calculation.

[0078] The specific implementation of step S05 is to establish a double-layer game optimization model. The objective function of the upper game model is:

[0079] ;

[0080] In the formula, is the charging efficiency objective function; is the charging pile output power, unit: kilowatt; is the charging current adjustment suggestion value, unit: ampere; is the charging time, unit: hour; is the battery voltage, unit: volt, derived from the real-time value of ; is the total input energy, unit: kilowatt-hour; is the heat loss, unit: kilowatt-hour; is the time cost, expressed in equivalent energy unit kilowatt-hour; is the penalty factor, the value range is 0.01 to 0.1; is the th constraint violation amount; is the total number of constraints. The objective function of the lower game model is:

[0081] ;

[0082] In the formula, is the battery safety objective function; is the battery temperature, unit: Celsius, derived from the real-time value of ; is the battery internal resistance, unit: milliohm, derived from the real-time value of ; is the voltage difference value, unit: volt; is the thermal safety evaluation value; is the voltage safety evaluation value; is the aging safety evaluation value; is the weight coefficient; is the risk penalty factor, the value range is 0.05 to 0.2; is the th safety risk amount, is the risk index; is the total number of risk factors. The charging current coupling function is:

[0083] ;

[0084] wherein, is the coordinated charging current set reference value, in amperes; is the linear coupling coefficient; is the nonlinear coupling coefficient.

[0085] The specific implementation of step S06 is the same as the aforementioned liquid timing adjustment model training, and will not be described in detail here.

[0086] The specific implementation of step S07 is to calculate the final output value through the charging current adaptive function, which adopts a piecewise linear control strategy and is specifically represented as follows:

[0087] ;

[0088] ;

[0089] wherein, is the final charging current output value, in amperes; is the rated charging current, in amperes; is the gain coefficient of different intervals, respectively taking values of 1.0, 1.2, and 1.5; is the bias term, in amperes; is the environmental compensation function of the interval, is the interval index; is the grid voltage, in volts; is the ambient temperature, in degrees Celsius. The environmental compensation function is represented as:

[0090] ;

[0091] wherein, is the compensation coefficient of the interval; is the rated grid voltage, in volts; is the reference temperature, taking a value of 25 degrees Celsius.

[0092] It needs to be explained that the battery state monitoring matrix The multi-dimensional time-series data is organized in matrix form, with different parameter types represented by row vectors and time series represented by column vectors, realizing the mathematical basis of multi-parameter fusion analysis. Compared with traditional single-parameter monitoring methods, this matrix structure can capture the time-series correlation and coupling relationship between parameters, providing comprehensive state information for subsequent intelligent decision-making, significantly improving the accuracy and robustness of charging control.

[0093] Power deviation calculation formula By introducing a noise term Considering the uncertainty in actual measurement, the power offset evaluation is more accurate and reliable. Charging power offset matrix The mapping relationship between power level and compensation strategy is established, which can adaptively select the optimal compensation strategy according to different power levels compared with traditional fixed compensation methods, effectively improving the accuracy and response speed of power regulation.

[0094] Current regulation range calculation formula Considering factors such as reference current, safety factor, aging correction, and temperature compensation, the formula realizes the synergistic effect of multiple factors through the combination of multiplication and addition. Compared with traditional single-factor regulation methods, this formula can more comprehensively reflect the actual charging capacity and safety boundary of the battery, ensuring dynamic balance between optimal efficiency and safety during the charging process.

[0095] Total harmonic distortion calculation formula The harmonic energy is calculated by squaring and taking the square root, and a measurement error term is introduced Improving the calculation accuracy. Harmonic interference probability matrix The probability mapping of harmonic frequency and interference strength is established, which can quantify the uncertainty of harmonic interference compared with traditional deterministic harmonic analysis methods, providing a probabilistic decision basis for the development of charging control strategies.

[0096] The objective function of the double-layer game model realizes the softening of constraints by introducing a penalty term, the upper efficiency objective function The efficiency index is defined by ratio, and the lower safety objective function Multiple safety factors are integrated in the form of weighted sum. Coupling function The coordination and optimization of two-layer objectives are realized through the combination of linear terms And nonlinear terms Compared with traditional single-objective optimization methods, it can achieve Pareto optimality between efficiency and safety, significantly improving the comprehensive performance of the charging strategy.

[0097] Charging current adaptive function Adopting a piecewise linear structure, different gain coefficients and compensation strategies are used in different current intervals, and the environmental compensation function The influence of grid voltage and ambient temperature is considered in the form of linear combination. Compared with the traditional fixed parameter control method, the adaptive mechanism can dynamically adjust the conservative degree of the control strategy according to the system state, maximize the system performance under the premise of ensuring safety, and realize the intelligentization and adaptability of the control strategy.

[0098] In order to better understand and implement the present application, the following provides an embodiment 2 of a specific application scenario of the present application: A certain city's electric vehicle charging station is equipped with 20 350-kilowatt high-power charging piles, serving more than 500 electric vehicles per day, facing technical challenges such as severe load fluctuation of the city's regional power grid in summer, diverse types of vehicle batteries, and complex environmental conditions. The technical team uses the charging current automatic adjustment method of the present application to build a complete intelligent charging control system.

[0099] The technical team first establishes a battery state monitoring matrix and deploys a high-precision sensor network to monitor the charging process in real time. The system collects four key parameters, including battery voltage, current, temperature, and internal resistance, with a sampling interval of 100 milliseconds. In a typical charging process, the voltage of a lithium iron phosphate battery ranges from 280 volts to 390 volts, the charging current ranges from 50 amperes to 300 amperes, the battery temperature is controlled between 15 degrees Celsius and 45 degrees Celsius, and the internal resistance varies between 0.8 milliohms and 1.5 milliohms. The parameter ranges of a ternary lithium battery are different, with a voltage range of 320 volts to 420 volts, a charging current of up to 350 amperes, and a relatively narrow temperature range of 20 degrees Celsius to 40 degrees Celsius. The monitoring matrix established by the system contains 3000 data points every 300 seconds, providing rich basic data for subsequent intelligent decision-making.

[0100] The establishment of the charging power offset matrix is based on a deep analysis of the grid conditions. The technical team found that the voltage of the regional power grid fluctuates greatly during the morning and evening peak periods, with a peak voltage of 108% of the rated value and a valley voltage of 92%. The power offset calculation results show that the offset is usually between 2% and 5% in low power mode, increases to 5% to 12% in medium power mode, and can reach more than 15% in high power mode. The system formulates corresponding compensation strategy weights for different offset conditions, as shown in Table 1.

[0101] Table 1 Power offset compensation strategy weight table

[0102] The construction of the current regulation adjustable matrix takes into account the special requirements of different battery types and charging stages. According to the battery charging characteristic curve, the technical team divides the charging process into three stages: the initial fast charging stage, the middle constant current stage, and the late trickle charging stage. The safety factor of the initial fast charging stage is set to 0.95, the aging correction factor is 0.90, and the temperature compensation range is negative 10 amperes to positive 15 amperes. The safety factor of the middle constant current stage is increased to 0.98, the aging correction factor is 0.95, and the temperature compensation range is reduced to negative 5 amperes to positive 8 amperes. The late trickle charging stage uses the most conservative parameter settings, with a safety factor of 1.0, an aging correction factor of 1.0, and a temperature compensation range of negative 2 amperes to positive 3 amperes.

[0103] The generation of the harmonic interference probability matrix is based on long-term monitoring of the harmonic characteristics of the regional power grid. The technical team uses a data acquisition system with a sampling frequency of 10.24 kHz to accurately analyze the harmonic components in the power grid. The monitoring results show that the total harmonic distortion rate of the regional power grid is 3.2% under normal conditions and can reach 4.8% during peak load periods. The main harmonic components include the 3rd, 5th, 7th, and 11th harmonics, with the 5th harmonic having the highest content, reaching 2.1% of the fundamental wave. The harmonic interference probability matrix established by the system shows that the probability of slight interference is 65%, the probability of moderate interference is 28%, and the probability of severe interference is 7%.

[0104] The establishment of the double-layer game optimization model is the core technical link of the system. The upper game model aims to maximize charging efficiency, and through analysis, it is found that the energy conversion efficiency of a typical charging process can reach 93.5%, the heat loss accounts for 4.2% of the total energy, and the time cost equivalent energy loss is 2.3%. The system sets the following constraints: the charging power does not exceed 95% of the rated power of the device, the charging current does not exceed 98% of the battery safety threshold, and the battery temperature does not exceed 50 degrees Celsius. The lower game model aims to optimize battery safety, and establishes a comprehensive safety evaluation system. The calculation of the thermal safety evaluation value takes into account the battery temperature distribution, heat dissipation efficiency, and thermal runaway risk, with a typical value range of 0.85 to 0.95. The voltage safety evaluation value is based on the single cell voltage uniformity and overvoltage risk, with a typical value range of 0.90 to 0.98. The aging safety evaluation value considers the capacity decay rate and internal resistance growth trend, with a typical value range of 0.88 to 0.96.

[0105] The training of the liquid timing adjustment model used a large dataset containing 50,000 charging cycles. The dataset covered 15 different types of electric vehicle batteries, charging data in different seasons, different time periods, and different load conditions. The liquid reservoir of the model contains 500 neurons, uses a small-world network topology, and has a connection density of 10%. After training, the model achieved a prediction accuracy of 97.3% on the test set, with a response time of 15 milliseconds, and could output real-time optimized charging current adjustment recommendations. As shown in Figure 3 , the charging current recommendation value output by the model showed a clear stage feature over time, with a higher value at the beginning of charging and gradually decreasing as the state of charge increased, reflecting an accurate understanding of the battery charging characteristics.

[0106] The charging current adaptive function uses different control strategies according to the range of the recommendation value. When the recommendation value is within the safe charging range, the system uses the standard control mode with a gain coefficient of 1.0 and a damping parameter of 0.5. When the recommendation value approaches the safety boundary, the system automatically switches to a conservative control mode, with a gain coefficient of 1.2 and a damping parameter of 0.7. When the recommendation value exceeds the safe range, the system starts the highest safety level control, with a gain coefficient of 1.5 and a damping parameter of 0.9. The environmental compensation mechanism adjusts dynamically according to the grid voltage fluctuations and environmental temperature changes. For every 1% fluctuation in grid voltage, the current output adjusts by 2%, and for every 5°C change in environmental temperature, the maximum output current adjusts by 3%.

[0107] The actual effect of the system after being put into operation verified the effectiveness of the technical scheme. During the 6-month operation monitoring, the charging process of the electric vehicles served by the system showed good stability and safety. The peak battery temperature during charging was reduced by 8% compared to the traditional control method, the charging time was shortened by 12%, and the energy utilization efficiency was improved by 5%. As shown in Figure 4 , the system showed good adaptability in the charging process of different types of batteries, with smooth and stable charging curves for lithium iron phosphate batteries and a good balance between efficiency and safety for ternary lithium batteries. The interference of grid harmonics on the charging process was effectively suppressed, with a charging current fluctuation amplitude of less than 3%.

[0108] The technical team also tested the system's performance in extreme working conditions. In summer high-temperature weather, when the environmental temperature reached 42°C, the system could automatically adjust the charging strategy by reducing the charging current and increasing the cooling interval to ensure that the battery temperature was always within the safe range. In winter low-temperature environment, when the environmental temperature dropped to -15°C, the system ensured the normal progress of the charging process by increasing the initial charging current and prolonging the preheating time. In the case of sudden changes in grid load, the response time of the system was within 50 milliseconds, and it could quickly adjust the charging parameters to adapt to changes in grid conditions.

[0109] The maintenance and upgrade of the system also reflect the intelligent features. The liquid time sequence adjustment model has online learning ability and can continuously optimize the control strategy according to new charging data. The technical team sets an automatic data updating mechanism, and after collecting 1000 new charging cycle data, the system automatically fine-tunes the model parameters. The weight coefficients of the double-layer game model are also adjusted regularly according to long-term running statistical data to ensure that the balance point of efficiency and safety is always optimal. As shown in FIG. 6, the performance indicators of the system after running for 6 months show a continuous improvement trend, and all key parameters show good convergence characteristics. Figure 5

[0110] The data statistical results of the multi-dimensional matrix fusion processing are shown in Table 2, which shows the comprehensive perception ability of the system to the complex charging environment.

[0111] Table 2 Multi-dimensional state parameter statistical table

[0112] The present application brings significant technical progress compared to traditional charging control methods. Traditional charging control methods usually use fixed control parameters and simple feedback control strategies, which cannot adapt to complex and variable charging environments and the individual needs of different types of batteries. The present application realizes real-time monitoring and analysis of the charging process from all angles through multi-dimensional state matrix fusion technology, which can capture more rich state information and coupling relationships between parameters compared to traditional single-parameter control methods. The application of the double-layer game optimization model breaks through the limitations of traditional single-objective optimization, and by establishing a hierarchical decision-making mechanism, it maximizes charging efficiency while ensuring battery safety, avoiding the problem of mutual restriction between efficiency and safety in traditional methods. The liquid time sequence adjustment model based on the time sequence modeling capability of neural networks can better handle the nonlinear dynamic characteristics and long-term dependencies in the charging process compared to traditional PID control or fuzzy control methods, realizing accurate prediction and control of the charging current. The adaptive threshold adjustment technology dynamically adjusts the control strategy according to the system state, which can automatically optimize the control parameters under different working conditions compared to the traditional fixed threshold control method, realizing the intelligentization and adaptability of the control strategy. The synergistic effect of these technological innovations makes the entire charging control system more robust and adaptable in complex environments, providing a new technical path for the development of high-power charging technology.

[0113] ====↓↓↓↓↓=====Variable Explanation====↓↓↓↓↓=====

[0114] It should be noted that the variables involved in the present application are explained in detail as shown in Tables 3 and 4.

[0115] ​Table 3 Variable explanation table (first part)

[0116] Table 4 Variable explanation table (second part)

[0117] The above merely provides a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for automatically adjusting the charging current of a high-power charging pile, characterized in that, The battery state monitoring matrix is constructed, and the basic data structure for battery state evaluation is established by real-time collection of parameters such as battery voltage, battery current, battery temperature and battery internal resistance; the charging power offset matrix is established, and the deviation between the charging pile output power and the target power is calculated according to the power grid voltage fluctuation and load change, forming the data basis for power offset compensation; The current regulation adjustable matrix is constructed, and the dynamic adjustment range matrix of the charging current is established based on the battery charging characteristic curve and the safety threshold constraint; the harmonic interference probability matrix is generated, the harmonic content of the power grid is detected by the frequency domain analysis method, the interference probability of different frequency harmonics on the charging process is calculated, and the harmonic influence evaluation model is established; the upper game model with the maximum charging efficiency as the target and the lower game model with the optimal battery safety as the target are established, and the two models are coordinated and optimized through the charging current coupling function; the pre-trained liquid state time sequence adjustment model is used to process the battery state monitoring matrix and the charging power offset matrix, and the optimized charging current regulation suggestion value is output; the final charging current output value is calculated through the charging current adaptive function.

2. The method of claim 1, wherein, The number of rows of the battery state monitoring matrix corresponds to the type of monitoring parameters, and the number of columns corresponds to the time sequence, which is used to store and organize the time sequence data of the key state parameters of the battery. Each matrix element represents the measurement value of the battery voltage, battery current, battery temperature and battery internal resistance at the moment, and multi-parameter fusion analysis is realized through matrix operation.

3. The method of claim 2, wherein, The charging power offset matrix is used to quantify the difference between the actual output power of the charging pile and the ideal output power, and to provide data support for power regulation. Each row of the charging power offset matrix corresponds to a different power level, and each column corresponds to a different compensation strategy.

4. The method of claim 3, wherein the method further comprises: The current regulation adjustable matrix is used to define the safe adjustment range of the charging current. The current regulation adjustable matrix elements represent the charging current adjustment boundaries under different working conditions. The current regulation adjustable matrix elements are determined according to the battery type, charging stage and environmental conditions to ensure that the charging process is carried out within the safety boundary.

5. The method of claim 4, wherein, The harmonic interference probability matrix is used to evaluate the potential impact of power grid harmonics on charging equipment. By statistically analyzing historical harmonic data and real-time monitoring results, the occurrence probability and impact degree of harmonic interference are predicted.

6. The method of claim 5, wherein, The objective function of the upper game model is the maximum charging efficiency function, and the input includes charging pile output power, charging current regulation suggestion value, charging time and battery voltage. The output is the charging efficiency optimization value. The constraint conditions include that the charging power does not exceed the rated power of the device, the charging current is within the safe range, and the battery temperature is below the safety threshold.

7. The method of claim 6, wherein, The objective function of the lower game model is the optimal battery safety function, and the input includes battery temperature, charging current regulation suggestion value, battery internal resistance and voltage difference. The output is the battery safety evaluation value. The constraint conditions include that the battery voltage does not exceed the cutoff voltage, the charging current change rate is limited, and the battery internal resistance change is within the normal range.

8. The method of claim 7, wherein, The charging current coupling function is used to coordinate the optimization process of the upper game model and the lower game model. The input includes the charging efficiency optimization value and the battery safety evaluation value, and the output is the coordinated charging current setting reference value.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores program instructions, and the program instructions are used for executing the method of automatically adjusting the charging current of the high-power charging pile according to any one of claims 1-8 when running in the computer.

10. A high-power charging pile charging current automatic adjustment system, characterized in that, The system is any one of a computer, a server and a single-chip microcomputer, and the computer readable storage medium is arranged in the system, and the system is provided with a microprocessor for executing the program instructions stored in the computer readable storage medium.

Images