A high-power DC charging pile overcurrent prediction method, medium and system

Abstract

The present invention provides a method, medium and system for predicting overcurrent of a high-power DC charging pile, which belongs to the technical field of charging piles, and includes: establishing a charging response equation group with error terms, including equations such as voltage balance, power transmission, current limitation, temperature change, battery state of charge and charging efficiency. The error terms of the equation group are fitted by using historical charging data with overcurrent phenomenon to obtain an optimized charging response equation group. According to the equation group, the influence of grid voltage, charging power, battery temperature, battery state of charge and the like on the charging current is analyzed. The characteristic parameters in the window, such as average charging current, current change rate, temperature change and the like, are extracted, and a neural network model for overcurrent prediction is established. The characteristic parameters of the charging pile to be tested are obtained in real time, and the input model is used to predict the possible overcurrent time and current value. The technical problem that the prior art is difficult to predict overcurrent of a high-power DC charging pile is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of charging piles, and in particular, relates to a method, medium and system for predicting overcurrent of a high-power DC charging pile. Background Art

[0002] With the rapid development of new energy vehicles, high-power DC charging technology has received increasing attention. Compared with the traditional slow charging mode, high-power DC charging can significantly shorten the charging time, improve charging efficiency, and meet consumers' demand for fast charging. The output power of a typical high-power DC charging pile is usually between 50kW and 350kW, and it can complete about 80% of the battery charge in 10-30 minutes. But at the same time, high-power charging also brings some technical challenges, the most prominent of which is the possible overcurrent problem.

[0003] During high-power DC charging, the current input from the power grid to the charging pile is often very large. Once an abnormal situation occurs, such as a sudden increase in the internal resistance of the battery or an excessively high charging environment temperature, the charging current may increase instantly, exceeding the safety limit of the charging pile and the on-board battery, thereby causing serious consequences such as equipment damage, overheating, and even fire. To address this problem, existing technologies usually adopt the following overcurrent protection measures:

[0004] 1) An overcurrent detection device is installed at the charging pile end. Once it detects that the charging current exceeds the threshold, the charging circuit is immediately cut off.

[0005] 2) Set up an overcurrent protection function in the vehicle's battery management system (BMS). When the BMS detects that the battery enters an overcurrent state, it will actively reduce or interrupt charging.

[0006] However, these prior art overcurrent protection methods need to detect overcurrent before they can be activated, and it is difficult to activate the protection method according to the predicted overcurrent.

[0007] In other words, it is difficult to predict the overcurrent of a high-power DC charging pile using existing technologies. Summary of the invention

[0008] In view of this, the present invention provides a high-power DC charging pile overcurrent prediction method, medium and system, which can solve the technical problem that the prior art is difficult to predict the overcurrent of a high-power DC charging pile.

[0009] The present invention is achieved in that:

[0010] A first aspect of the present invention provides a high-power DC charging pile overcurrent prediction method, which includes the following steps:

[0011] S10. Establish a charging response equation group with error terms that considers the grid power input parameters, charging parameters, and battery parameters of the new energy vehicle of the high-power DC charging pile, including a voltage balance equation, a power transmission equation, a current limit equation, a temperature change equation, a battery state of charge equation, and a charging efficiency equation;

[0012] S20, acquiring multiple groups of historical charging process data with overcurrent phenomena, including grid power input data, charging data, and battery data, and fitting the error term of each equation in the charging response equation group according to the charging process data to obtain a fitted charging response equation group;

[0013] S30, taking the charging current change as the dependent variable, solving the charging response equation group to obtain the charging current influence equation group, including the influence equation of the grid voltage on the charging current, the influence equation of the charging power on the charging current, the influence equation of the battery temperature on the charging current, the influence equation of the battery state of charge on the charging current, the influence equation of the charging efficiency on the charging current, and the influence equation of the charging stage on the charging current;

[0014] S40, according to the historical charging process data, set a time window, and extract characteristic parameters within the window, including average charging current, maximum charging current, current change rate, grid voltage fluctuation, battery temperature change rate, state of charge change speed, charging efficiency change, and charging stage conversion frequency;

[0015] S50, establishing a training data set according to the multiple groups of historical charging process data with overcurrent phenomena, including the characteristic parameters, and the time when the overcurrent phenomenon occurs and the corresponding current, and using the training data set to train a neural network model with a multi-branch structure to obtain an overcurrent prediction model, wherein the training input is the characteristic parameters, and the training output is the time when the overcurrent phenomenon occurs and the corresponding current;

[0016] S60, acquiring the charging process data of the high-power DC charging pile to be tested in real time, and extracting characteristic parameters of the charging process data of the high-power DC charging pile to be tested by using the step of S40, recording them as characteristic parameters to be tested, calculating the characteristic parameters to be tested by using the overcurrent prediction model, and outputting the predicted time when overcurrent will occur in the high-power DC charging pile to be tested and the corresponding current.

[0017] Specifically, the step S10 specifically includes: establishing a charging response equation group with error terms that considers the grid power-in parameters, charging parameters, and battery parameters of the new energy vehicle of the high-power DC charging pile, including a voltage balance equation, a power transmission equation, a current limit equation, a temperature change equation, a battery state of charge equation, and a charging efficiency equation. Among them, the grid power-in parameters include input voltage, input frequency, and input harmonics; the charging parameters include charging time, charging voltage, charging current, charging stage, charging power, charging efficiency, and charging interface temperature; the battery parameters include battery capacity, battery internal resistance, battery temperature, state of charge, health status, and charging rate.

[0018] Wherein, the step S20 specifically includes: obtaining multiple groups of historical charging process data with overcurrent phenomena, including grid power-in data, charging data, and battery data, and fitting the error term of each equation in the charging response equation group according to the charging process data to obtain the fitted charging response equation group. Wherein, the grid power-in data is data corresponding to the grid power-in parameters, the charging data is data corresponding to the charging parameters, and the battery data is data corresponding to the battery parameters.

[0019] Wherein, the step S30 specifically includes: taking the charging current change as the dependent variable, solving the charging response equation group, and obtaining the charging current influence equation group, including the influence equation of the grid voltage on the charging current, the influence equation of the charging power on the charging current, the influence equation of the battery temperature on the charging current, the influence equation of the battery state of charge on the charging current, the influence equation of the charging efficiency on the charging current, and the influence equation of the charging stage on the charging current.

[0020] Wherein, the step S40 specifically includes: according to the historical charging process data, setting a time window, extracting characteristic parameters within the window, including average charging current, maximum charging current, current change rate, grid voltage fluctuation, battery temperature change rate, charge state change speed, charging efficiency change and charging stage conversion frequency. Wherein, the predetermined collection time threshold is 10 minutes to 30 minutes.

[0021] Wherein, the step S50 specifically includes: establishing a training data set according to the multiple groups of historical charging process data with overcurrent phenomena, including the characteristic parameters, the time when the overcurrent phenomenon occurs and the corresponding current, and using the training data set to train a neural network model with a multi-branch structure to obtain an overcurrent prediction model, wherein the training input is the characteristic parameters, and the training output is the time when the overcurrent phenomenon occurs and the corresponding current. The neural network model includes a long-term characteristic subnetwork, a short-term characteristic subnetwork, a time series characteristic subnetwork and a fusion subnetwork.

[0022] Wherein, the step S50 specifically includes:

[0023] 1) Collect the charging process data of the high-power DC charging pile to be tested in real time through the charging pile port, including grid input voltage, grid input frequency, grid input harmonics, charging time, charging voltage, charging current, charging stage, charging power, charging efficiency, charging interface temperature, battery capacity, battery internal resistance, battery temperature, battery state of charge, battery health status, charging rate and other parameters.

[0024] 2) Use the window method to extract the characteristic parameters of these charging process data, including average charging current, maximum charging current, current change rate, grid voltage fluctuation, battery temperature change rate, charge state change speed, charging efficiency change and charging stage conversion frequency. The window length can be set to 30 seconds to 1 minute, that is, the charging process data is divided into a continuous time window of 0.5-1 minute for characteristic parameter extraction.

[0025] 3) The extracted characteristic parameters are recorded as the characteristic parameters to be measured for subsequent overcurrent prediction.

[0026] 4) The obtained characteristic parameters to be measured are input into the overcurrent prediction model trained in step S50 for calculation, and the model will output the predicted time when overcurrent will occur and the corresponding current of the high-power DC charging pile to be measured.

[0027] Among them, the grid power input parameters include input voltage, input frequency, and input harmonics; the charging parameters include charging time, charging voltage, charging current, charging stage, charging power, charging efficiency, and charging interface temperature; the battery parameters include battery capacity, battery internal resistance, battery temperature, state of charge, health status, and charging rate.

[0028] Furthermore, the grid power input data is data corresponding to the grid power input parameters, the charging data is data corresponding to the charging parameters, and the battery data is data corresponding to the battery parameters.

[0029] Furthermore, the neural network model includes a long-term feature sub-network, a short-term feature sub-network, a temporal feature sub-network and a fusion sub-network.

[0030] Furthermore, the long-term feature refers to a feature whose continuous collection time exceeds a predetermined collection time threshold, and the short-term feature refers to a feature whose continuous collection time is less than a predetermined collection time threshold.

[0031] Furthermore, the predetermined collection time threshold is 10 minutes to 30 minutes.

[0032] The long-term feature subnetwork is used to capture the long-term trends and patterns in the charging process. The training input during training is the long-term feature parameters in the historical charging data, and the training output is the long-term feature vector. A recurrent neural network (RNN) or a long short-term memory network (LSTM) is used. Specifically, the structure is a multi-layer LSTM network, including an input layer, a hidden layer, and an output layer.

[0033] The short-term feature subnetwork is used to capture short-term fluctuations and mutations in the charging process. The training input during training is the characteristic parameters of the charging data in the most recent period of time, and the training output is a short-term feature vector. A convolutional neural network (CNN) is used, and its structure specifically consists of multiple convolutional layers, pooling layers, and fully connected layers. In the convolutional layer, the charging current influence equation group is used as the initialization value of the convolution kernel to improve the sensitivity of the model to charging current changes;

[0034] The time series feature sub-network is used to capture the time series features of the charging process. The training input during training is the charging data feature parameters arranged in chronological order, and the training output is the time series feature vector. The time convolution network (TCN) or attention mechanism is adopted, and its structure is specifically multi-layer causal convolution layers and residual connections;

[0035] The fusion feature sub-network is used to integrate long-term, short-term and time series features. The training input during training is a combination of long-term feature vectors, short-term feature vectors and time series feature vectors. The training output is an overcurrent prediction result. A fully connected neural network is used. Its structure is specifically a multi-layer fully connected layer, including an input layer, multiple hidden layers and an output layer.

[0036] The difference between the long-term feature sub-network and the short-term feature sub-network lies in the different time scales they process. The long-term feature sub-network focuses on the trend and pattern of the entire charging process, while the short-term feature sub-network focuses on capturing fluctuations and mutations in a short period of time, thereby improving the model's ability to recognize features of different time scales.

[0037] Furthermore, the predetermined collection time threshold is 10 minutes.

[0038] Furthermore, the time window is set to 30 seconds to 1 minute.

[0039] The charging response equations include:

[0040] 1. Voltage balance equation:

[0041] ;

[0042] In the formula, is the grid input voltage (V), which is obtained by real-time measurement of the grid monitoring equipment; is the charging pile output voltage (V), measured by the charging pile output voltage sensor; is the charging current (A), measured by the current sensor at the output end of the charging pile; is the equivalent resistance (Ω), including the line resistance and the internal resistance of the charging pile, which is obtained by pre-measurement and calculation; is the equivalent inductance (H), including the line inductance and the internal inductance of the charging pile, obtained by pre-measurement and calculation; is the charging current change rate (A / s), which is calculated by continuously measuring the charging current; is the voltage error term (V), which is used to compensate for model error and measurement error.

[0043] 2. Power transfer equation:

[0044] ;

[0045] In the formula, is the charging power (W); is the output voltage of the charging pile (V); is the charging current (A); is the charging efficiency, ranging from [0,1], calculated by measuring the input and output power in real time; is the grid input voltage (V); is the grid input current (A), measured by the grid-side current sensor; is the power loss (W), including transformer loss, switch loss, etc., obtained through thermal measurement and calculation; is the power error term (W).

[0046] 3. Current limit equation:

[0047] ;

[0048] In the formula, is the charging current (A); The maximum output current of the charging pile (A), which is determined by the specifications of the charging pile; The maximum output power of the charging pile (W), which is determined by the specifications of the charging pile; is the output voltage of the charging pile (V); is the rated capacity of the battery (Ah), obtained from the vehicle BMS; is the charging time (h), counting from the start of charging; is the charging rate factor, ranging from [0,1], set according to the battery type and manufacturer's recommendations; is the current error term (A).

[0049] 4. Temperature change equation:

[0050] ;

[0051] In the formula, is the battery temperature (K), measured by the battery temperature sensor; is time(s); is the battery mass (kg), obtained from the vehicle BMS; is the specific heat capacity of the battery (J / (kg·K)), which can be obtained by looking up the table according to the battery type; is the charging current (A); is the battery internal resistance (Ω), obtained from the vehicle BMS or measured by the pulse test method; is the heat transfer coefficient (W / (m²·K)), which is determined according to the characteristics of the battery cooling system; is the battery heat dissipation surface area (m²), calculated based on the battery pack size; is the ambient temperature (K), measured by the ambient temperature sensor; is the temperature error term (K / s).

[0052] 5. Battery charge state equation:

[0053] ;

[0054] In the formula, is the current state of charge, range [0,1]; is the initial state of charge, obtained from the vehicle's BMS; is the nominal capacity of the battery (Ah), obtained from the vehicle BMS; is the Coulomb efficiency, ranging from [0,1], and is dynamically calculated according to the charging conditions; is the charging current (A); is the charging time (h); is the SOC error term.

[0055] 6. Charging efficiency equation:

[0056] ;

[0057] In the formula, is the charging efficiency, ranging from [0,1]; is the charging power (W); is the grid input power (W); is the output voltage of the charging pile (V); is the charging current (A); is the grid input voltage (V); is the grid input current (A); is the power loss (W); is the efficiency error term.

[0058] The charging current affects the equations:

[0059] 1. The influence equation of grid voltage on charging current:

[0060] ;

[0061] In the formula, is the charging current (A); is the grid input voltage (V); is the equivalent resistance (Ω); is the equivalent inductance (H); is the system response time (s), obtained through step response test; is an imaginary unit; is the grid angular frequency (rad / s), , is the grid frequency (Hz); is the voltage-current error term (A / V).

[0062] 2. The effect of charging power on charging current:

[0063] ;

[0064] In the formula, is the charging current (A); is the charging power (W); is the output voltage of the charging pile (V); is the equivalent resistance (Ω); is the power-current effect error term (A / W).

[0065] 3. The effect of battery temperature on charging current equation:

[0066] ;

[0067] In the formula, is the charging current (A); is the battery temperature (K); is the charging current at reference temperature (A), usually measured at 25°C; is the reference temperature (K), usually 298.15K (25°C); is the temperature influence coefficient, obtained by fitting the experimental data; is the temperature-current effect error term (A / K).

[0068] 4. The influence equation of battery state of charge on charging current:

[0069] ;

[0070] In the formula, is the charging current (A); is the battery charge state, range [0,1]; is the maximum output current of the charging pile (A); is the SOC influence coefficient, obtained by fitting the experimental data; is the SOC-current influence error term (A).

[0071] 5. The effect of charging efficiency on charging current:

[0072] ;

[0073] In the formula, is the charging current (A); is the charging efficiency, ranging from [0,1]; is the grid input power (W); is the output voltage of the charging pile (V); is the efficiency-current effect error term (A).

[0074] 6. The effect of charging stage on charging current:

[0075] ;

[0076] In the formula, is the charging current (A); is the maximum output current of the charging pile (A);

[0077] is the charging stage function, defined as follows:

[0078]

[0079] is the charging time (h); is the charging phase conversion time point (h); is the charging stage parameter, obtained by fitting the experimental data; is the error term affecting the charging stage (A).

[0080] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned high-power DC charging pile overcurrent prediction method.

[0081] A third aspect of the present invention provides a high-power DC charging pile overcurrent prediction system, which includes the above-mentioned computer-readable storage medium.

[0082] Compared with the prior art, the high-power DC charging pile overcurrent prediction method, medium and system provided by the present invention have the following beneficial effects:

[0083] 1. Establish a set of charging response equations to accurately describe the dynamic changes of electrical and thermal parameters during the charging process

[0084] In step S10, the present invention establishes a charging response equation group including a voltage balance equation, a power transmission equation, a current limit equation, a temperature change equation, a battery charge state equation and a charging efficiency equation. The equation group not only takes into account the grid input parameters, the charging pile output parameters and the battery parameters, but also introduces various error terms to compensate for the measurement error and modeling deviation, which can more accurately reflect the complex dynamic characteristics of the high-power DC charging process. Compared with the simple overcurrent threshold monitoring in the prior art, the charging response equation group of the present invention can more comprehensively characterize the actual working state of the charging system.

[0085] 2. Use a multi-branch neural network model to predict overcurrent and improve prediction accuracy

[0086] The present invention adopts a neural network model with a multi-branch structure in step S50, including a long-term feature sub-network, a short-term feature sub-network, a time series feature sub-network and a fusion sub-network. This multi-branch structure can fully capture the characteristics of different time scales in the charging process, and can not only focus on the long-term trend of the entire charging process, but also finely analyze short-term fluctuations and mutations. Compared with the single overcurrent detection method in the prior art, the neural network model of the present invention has stronger feature extraction and prediction capabilities, and can more accurately predict the impending overcurrent situation.

[0087] 3. Real-time monitoring of charging process data, timely implementation of targeted over-current protection measures

[0088] The present invention collects the charging process data of the charging pile to be tested in real time in step S60, and extracts key characteristic parameters to provide input for subsequent overcurrent prediction. Once the overcurrent is predicted to occur, the present invention can take targeted protection measures such as power reduction, switching charging mode or emergency charging stop in time in step S80, so as to avoid serious damage to the charging equipment and battery caused by overcurrent. Compared with the prior art that passively waits for overcurrent to occur before performing protection, the method of the present invention can predict in advance and take active precautions, greatly improving the safety of the charging system.

[0089] In general, the high-power DC charging overcurrent prediction method proposed in the present invention makes full use of the dynamic characteristics of various physical and electrical parameters in the charging process, adopts precise mathematical modeling and high-performance neural network algorithms, and realizes accurate prediction of impending overcurrent, thus solving the technical problem that the existing technology is difficult to predict overcurrent of high-power DC charging piles. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0091] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0092] like Figure 1 As shown, it is a flow chart of a high-power DC charging pile overcurrent prediction method provided by the first aspect of the present invention, and the method comprises the following steps:

[0093] S10. Establish a charging response equation group with error terms that considers the grid power input parameters, charging parameters, and battery parameters of the new energy vehicle of the high-power DC charging pile, including a voltage balance equation, a power transmission equation, a current limit equation, a temperature change equation, a battery state of charge equation, and a charging efficiency equation;

[0094] S20, obtaining multiple groups of historical charging process data with overcurrent phenomena, including grid power input data, charging data, and battery data, and fitting the error term of each equation in the charging response equation group according to the charging process data to obtain a fitted charging response equation group;

[0095] S30, taking the charging current change as the dependent variable, solving the charging response equation group, and obtaining the charging current influence equation group, including the influence equation of the grid voltage on the charging current, the influence equation of the charging power on the charging current, the influence equation of the battery temperature on the charging current, the influence equation of the battery state of charge on the charging current, the influence equation of the charging efficiency on the charging current, and the influence equation of the charging stage on the charging current;

[0096] S40, according to the historical charging process data, set a time window, and extract characteristic parameters within the window, including average charging current, maximum charging current, current change rate, grid voltage fluctuation, battery temperature change rate, state of charge change speed, charging efficiency change, and charging stage conversion frequency;

[0097] S50, establishing a training data set based on multiple groups of historical charging process data with overcurrent phenomena, including characteristic parameters, and the time when the overcurrent phenomenon occurs and the corresponding current, and using the training data set to train a neural network model with a multi-branch structure to obtain an overcurrent prediction model, wherein the training input is the characteristic parameters, and the training output is the time when the overcurrent phenomenon occurs and the corresponding current;

[0098] S60, acquiring the charging process data of the high-power DC charging pile to be tested in real time, and extracting characteristic parameters of the charging process data of the high-power DC charging pile to be tested by using the step of S40, recording them as characteristic parameters to be tested, calculating the characteristic parameters to be tested by using an overcurrent prediction model, and outputting the predicted time when overcurrent will occur in the high-power DC charging pile to be tested and the corresponding current.

[0099] The specific implementation methods of the above steps are described in detail below:

[0100] Step S10: Establish a charging response equation group with error terms considering the grid power input parameters, charging parameters, and battery parameters of the new energy vehicle of the high-power DC charging pile

[0101] In step S10, a mathematical model covering various key parameters of a high-power DC charging pile needs to be established. The model can accurately describe the relationship between variables such as voltage, current, power, temperature, and state of charge during the charging process, and take into account measurement and model errors. This mathematical model is called the charging response equation group, which includes the following equations:

[0102] 1. Voltage balance equation:

[0103] ;

[0104] In the formula, is the grid input voltage (V), which is obtained by real-time measurement of the grid monitoring equipment; is the output voltage of the charging pile (V), measured by the voltage sensor at the output end of the charging pile; is the charging current (A), measured by the current sensor at the output end of the charging pile; is the equivalent resistance (Ω), including the line resistance and the internal resistance of the charging pile, obtained by pre-measurement and calculation; is the equivalent inductance (H), including the line inductance and the internal inductance of the charging pile, obtained by pre-measurement and calculation; is the charging current change rate (A / s), which is calculated by continuously measuring the charging current; is the voltage error term (V), which is used to compensate for model error and measurement error.

[0105] 2. Power transfer equation:

[0106] ;

[0107] In the formula, is the charging power (W); is the output voltage of the charging pile (V); is the charging current (A); is the charging efficiency, in the range [0,1], calculated by measuring the input and output power in real time; is the grid input voltage (V); is the grid input current (A), measured by the grid-side current sensor; is the power loss (W), including transformer loss, switch loss, etc., obtained through heat measurement and calculation; is the power error term (W).

[0108] 3. Current Limit Equation:

[0109] ;

[0110] In the formula, is the charging current (A); is the maximum output current of the charging pile (A), which is determined by the specifications of the charging pile; is the maximum output power of the charging pile (W), which is determined by the specifications of the charging pile; is the output voltage of the charging pile (V); is the rated capacity of the battery (Ah), obtained from the vehicle BMS; is the charging time (h), counting from the start of charging; is the charging rate factor, ranging from [0,1], set according to the battery type and manufacturer's recommendations; is the current error term (A).

[0111] 4. Temperature change equation:

[0112] ;

[0113] In the formula, is the battery temperature (K), measured by the battery temperature sensor; is time(s); is the battery mass (kg), obtained from the vehicle BMS; is the specific heat capacity of the battery (J / (kg·K)), which can be obtained by looking up the table according to the battery type; is the charging current (A); is the battery internal resistance (Ω), obtained from the vehicle BMS or measured by the pulse test method; is the heat transfer coefficient (W / (m²·K)), which is determined according to the characteristics of the battery cooling system; is the heat dissipation surface area of ​​the battery (m²), calculated based on the battery pack size; is the ambient temperature (K), measured by the ambient temperature sensor; is the temperature error term (K / s).

[0114] 5.Battery charge state equation:

[0115] ;

[0116] In the formula, is the current state of charge, range [0,1]; is the initial state of charge, obtained from the vehicle BMS; is the nominal capacity of the battery (Ah), obtained from the vehicle BMS; is the coulombic efficiency, ranging from [0,1], and is dynamically calculated according to the charging conditions; is the charging current (A); is the charging time (h); is the SOC error term.

[0117] 6. Charging efficiency equation:

[0118] ;

[0119] In the formula, is the charging efficiency, range [0,1]; is the charging power (W); is the grid input power (W); is the output voltage of the charging pile (V); is the charging current (A); is the grid input voltage (V); is the grid input current (A); is the power loss (W); is the efficiency error term.

[0120] The above six equations constitute the charging response equation group, which describes the complex relationship between grid input parameters, charging parameters and battery parameters during high-power DC charging, and takes into account various measurement and model errors. This equation group provides a mathematical basis for subsequent overcurrent prediction. It should be pointed out that the variables and their physical meanings in each equation are clearly defined and there is no ambiguity. Moreover, even if some uncommon functions are involved, such as the effect of temperature on charging current, detailed descriptions will be given in the corresponding equations.

[0121] Step S20: Acquire multiple groups of historical charging process data with overcurrent phenomena, and fit the error term of each equation in the charging response equation group according to the charging process data to obtain a fitted charging response equation group.

[0122] In step S20, a large amount of actual charging process data with overcurrent phenomenon needs to be collected, and the charging response equation group established in step S10 is fitted based on these data to more accurately determine the error terms in each equation.

[0123] Specifically, the following three types of data need to be collected:

[0124] 1) Grid power data: including input voltage , Input frequency And parameters such as input harmonics;

[0125] 2) Charging data: including charging time , Charging voltage , Charging current , Charging stage , Charging power , Charging efficiency And the charging port temperature etc. parameters;

[0126] 3) Battery data: including battery capacity , Battery internal resistance , Battery temperature , Battery state of charge , Battery health status And charging rate And other parameters.

[0127] After collecting these historical charging process data, it is necessary to calculate the error term in each equation For example, the error term of the voltage balance equation is , a multivariate linear model can be fitted using the least squares method:

[0128] ;

[0129] In the formula, is the fitting coefficient. Similarly, the error terms of other equations can also be fitted. Through such a fitting process, a more accurate charging response equation group is obtained, which can better describe the various dynamic characteristics of the actual charging process.

[0130] Step S30: Taking the charging current change as the dependent variable, solving the charging response equation group to obtain the charging current influence equation group

[0131] In step S30, the charging current As the dependent variable, the charging response equations established in steps S10 and S20 are solved to obtain a set of equations representing the influence of various factors on the charging current, namely, the charging current influence equations. These equations can be used to analyze and predict the changing law of the charging current.

[0132] 1. The influence equation of grid voltage on charging current:

[0133] ;

[0134] In the formula, is the charging current (A); is the grid input voltage (V); is the equivalent resistance (Ω); is the equivalent inductance (H); is the system response time (s), obtained through step response test; is an imaginary unit; is the grid angular frequency (rad / s), , is the grid frequency (Hz); is the voltage-current error term (A / V).

[0135] 2. The effect equation of charging power on charging current:

[0136] ;

[0137] In the formula, is the charging current (A); is the charging power (W); is the output voltage of the charging pile (V); is the equivalent resistance (Ω); is the power-current effect error term (A / W).

[0138] 3.Equation of the effect of battery temperature on charging current:

[0139] ;

[0140] In the formula, is the charging current (A); is the battery temperature (K); is the charging current at reference temperature (A), usually measured at 25°C; is the reference temperature (K), usually 298.15K (25°C); is the temperature influence coefficient, obtained by fitting the experimental data; is the temperature-current effect error term (A / K).

[0141] 4. The influence equation of battery state of charge on charging current:

[0142] ;

[0143] In the formula, is the charging current (A); is the battery charge state, range [0,1]; is the maximum output current of the charging pile (A); is the SOC influence coefficient, obtained by fitting the experimental data; is the SOC-current influence error term (A).

[0144] 5. The effect of charging efficiency on charging current:

[0145] ;

[0146] In the formula, is the charging current (A); is the charging efficiency, range [0,1]; is the grid input power (W); is the output voltage of the charging pile (V); is the efficiency-current effect error term (A).

[0147] 6. The charging stage affects the charging current equation:

[0148] ;

[0149] In the formula, is the charging current (A); is the maximum output current of the charging pile (A);

[0150] is the charging stage function, which is defined as follows:

[0151] ;

[0152] is the charging time (h); is the time point for the charging phase transition (h); For charging stage

[0153] Step S40: According to the historical charging process data, set a time window and extract characteristic parameters within the window

[0154] In step S40, the historical charging process data collected in step S20 is used to extract some key characteristic parameters. The specific implementation is as follows:

[0155] 1. Set the time window: According to the actual situation, the charging process data can be divided into continuous time windows ranging from 30 seconds to 1 minute. The selection of the length of this time window requires a trade-off between analysis accuracy and computational complexity.

[0156] 2. Extract characteristic parameters: For the charging process data in each time window, extract the following characteristic parameters:

[0157] (1) Average charging current : Calculate the charging current within this time window The average value of .

[0158] (2) Maximum charging current : Find the charging current within this time window The maximum value of .

[0159] (3) Current change rate : Calculate the charging current within this time window The rate of change.

[0160] (4) Grid voltage fluctuation : Calculate the grid input voltage within this time window The fluctuation range.

[0161] (5) Battery temperature change rate : Calculate the battery temperature within the time window The rate of change.

[0162] (6) Speed ​​of charge state change : Calculate the battery state of charge within the time window The speed of change.

[0163] (7) Changes in charging efficiency : Calculate the charging efficiency within this time window The range of change.

[0164] (8) Conversion frequency during charging phase : Count the number of conversions in the charging phase within this time window.

[0165] 3. Record characteristic parameters: Record the characteristic parameters extracted above as input for the subsequent overcurrent prediction model.

[0166] Through such a feature extraction process, a set of key parameters that can reflect the dynamic characteristics of the charging process can be obtained, laying the foundation for subsequent overcurrent prediction. The selection of the time window needs to be weighed according to the actual situation, usually 30 seconds to 1 minute is a reasonable range.

[0167] Step S50: Establish a training data set based on the multiple groups of historical charging process data with overcurrent phenomena, and use the training data set to train a neural network model with a multi-branch structure to obtain an overcurrent prediction model

[0168] In step S50, a training data set is established using the characteristic parameters extracted in step S40, as well as the time and current value of overcurrent phenomena during the historical charging process. Then, a neural network model with a multi-branch structure is used for training to obtain the final overcurrent prediction model. The specific implementation is as follows:

[0169] 1. Build training dataset:

[0170] (1) The feature parameters extracted in step S40 are used as training input, including the average charging current , Maximum charging current , Current change rate , Grid voltage fluctuation , Battery temperature change rate , Charge state change speed , Charging efficiency changes and charging phase switching frequency wait.

[0171] (2) The time when the overcurrent phenomenon will occur during the historical charging process and the corresponding overcurrent as training output.

[0172] (3) Combine the above input and output data into a training data set.

[0173] 2. Training multi-branch neural network model:

[0174] (1) The neural network model consists of four sub-networks: long-term feature sub-network, short-term feature sub-network, temporal feature sub-network and fusion sub-network.

[0175] (2) The long-term feature sub-network uses a recurrent neural network (RNN) or a long short-term memory network (LSTM) to capture long-term trends and patterns in the charging process.

[0176] (3) The short-term feature subnetwork uses a convolutional neural network (CNN) to capture short-term fluctuations and mutations during the charging process. In the convolution layer, the charging current influence equations obtained in step S30 are used as the initialization value of the convolution kernel to improve the model's sensitivity to charging current changes.

[0177] (4) The time series feature sub-network adopts the temporal convolutional network (TCN) or attention mechanism to capture the time series features of the charging process.

[0178] (5) The fusion sub-network uses a fully connected neural network to integrate long-term, short-term and time series characteristics and output overcurrent prediction results.

[0179] (6) Use the training data set collected in step S20 to perform end-to-end training on the above-mentioned multi-branch neural network model to obtain the final overcurrent prediction model.

[0180] Through such a multi-branch neural network structure, the characteristics of different time scales in the charging process can be fully utilized to improve the accuracy of overcurrent prediction. Among them, the difference between the long-term and short-term feature sub-networks lies in the different time scales they process. The long-term characteristics focus on the trends and patterns of the entire charging process, while the short-term characteristics focus on capturing fluctuations and mutations in a short period of time, thereby improving the model's ability to recognize characteristics of different time scales.

[0181] Step S60: Acquire the charging process data of the high-power DC charging pile to be tested in real time, and extract the characteristic parameters of the charging process data of the high-power DC charging pile to be tested by using the step S40, and record them as the characteristic parameters to be tested.

[0182] In step S60, it is necessary to obtain the charging process data of the high-power DC charging pile to be tested in real time, and extract the same characteristic parameters as step S40 as input for subsequent overcurrent prediction. The specific implementation is as follows:

[0183] 1. Real-time data collection:

[0184] (1) Collect the charging process data of the high-power DC charging pile to be tested in real time through the charging pile port, including the grid input voltage , Grid input frequency , grid input harmonics, charging time , Charging voltage , Charging current , Charging stage , Charging power , Charging efficiency , Charging port temperature , Battery capacity , Battery internal resistance , Battery temperature , Battery state of charge , Battery health status and charging rate And other parameters.

[0185] 2. Feature parameter extraction:

[0186] (1) Set a time window ranging from 30 seconds to 1 minute to segment the collected charging process data.

[0187] (2) Extract the average charging current from the data in each time window , Maximum charging current , Current change rate , Grid voltage fluctuation , Battery temperature change rate , Charge state change speed , Charging efficiency changes and charging phase switching frequency And other characteristic parameters.

[0188] (3) The extracted characteristic parameters are recorded as "characteristic parameters to be measured" for subsequent overcurrent prediction.

[0189] Through such real-time data collection and feature parameter extraction, the dynamic characteristics of the current charging process of the high-power DC charging pile to be tested can be obtained, providing the required input data for subsequent overcurrent prediction. The selection of the time window is the same as step S40, and a trade-off between analysis accuracy and computational complexity is required.

[0190] In general, steps S10 to S60 describe a high-power DC charging overcurrent prediction method based on a charging response equation group and a multi-branch neural network model. This method makes full use of the dynamic characteristics of various physical and electrical parameters in the charging process, and by establishing an accurate mathematical model and training a high-performance neural network, it can accurately predict the impending overcurrent phenomenon and provide protection for the safe operation of the charging system.

[0191] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned high-power DC charging pile overcurrent prediction method.

[0192] A third aspect of the present invention provides a high-power DC charging pile overcurrent prediction system, which includes the above-mentioned computer-readable storage medium.

[0193] Specifically, the principle of the present invention is:

[0194] 1. Establish a charging response equation group to describe the dynamic relationship between various parameters during the charging process

[0195] In step S10, a charging response equation group is established, including voltage balance, power transmission, current limit, temperature change, battery charge state and charging efficiency, etc. These equations not only cover the grid input parameters, charging pile output parameters and battery parameters, but also introduce various error terms, which can more accurately reflect the complex dynamic relationship between these physical quantities in the high-power DC charging process.

[0196] The significance of establishing such a charging response equation group is:

[0197] (1) It provides a mathematical basis for subsequent overcurrent prediction. The changes in various parameters during the charging process will affect the charging current, and the charging response equations describe the functional relationship between these parameters and the current.

[0198] (2) It can make up for the deficiency of existing technologies that only focus on monitoring a single parameter. By establishing a mathematical model covering multiple factors including the power grid, charging piles and batteries, we can have a more comprehensive understanding of the working status of the charging process.

[0199] (3) It integrates the measured data and compensates for the limitations of theoretical modeling by fitting the error term. This not only retains the interpretability of the physical model, but also improves the accuracy of the model, laying a good foundation for subsequent overcurrent prediction.

[0200] 2. Use multi-branch neural network model for overcurrent prediction

[0201] In step S50, a multi-branch neural network model is trained to predict overcurrent. The model includes four sub-networks: a long-term feature sub-network, a short-term feature sub-network, a time series feature sub-network, and a fusion sub-network. The design of this multi-branch structure conforms to the basic principle of "combination is better than single" in the field of artificial intelligence.

[0202] Specifically:

[0203] (1) The long-term feature sub-network uses a recurrent neural network or a long short-term memory network, which is good at capturing long-term trends and patterns in the charging process.

[0204] (2) The short-term feature subnetwork uses a convolutional neural network, which is good at analyzing short-term fluctuations and mutations in the charging process. The charging current influence equation group obtained in step S30 is also used as the initialization value of the convolution kernel to improve the model's sensitivity to charging current changes.

[0205] (3) The time series feature sub-network adopts a temporal convolutional network or an attention mechanism, and is good at analyzing the time series characteristics of the charging process.

[0206] (4) The fusion sub-network adopts a fully connected neural network, which is responsible for integrating the features extracted by the first three sub-networks and outputting the final overcurrent prediction results.

[0207] This multi-branch structure design fully utilizes the advantages of neural networks in feature extraction and pattern recognition. Different sub-networks focus on capturing the features of different time scales during the charging process. By combining these features, the accuracy of overcurrent prediction can be greatly improved. Compared with the single overcurrent detection method in the prior art, the neural network model of the present invention has stronger functions and performance.

[0208] 3. Real-time monitoring of charging process data, timely implementation of targeted over-current protection measures

[0209] In step S60, the charging process data of the charging pile to be tested is collected in real time, and key characteristic parameters are extracted. These characteristic parameters include both the statistics of the charging current and other related parameters such as the grid voltage and battery temperature. Then in step S70, these characteristic parameters are input into the trained neural network model to predict the impending overcurrent situation.

[0210] Once the possible overcurrent is predicted, corresponding protection measures can be taken in step S80, such as reducing the charging power, switching the charging mode, or emergency stopping charging. This advance prediction and active protection method can greatly shorten the response time of the system and effectively prevent overcurrent from causing serious damage to the equipment and battery. Compared with the prior art that passively waits for overcurrent to occur before performing protection, the method of the present invention significantly improves the safety of the charging system.

[0211] In order to better understand and implement the present invention, an embodiment of a specific application scenario of the present invention is provided below: A new energy vehicle company is developing a high-performance pure electric vehicle model, and its large-capacity power battery pack needs to adopt fast charging technology. To this end, the company cooperated with a professional DC charging pile manufacturer to jointly develop a 350kW high-power DC charging system. This charging system consists of a grid-side transformer, a charging control unit, a charging interface, and an on-board battery management system. It can fully charge the electric vehicle battery pack from an initial charge state of 20% to 90% within 20 minutes.

[0212] However, in actual application, it is found that the charging current occasionally surges during the charging process, causing the overcurrent protection device of the charging pile to frequently operate, seriously affecting the continuity and stability of the charging service. In order to solve this problem, the researchers decided to use the high-power DC charging overcurrent prediction method proposed in this invention to optimize the charging system.

[0213] Step S10: Establishing a charging response equation group

[0214] First, the researchers established the following charging response equations based on various parameters of the charging system:

[0215] 1. Voltage balance equation:

[0216] ;

[0217] in, is the grid input voltage, is the output voltage of the charging pile, is the charging current, is the equivalent resistance, is the equivalent inductance, is the charging current change rate, is the voltage error term.

[0218] 2. Power transfer equation:

[0219] ;

[0220] in, is the charging power, is the charging efficiency, is the grid input voltage, is the grid input current, is the power loss, is the power error term.

[0221] 3. Current Limit Equation:

[0222] ;

[0223] in, is the maximum output current of the charging pile, is the maximum output power of the charging pile, is the rated capacity of the battery, For charging time, is the charging rate coefficient, is the current error term.

[0224] 4. Temperature change equation:

[0225] ;

[0226] in, is the battery temperature, is the battery quality, is the specific heat capacity of the battery, is the internal resistance of the battery, is the heat transfer coefficient, is the heat dissipation surface area of ​​the battery, is the ambient temperature, is the temperature error term.

[0227] 5.Battery charge state equation:

[0228] ;

[0229] in, is the initial state of charge, is the nominal capacity of the battery, is the Coulomb efficiency, is the SOC error term.

[0230] 6. Charging efficiency equation:

[0231] ;

[0232] in, The input power to the grid is is the efficiency error term.

[0233] By establishing such a set of charging response equations covering all aspects of the power grid, charging piles and batteries, the researchers laid a mathematical foundation for subsequent overcurrent prediction and provided theoretical support for the safe operation of the charging system.

[0234] Step S20: Obtain historical charging process data and fit the charging response equation group

[0235] After establishing the charging response equations, the researchers collected historical charging data of overcurrent phenomena in the actual application of the charging system. These data include grid input voltage, grid input frequency, grid input harmonics, charging time, charging voltage, charging current, charging stage, charging power, charging efficiency, charging interface temperature, battery capacity, battery internal resistance, battery temperature, battery state of charge, battery health status and charging rate.

[0236] By analyzing these historical charging data, the researchers found that in some charging processes, when the battery internal resistance suddenly increases or the ambient temperature is too high, the charging current will increase sharply for a short period of time, thereby triggering the overcurrent protection device of the charging pile. In order to more accurately describe the dynamic change characteristics of each parameter in this charging process, the researchers fitted and optimized the charging response equations established in step S10. Specifically, they used the least squares method to fit the error term in each equation. Parameter estimation was performed and a set of charging response equations that are closer to reality was obtained.

[0237] For example, for the voltage error term in the voltage balance equation , after fitting, we get:

[0238] ;

[0239] This shows that grid voltage fluctuations, charging current changes, and charging current change rates will all have a certain error effect on the output voltage of the charging pile. Similarly, the error terms in other equations have also been fitted and optimized accordingly. In this way, the researchers made the charging response equations more accurately describe the complex dynamic characteristics of each parameter in the high-power DC charging process.

[0240] Step S30: Solve the charging response equations to obtain the charging current influence equations

[0241] Based on the optimized charging response equations in step S20, the researchers further solved a set of equations that characterize the influence of various factors on the charging current, namely, the charging current influence equations. These equations are as follows:

[0242] 1. The influence equation of grid voltage on charging current:

[0243] ;

[0244] in, is the system response time, is the grid angular frequency.

[0245] 2. The effect equation of charging power on charging current:

[0246] ;

[0247] 3.Equation of the effect of battery temperature on charging current:

[0248] ;

[0249] in, is the charging current at reference temperature, reference temperature .

[0250] 4. The influence equation of battery state of charge on charging current:

[0251] ;

[0252] 5. The effect of charging efficiency on charging current:

[0253] ;

[0254] 6. The charging stage affects the charging current equation:

[0255]

[0256] in, For the following piecewise function:

[0257] ;

[0258] By solving the charging response equations, the researchers obtained a set of mathematical models that can describe the influence of various factors on the charging current. These models not only take into account parameters such as grid voltage, charging power, battery temperature, battery state of charge and charging efficiency, but also introduce the charging stage factor based on the characteristics of the actual charging process. Through these charging current influence equations, the researchers laid a good foundation for subsequent overcurrent prediction.

[0259] Step S40: Extracting characteristic parameters of charging process data

[0260] With the charging current influence equations established in step S30, the researchers then proceeded to extract the key characteristic parameters of the charging process data. Specifically, they set a 1-minute time window and extracted the following 8 characteristic parameters in each time window:

[0261] 1. Average charging current : Count the average value of the charging current in this time window.

[0262] 2. Maximum charging current : Find the maximum value of the charging current within the time window.

[0263] 3. Current change rate : Calculate the rate of change of the charging current within this time window.

[0264] 4. Grid voltage fluctuations : Calculate the fluctuation amplitude of the grid input voltage within the time window.

[0265] 5. Battery temperature change rate : Calculate the rate of change of battery temperature within the time window.

[0266] 6. Charge state change speed : Calculate the rate of change of the battery state of charge within the time window.

[0267] 7.Charging efficiency changes : Calculate the change in charging efficiency within this time window.

[0268] 8. Conversion frequency during charging phase : Count the number of conversions in the charging phase within this time window.

[0269] By extracting these characteristic parameters, researchers can fully reflect the dynamic change trends of various physical quantities in the high-power DC charging process, laying the foundation for subsequent overcurrent prediction. The 1-minute time window length is determined after weighing the analysis accuracy and computational complexity.

[0270] Step S50: Training a multi-branch neural network model for overcurrent prediction

[0271] With the characteristic parameters of the charging process data extracted in step S40, the researchers began to train a multi-branch neural network model to predict the impending overcurrent situation. The specific steps are as follows:

[0272] 1. Build a training dataset

[0273] (1) The eight feature parameters extracted in step S40 are used as training input, including .

[0274] (2) The time when the overcurrent phenomenon occurred during the historical charging process and the corresponding overcurrent as training output.

[0275] (3) Combine the above input and output data into a training data set.

[0276] 2. Design a multi-branch neural network model

[0277] (1) The model consists of four sub-networks: long-term feature sub-network, short-term feature sub-network, temporal feature sub-network and fusion sub-network.

[0278] (2) The long-term feature subnetwork adopts a three-layer LSTM network. The input layer accepts the feature parameters within 20 minutes of history, extracts long-term features through the LSTM hidden layer, and the output layer gives the long-term feature vector.

[0279] (3) The short-term feature subnetwork adopts a three-layer convolutional neural network. The input layer accepts the feature parameters within the last 5 minutes. The charging current influence equation group obtained in step S30 is used as the initial value of the convolution kernel in the convolution layer to improve the sensitivity to short-term fluctuations. The output layer gives a short-term feature vector.

[0280] (4) The time series feature subnetwork adopts a three-layer temporal convolutional network. The input layer arranges the feature parameter sequence within 20 minutes in chronological order, extracts the time series features through causal convolutional layers and residual connections, and the output layer gives the time series feature vector.

[0281] (5) The fusion sub-network adopts a 4-layer fully connected network. The input layer accepts the concatenation of long-term feature vectors, short-term feature vectors and time series feature vectors. After fusion processing by multiple hidden layers, the overcurrent prediction result is finally output, including the predicted overcurrent occurrence time. and the corresponding overcurrent .

[0282] 3. Training the Neural Network Model

[0283] (1) Using the historical charging process data collected in step S20 as a training set, the multi-branch neural network model is trained end-to-end.

[0284] (2) The training objective is to minimize the mean square error between the predicted overcurrent time and current and the actual overcurrent time and current.

[0285] (3) The Adam optimization algorithm was used, the learning rate was set to 0.001, and after 500 epochs of training, the overcurrent prediction model was finally obtained.

[0286] Through this multi-branch neural network structure, the researchers took full advantage of the complementary advantages of different time scale features. The long-term feature sub-network is good at capturing the trends and patterns of the entire charging process, the short-term feature sub-network focuses on analyzing local fluctuations and mutations, and the time series feature sub-network extracts time series features. The fusion sub-network combines these features and ultimately achieves accurate prediction of the overcurrent occurrence time and current value. Compared with the single overcurrent detection method in the existing technology, this multi-branch neural network model has stronger feature learning and prediction capabilities.

[0287] Step S60: Real-time acquisition of charging pile data to be tested and extraction of characteristic parameters

[0288] With the trained overcurrent prediction model, the researchers then began to monitor the high-power DC charging pile to be tested in real time. The specific steps are as follows:

[0289] 1. Real-time data collection

[0290] (1) Through the charging pile port, real-time data on the charging process of the charging pile to be tested is collected, including the grid input voltage , Grid input frequency , grid input harmonics, charging time , Charging voltage , Charging current , Charging stage , Charging power , Charging efficiency , Charging port temperature , Battery capacity , Battery internal resistance , Battery temperature , Battery state of charge , Battery health status and charging rate wait.

[0291] 2. Feature parameter extraction

[0292] (1) Set a 1-minute time window and segment the collected charging process data.

[0293] (2) Extract the average charging current from the data in each time window , Maximum charging current , Current change rate , Grid voltage fluctuation , Battery temperature change rate , Charge state change speed , Charging efficiency changes and charging phase switching frequency And other 8 characteristic parameters.

[0294] (3) The extracted characteristic parameters are recorded as "characteristic parameters to be measured" for subsequent overcurrent prediction.

[0295] Through such real-time data collection and feature parameter extraction, researchers can monitor the operating status of the charging pile to be tested in real time and provide the required input data for subsequent overcurrent prediction. The 1-minute time window length is consistent with the setting in step S40, which can ensure the analysis accuracy while controlling the calculation complexity within an acceptable range.

[0296] Step S70: Use the trained neural network model to predict overcurrent

[0297] With the "characteristic parameters to be measured" extracted in step S60, the researchers then input them into the overcurrent prediction model trained in step S50 to obtain the corresponding prediction results. The specific process is as follows:

[0298] 1. The eight characteristic parameters extracted in step S60 Input into the trained multi-branch neural network model.

[0299] 2. The neural network model first extracts the long-term, short-term and time series features of the charging process through the long-term feature sub-network, short-term feature sub-network and time series feature sub-network.

[0300] 3. Then, the fusion sub-network conducts a comprehensive analysis of these three types of features and finally outputs the corresponding overcurrent prediction results, including the predicted overcurrent occurrence time. and overcurrent .

[0301] For example, for a certain charging process, the prediction results of the neural network model show that after 16 minutes of charging, the charging current is expected to exceed the safety threshold of 540A and reach a peak of 580A. This advance prediction method reserves sufficient reaction time for subsequent overcurrent protection measures.

[0302] Step S80: Take targeted protection measures for the predicted overcurrent

[0303] Once the neural network model predicts an impending overcurrent condition, researchers can take the following targeted overcurrent protection measures appropriately in advance of the prediction moment:

[0304] 1. Appropriately reduce the charging power: According to the predicted overcurrent value, appropriately reduce the charging power , to limit the charging current Do not exceed the safety threshold. For example, the charging power can be reduced from 350kW to 280kW.

[0305] 2. Switch charging mode: When it is predicted that the constant voltage charging stage is about to begin, you can switch to the constant current charging mode in advance to avoid overcurrent caused by a sudden increase in the internal resistance of the battery.

[0306] 3. Pause charging: If it is predicted that there is a high risk of overcurrent, the charging process can be temporarily interrupted and the charging can be restarted after the battery temperature drops.

[0307] 4. Emergency stop charging: For severe overcurrent conditions that cannot be effectively controlled by the above measures, the overcurrent protection device of the charging pile can be immediately triggered to urgently cut off the charging circuit.

[0308] Through these targeted overcurrent protection measures, researchers can take corresponding preventive and control actions in a timely manner before predicting that overcurrent is about to occur, thereby minimizing damage to charging equipment and batteries caused by overcurrent and ensuring the safety of high-power DC charging process.

[0309] In summary, the high-power DC charging overcurrent prediction method proposed in the present invention makes full use of the dynamic characteristics of various physical and electrical parameters in the charging process, and realizes accurate prediction and effective protection of impending overcurrent by establishing an accurate mathematical model and training a high-performance neural network algorithm. Compared with the prior art method of passively waiting for overcurrent to occur before performing protection, this method of early prediction and active prevention can greatly improve the safety and reliability of the charging system.

[0310] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A high-power DC charging pile overcurrent prediction method, characterized in that: The following steps are involved: S10. Establish a charging response equation group with error terms that considers the grid power input parameters, charging parameters, and battery parameters of the new energy vehicle of the high-power DC charging pile, including a voltage balance equation, a power transmission equation, a current limit equation, a temperature change equation, a battery state of charge equation, and a charging efficiency equation; S20, acquiring multiple groups of historical charging process data with overcurrent phenomena, including grid power input data, charging data, and battery data, and fitting the error term of each equation in the charging response equation group according to the charging process data to obtain a fitted charging response equation group; S30, taking the charging current change as the dependent variable, solving the charging response equation group to obtain the charging current influence equation group, including the influence equation of the grid voltage on the charging current, the influence equation of the charging power on the charging current, the influence equation of the battery temperature on the charging current, the influence equation of the battery state of charge on the charging current, the influence equation of the charging efficiency on the charging current, and the influence equation of the charging stage on the charging current; S40, according to the historical charging process data, set a time window, and extract characteristic parameters within the window, including average charging current, maximum charging current, current change rate, grid voltage fluctuation, battery temperature change rate, state of charge change speed, charging efficiency change, and charging stage conversion frequency; S50, establishing a training data set according to the multiple groups of historical charging process data with overcurrent phenomena, including the characteristic parameters, and the time when the overcurrent phenomenon occurs and the corresponding current, and using the training data set to train a neural network model with a multi-branch structure to obtain an overcurrent prediction model, wherein the training input is the characteristic parameters, and the training output is the time when the overcurrent phenomenon occurs and the corresponding current; S60, acquiring the charging process data of the high-power DC charging pile to be tested in real time, and extracting characteristic parameters of the charging process data of the high-power DC charging pile to be tested by using the step of S40, recording them as characteristic parameters to be tested, calculating the characteristic parameters to be tested by using the overcurrent prediction model, and outputting the predicted time when overcurrent will occur in the high-power DC charging pile to be tested and the corresponding current.

2. A high-power DC charging pile overcurrent prediction method according to claim 1, characterized in that: The grid power input parameters include input voltage, input frequency, and input harmonics; the charging parameters include charging time, charging voltage, charging current, charging stage, charging power, charging efficiency, and charging interface temperature; the battery parameters include battery capacity, battery internal resistance, battery temperature, state of charge, health status, and charging rate.

3. A high-power DC charging pile overcurrent prediction method according to claim 2, characterized in that: The grid power-in data is data corresponding to the grid power-in parameters, the charging data is data corresponding to the charging parameters, and the battery data is data corresponding to the battery parameters.

4. A high-power DC charging pile overcurrent prediction method according to claim 3, characterized in that: The neural network model includes a long-term feature sub-network, a short-term feature sub-network, a temporal feature sub-network and a fusion sub-network.

5. A high-power DC charging pile overcurrent prediction method according to claim 4, characterized in that: The long-term feature refers to a feature whose continuous collection time exceeds a predetermined collection time threshold, and the short-term feature refers to a feature whose continuous collection time is less than a predetermined collection time threshold.

6. A high-power DC charging pile overcurrent prediction method according to claim 5, characterized in that: The predetermined collection time threshold is 10 minutes to 30 minutes.

7. A high-power DC charging pile overcurrent prediction method according to claim 6, characterized in that: The predetermined collection time threshold is 10 minutes.

8. A high-power DC charging pile overcurrent prediction method according to claim 7, characterized in that: The time window is set to 30 seconds to 1 minute.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the high-power DC charging pile overcurrent prediction method according to any one of claims 1 to 8.

10. A high-power DC charging pile overcurrent prediction system, characterized in that: A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Images