New energy power supply multi-working-condition modeling method based on identification algorithm and automatic process

Abstract

The invention provides a new energy power supply multi-working-condition modeling method based on an identification algorithm and an automatic process. The method comprises the steps that the power level of an actual measurement experiment of a new energy power supply to be identified is determined; based on the power level, an actual measurement experiment type of the new energy power supply is selected, and corresponding experiment operation parameters are automatically configured; operating an actual measurement experiment of the new energy power supply based on the actual measurement experiment type, and recording actual measurement data; on the basis of the actually measured data, performing new energy power supply control parameter identification by adopting multiple algorithms to obtain respective identification results; based on each identification result, configuring operation parameters which are the same as the actual measurement experiment in the new energy power supply simulation environment, operating the new energy power supply simulation experiment, and recording simulation data; deviation between the actually measured data and the simulation data is determined, the algorithm corresponding to the identification result with the minimum deviation and the identification parameters of the algorithm are selected, actually measured modeling of the new energy power supply is completed, and the accuracy and efficiency of modeling of the new energy power supply are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of new energy power supply modeling, and particularly relates to a new energy power supply multi-working condition modeling method based on an identification algorithm and an automatic process. BACKGROUND

[0002] With the rapid development of new energy power generation technology, the penetration rate of renewable energy such as photovoltaic and wind power in the power system is continuously increasing. However, the strong nonlinear characteristics, multi-working condition operation characteristics and complex control strategies of new energy power supply make its accurate modeling face great challenges.

[0003] Traditional modeling methods mainly rely on a single algorithm and experimental data under limited working conditions, and it is difficult to comprehensively cover various working conditions that new energy inverters may encounter in actual operation. In addition, there are certain limitations in the parameter identification of new energy power supply in the prior art: first, a large number of iterative calculations are required in the parameter identification process, and the algorithm converges slowly; second, the experimental design process of the existing method relies too much on the experience of engineers, and it is difficult to automatically generate a systematic and standardized full-working condition test sequence; in addition, the entire process from experimental process generation to parameter optimization lacks intelligent collaborative mechanism, and the deviation correction between simulation results and measured data still needs to rely on manual experience, which is low in efficiency and cannot meet the actual production needs. SUMMARY

[0004] The present application provides a new energy power supply multi-working condition modeling method based on an identification algorithm and an automatic process, to solve the technical problem of low modeling efficiency of new energy power supply in the prior art.

[0005] In one aspect, the present application provides a new energy power supply multi-working condition modeling method based on an identification algorithm and an automatic process, comprising: determining the power level of the measured experiment of the new energy power supply to be identified; based on the power level, selecting the measured experiment type of the new energy power supply and automatically configuring the corresponding experimental operation parameters; based on the measured experiment type, running the measured experiment of the new energy power supply and recording the measured data; based on the measured data, using multiple algorithms to respectively identify the control parameters of the new energy power supply, and obtaining respective identification results; based on each of the identification results, configuring the same operation parameters in the new energy power supply simulation environment as the measured experiment, running the new energy power supply simulation experiment, and recording the simulation data; determining the deviation between the measured data and the simulation data, selecting the algorithm and its identification parameters corresponding to the smallest deviation, and completing the measured modeling of the new energy power supply.

[0006] According to the new energy power supply multi-working condition modeling method based on the identification algorithm and the automatic process provided by the application, based on the power level, the measured experiment type of the new energy power supply is selected, including: If the power level is greater than the preset power threshold, a semi-physical measured experiment is selected. If the power level is less than or equal to the preset power threshold, a full-physical measured experiment is selected.

[0007] According to the new energy power supply multi-working condition modeling method based on the identification algorithm and the automatic process provided by the application, the deviation between the measured data and the simulation data is determined, the algorithm and the identification parameter corresponding to the identification result with the minimum deviation are selected, and the measured modeling of the new energy power supply is completed, including: The measured data and the simulation data are compared, and the deviation of each parameter between the two is calculated. Based on the deviation of each parameter, the total deviation is calculated by using the weighted average method. The total deviation corresponding to different identification algorithms is compared, and the identification algorithm with the minimum total deviation and the corresponding identification parameter are selected. The selected identification algorithm and the identification parameter are taken as the final modeling result, and the measured modeling of the new energy power supply is completed.

[0008] According to the new energy power supply multi-working condition modeling method based on the identification algorithm and the automatic process provided by the application, the deviation of each parameter between the measured data and the simulation data is calculated, including: When the running state is a normal working condition state, the running process of the new energy power supply is divided into a steady state interval and a transient state interval. The average deviation and the maximum deviation of the steady state interval are calculated. The average deviation of the transient state interval is calculated. When the running state is a fault working condition, the running process is divided into a pre-disturbance stage, a disturbance period stage and a post-disturbance stage. For the pre-disturbance stage, the average deviation is calculated according to the steady state. For the disturbance period stage, the average deviation is calculated according to the transient state. For the post-disturbance stage, the average deviation is calculated according to the steady state. Among them, the weight of the disturbance period stage is greater than the weight of the post-disturbance stage, and the weight of the pre-disturbance stage is greater than the weight of the post-disturbance stage.

[0009] According to the new energy power supply multi-working condition modeling method based on the identification algorithm and the automatic process provided by the application, the corresponding experimental running parameters are automatically configured, including: The voltage level and the fault ride-through type suitable for the power level are automatically determined. According to the power level, the voltage level and the fault ride-through type, an experimental test sequence is generated. For each test point in the experimental test sequence, automatically adjust the parameter settings of the experimental equipment.

[0010] According to the new energy power supply multi-working condition modeling method based on identification algorithm and automatic process provided by the application, the deviation between the measured data and the simulation data further comprises: For the deviation of each parameter, confidence interval analysis is introduced to obtain confidence interval analysis results; According to the confidence interval analysis results, the deviation is classified to obtain accidental deviation and systematic deviation; For the parameters with a systematic deviation greater than a preset deviation threshold, the causes are analyzed and improvement measures are proposed.

[0011] According to the new energy power supply multi-working condition modeling method based on identification algorithm and automatic process provided by the application, according to the confidence interval analysis results, the deviation is classified to obtain accidental deviation and systematic deviation, which comprises: If the deviation value continuously falls outside the confidence interval, it is determined as a systematic deviation; If the deviation value is randomly distributed inside and outside the confidence interval, it is determined as accidental deviation.

[0012] According to the new energy power supply multi-working condition modeling method based on identification algorithm and automatic process provided by the application, before determining the deviation between the measured data and the simulation data, further comprising: The measured data and the simulation data are cut into multiple data segments according to a time window; The similarity of the data segments in each window is calculated one by one; Data segments with a similarity lower than a matching threshold are marked as abnormal segments, and the experimental operation parameters corresponding to the abnormal segments are traced back; Based on the parameter distribution characteristics of the abnormal segments, an experimental retest suggestion list is automatically generated.

[0013] According to the new energy power supply multi-working condition modeling method based on identification algorithm and automatic process provided by the application, the plurality of algorithms comprises at least two of the following algorithms: Particle swarm optimization algorithm, least square method and BP neural network.

[0014] According to the new energy power supply multi-working condition modeling method based on identification algorithm and automatic process provided by the application, the particle swarm optimization algorithm is used for parameter identification, which comprises: Step 1, design the objective function to minimize the error between the model output and the measured data; Step 2, set the algorithm parameters, specifically including the number of particles corresponding to each inverter control parameter, the maximum number of iterations, the inertia weight, the learning factor and the speed range; Step three, initialize the particle swarm, including the position and velocity of particles and other information; Step four, for each particle, calculate its fitness value, that is, the value of the objective function; Step five, update the global optimal position and the historical optimal position of each particle; Step six, update the velocity and position of each particle according to the global optimal position and the historical optimal position; Step seven, check the termination condition, if the maximum number of iterations is reached or the difference between the objective function values of two consecutive iterations is less than the precision threshold, output the optimal solution, otherwise return to step four; The identification of BP neural network includes: Step one, input sample data, pass it to the first layer of the network, that is, the input layer, and each input variable corresponds to an input neuron; Step two, pass the input data through the weighted sum and nonlinear transformation of the input layer to the next layer of the network; Step three, repeat step two until the data is passed to the last layer of the network, that is, the output layer, and the output result of the network is calculated; Step four, compare the output result with the actual label and calculate the error; Step five, according to the error value and chain rule, calculate the gradient of each neuron from the output layer to the input layer layer by layer; Step six, according to the calculated gradient information, adjust the weight and bias of each neuron to reduce the error value; Parameter identification using least squares method includes: Based on n experiments, the multiple linear regression model containing m+1 parameters to be solved can be expressed as follows: ; Where, is the response of the model under the b th experiment, is the b th independent variable at the d th experiment, is the constant to be solved, is the corresponding regression coefficient to be solved, is the random error under the b th experiment; The estimated value of the parameter that minimizes the sum of squares of random errors is The least squares estimation result corresponding to the to-be-solved parameters of the model is the estimated value vector , which can be expressed as follows: ; Where,X is n ∙ m a matrix of arguments of order n, Y is n a response vector of order n.

[0015] The application provides a new energy power supply multi-working condition modeling method based on an identification algorithm and an automatic process, and automatic configuration of multi-working condition parameters is realized through automatic process generation; a plurality of algorithms are used for identification of new energy power supply control parameters respectively, identification results of the respective algorithms are obtained, and intelligent identification of new energy power supply control parameters is realized; a measured-simulation data verification architecture is established, a deviation calculation result is used as an evaluation standard for algorithm selection, an identification algorithm and an identification result with minimum output deviation are selected, and thus new energy power supply measured modeling is completed, technical problems such as incomplete working condition coverage, low efficiency and dependence on manual intervention in a traditional method are effectively solved, and the accuracy and efficiency of new energy power supply modeling are improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0017] Figure 1 is a flowchart of the new energy power supply multi-working condition modeling method based on the identification algorithm and the automatic process provided by the embodiment of the application; Figure 2 is a structural schematic diagram of the new energy power supply multi-working condition modeling device based on the identification algorithm and the automatic process provided by the embodiment of the application; Figure 3 is a schematic diagram of particle position updating provided by the embodiment of the application; Figure 4 is a structural schematic diagram of an electronic device provided by the embodiment of the application. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the application clearer, the technical solutions in the application will be described clearly and completely below with reference to the drawings in the application. Obviously, the described embodiments are some embodiments of the application, but not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0019] Figure 1is a flowchart of a new energy power supply multi-working condition modeling method based on identification algorithm and automatic process provided by an embodiment of the application.

[0020] Referring to Figure 1 The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process can include the following steps.

[0021] Step 101, determining the power level of the measured experiment of the new energy power supply to be identified.

[0022] In this step, before modeling the new energy power supply, the model and power level of the new energy inverter to be tested need to be determined. The power level is an important basis for experimental design, as it determines the type and parameter configuration of subsequent experiments. For example, the power level can be divided into three levels: large, medium and small. The specific values can be determined according to the actual application scenario and equipment specifications.

[0023] Step 102, based on the power level, selecting the measured experiment type of the new energy power supply and automatically configuring the corresponding experiment running parameters.

[0024] In this step, according to the determined power level, the appropriate measured experiment type is selected. The experiment type can include semi-physical measured experiment and full-physical measured experiment. For example, for a power level greater than 50kW, a semi-physical measured experiment is used; for a power level less than 50kW, a full-physical measured experiment is used. Experiment running parameters include voltage level, fault ride-through type, fault ride-through depth, fault duration, etc. These parameters need to be configured according to industry specifications. The fault ride-through type includes low penetration symmetric fault, low penetration asymmetric fault, and high penetration symmetric fault. The fault ride-through depth includes 5 low penetration conditions and 2 high penetration conditions, i.e. low penetration 5%, 20%, 40%, 60%, 80%, and high penetration 115%, 130%.

[0025] Step 103, based on the measured experiment type, running the measured experiment of the new energy power supply, and recording the measured data.

[0026] In this step, according to the selected experiment type and configured running parameters, the actual experiment is run. During the experiment, various key data such as voltage, current, power, etc. need to be recorded. These data will serve as the basis for subsequent parameter identification.

[0027] Step 104, based on the measured data, using multiple algorithms to identify the control parameters of the new energy power supply respectively, and obtaining respective identification results.

[0028] In this step, the measured data is processed using multiple algorithms (such as at least two of particle swarm optimization algorithm, least squares method, BP neural network algorithm, etc.) to identify the control parameters of the new energy power source. Each algorithm has its unique advantages and applicable scenarios, and through the comparison of multiple algorithms, the accuracy and reliability of parameter identification can be improved.

[0029] Step 105, based on the identification results, configure the same operating parameters as the measured experiment in the new energy power source simulation environment, run the new energy power source simulation experiment, and record the simulation data.

[0030] In this step, the new energy power source simulation environment is a virtual, computer-based simulation test platform that simulates the operation of new energy power sources (such as photovoltaic inverters, wind turbines, etc.) under various operating conditions. It reproduces the actual operating characteristics of new energy power sources through software tools and mathematical models, including their electrical performance, control strategies, and interactions with the power grid.

[0031] The main functions of the simulation environment can include: parameter configuration function: can accurately configure according to the operating parameters of the measured experiment (such as power level, voltage level, fault ride-through type, etc.), to ensure that the simulation environment is consistent with the measured experimental conditions. Model verification function: run the simulation experiment by inputting the control parameters identified, generate simulation data, and compare with measured data for analysis to verify the accuracy of the model. Dynamic response simulation function: can simulate the dynamic response of new energy power sources under different operating conditions, including steady-state operation, transient changes (such as fault ride-through process) and recovery process after fault. Data recording function: can record key data such as voltage, current, power, etc. during the simulation process for subsequent analysis and deviation calculation.

[0032] Step 106, determine the deviation between the measured data and the simulation data, select the algorithm and its identification parameters corresponding to the smallest deviation, and complete the measured modeling of the new energy power source.

[0033] In this step, the deviation between the measured data and the simulation data is calculated, including the average deviation and maximum deviation of key indicators such as voltage, current, power, etc.

[0034] In this embodiment, the automatic configuration of multiple working condition parameters such as power level, voltage fluctuation range, fault ride-through type and depth is realized through automatic process generation; the control parameter identification of new energy power supply is respectively carried out by using multiple algorithms to obtain respective identification results, thereby realizing the intelligent identification of the control parameter of new energy power supply; the deviation calculation result is taken as the evaluation standard for algorithm selection by establishing a measured-simulation data verification architecture, the identification algorithm and identification result with the minimum output deviation are selected, thereby completing the measured modeling of new energy power supply, effectively solving the technical problems such as incomplete working condition coverage, low efficiency and dependence on manual intervention of the traditional method, and improving the accuracy and efficiency of new energy power supply modeling.

[0035] In an embodiment of the present specification, based on the power level, the type of measured experiment of new energy power supply is selected, including: Step one, if the power level is greater than the preset power threshold, a semi-physical measured experiment is selected; In this step, when the power level is greater than the preset power threshold (for example, 50kW), a semi-physical measured experiment is selected. The semi-physical measured experiment usually combines actual hardware devices and virtual simulation environment, and is suitable for high-power level testing. For high-power level devices, semi-physical measured experiment can more efficiently test while reducing experimental cost and risk.

[0036] Step two, if the power level is less than or equal to the preset power threshold, a full-physical measured experiment is selected; In this step, when the power level is less than or equal to the preset power threshold, a full-physical measured experiment is selected. The full-physical measured experiment completely uses actual hardware devices for testing, and is suitable for low-power level testing. For low-power level devices, full-physical measured experiment can provide a more realistic test environment to ensure the accuracy and reliability of test results.

[0037] In this embodiment, by setting the power threshold, semi-physical measured experiment or full-physical measured experiment is automatically selected, avoiding the subjectivity and uncertainty of manual selection of experiment type. For high-power level devices, selecting semi-physical measured experiment can reduce experimental cost and time; for low-power level devices, selecting full-physical measured experiment can ensure the accuracy of test results, which can adapt to new energy power supplies of different power levels, improve the universality and flexibility of experiments, and ensure the reliability and practicality of experimental results.

[0038] In an embodiment of the present specification, the deviation between the measured data and the simulation data is determined, the algorithm and its identification parameters corresponding to the identification result with the minimum deviation are selected, and the measured modeling of new energy power supply is completed, including: Step one, compare the measured data and the simulation data, and calculate the deviation of each parameter between them; In this step, the parameters include voltage, current, power, and other key indicators. By calculating the deviation, the accuracy of the model can be quantified, providing a basis for subsequent algorithm selection.

[0039] Step two, based on the deviation of each parameter, the total deviation is calculated by weighted average method; In this step, the deviation of each parameter is calculated by weighted average method, and a comprehensive total deviation value is obtained. The weight can be allocated according to the importance and influence degree of different parameters. The weighted average method can more scientifically reflect the overall deviation of the model, avoiding the influence of the overall evaluation result due to the large deviation of some parameters.

[0040] Step three, compare the total deviation of different recognition algorithms, and select the recognition algorithm with the smallest total deviation and its corresponding recognition parameter; In this step, by comparing the total deviation, the optimal algorithm can be objectively selected to ensure the accuracy and reliability of the model.

[0041] Step four, the selected recognition algorithm and its recognition parameter are taken as the final modeling result, and the real-time modeling of new energy power supply is completed; In this step, the algorithm with the smallest total deviation and its recognition parameter are determined as the final modeling result, ensuring the accuracy and reliability of the final model under various working conditions, and providing support for subsequent application.

[0042] In this embodiment, by comparing the measured data and the simulation data, the deviation of each parameter is calculated, and the total deviation is calculated by weighted average method, which can more scientifically evaluate the accuracy of the model. By comparing the total deviation of different algorithms, the algorithm with the smallest total deviation and its recognition parameter are selected, ensuring the objectivity and reliability of the modeling result. Through scientific deviation evaluation and objective algorithm selection, the final model can accurately reflect the characteristics of new energy power supply under various working conditions, providing reliable model support for system design, operation control and fault diagnosis of new energy power supply.

[0043] In an embodiment of the present application, the deviation of each parameter between the measured data and the simulation data is calculated, including: First case: when the running state is normal working condition, the running process of new energy power supply is divided into steady state interval and transient state interval; Calculate the average deviation and maximum deviation of the steady state interval; Calculate the average deviation of the transient state interval.

[0044] Specifically, under normal operating conditions, the operation process of a new energy power source can be divided into a steady-state interval and a transient-state interval. The steady-state interval refers to a stage where the system runs smoothly and the parameters change little, which can be divided according to experience. The transient-state interval refers to a stage where the parameters change rapidly after the system is disturbed, which can also be divided according to experience. By distinguishing the steady-state and transient-state intervals, the performance of the model under different dynamic conditions can be more accurately evaluated. In the steady-state interval, the average deviation and the maximum deviation between the model output and the measured data are calculated. The average deviation reflects the overall error level of the model under steady-state, while the maximum deviation reflects the maximum error of the model under steady-state. By calculating the average deviation and the maximum deviation, the accuracy of the model under steady-state can be comprehensively evaluated. In the transient-state interval, the average deviation between the model output and the measured data is calculated. The average deviation in the transient-state interval reflects the error level of the model during dynamic changes. By calculating the average deviation in the transient-state interval, the adaptability and accuracy of the model during dynamic changes can be evaluated.

[0045] The second case: when the operating state is a fault condition, the operation process is divided into a pre-disturbance stage, a disturbance period stage, and a post-disturbance stage; For the pre-disturbance stage, the average deviation is calculated according to the steady state; For the disturbance period stage, the average deviation is calculated according to the transient state; For the post-disturbance stage, the average deviation is calculated according to the steady state; Among them, the weight of the disturbance period stage (for example, set to 0.6) is greater than the weight of the post-disturbance stage (for example, set to 0.3) which is greater than the weight of the pre-disturbance stage (for example, set to 0.1).

[0046] Specifically, the pre-disturbance stage refers to the normal operation stage before the fault occurs; the disturbance period stage refers to the dynamic change stage after the fault occurs; and the post-disturbance stage refers to the stage where the system returns to normal operation after the fault ends. By distinguishing different stages, the performance of the model under fault conditions can be more accurately evaluated. In the pre-disturbance stage, the system is in normal operation state, so the average deviation is calculated according to the steady state, ensuring that the accuracy of the model before the fault occurs can be accurately evaluated. In the disturbance period stage, the system is in a dynamic change state, so the average deviation is calculated according to the transient state, evaluating the adaptability and accuracy of the model during dynamic changes. In the post-disturbance stage, the system returns to normal operation state, so the average deviation is calculated according to the steady state, ensuring that the accuracy of the model after the fault ends can be accurately evaluated.

[0047] In this embodiment, by distinguishing between normal and fault conditions, and further subdividing into steady-state interval, transient-state interval, pre-disturbance phase, during-disturbance phase and post-disturbance phase, the performance of the model under different dynamic conditions can be more comprehensively evaluated. Different weights are given to the deviations of different phases, ensuring that the accuracy of the model in key phases (such as the during-disturbance phase) is given priority, improving the overall reliability of the model. Through this detailed deviation calculation method, the model can better adapt to different operating conditions, including normal operation and fault conditions, ensuring accurate prediction and control in all situations.

[0048] In an embodiment of the present specification, the corresponding experimental operation parameters are automatically configured, including: Step one, automatically determine the voltage level and fault ride-through type that match the power level; In this step, before conducting the actual experiment, the voltage level and fault ride-through type that match the power level of the new energy power source need to be determined automatically. Voltage level and fault ride-through type are important parameters for experimental design, which directly affect the accuracy and reliability of the experiment. For example, input the basic information of the new energy power source such as power level, rated voltage, control strategy, etc., and automatically calculate the matching voltage level and fault ride-through type according to the preset rules (such as the mapping relationship between power level and voltage level, the adaptation relationship between fault ride-through type and power level).

[0049] Step two, generate a standardized experimental test sequence according to the power level, voltage level and fault ride-through type; In this step, based on the determined power level, voltage level and fault ride-through type, a standardized experimental test sequence is generated. The experimental test sequence includes a series of specific test points, each of which corresponds to a specific set of experimental parameters such as voltage level, fault ride-through type, fault ride-through depth, fault duration, etc. Through the standardized experimental test sequence, the systematicness and comprehensiveness of the experimental process are ensured, avoiding modeling errors caused by incomplete experimental design.

[0050] Step three, for each test point in the experimental test sequence, automatically adjust the parameter settings of the experimental equipment to ensure the accuracy and repeatability of the experiment; In this step, for example, the parameters of each test point in the generated experimental test sequence are mapped to the control interface of the experimental equipment, including voltage level, fault ride-through type, fault ride-through depth, fault duration, etc. According to the mapping result, the parameter settings of the experimental equipment are automatically adjusted. This is usually achieved through automated control software or hardware interfaces, for example, through PLC (Programmable Logic Controller) or SCADA (Supervisory Control and Data Acquisition) system to automatically adjust voltage, current, power and other parameters.

[0051] In this embodiment, by automatically determining the voltage level and fault ride-through type adapted to the power level, manual intervention is reduced, and the efficiency and accuracy of experimental design are improved. The standardized experimental test sequence is generated to ensure the systematicness and comprehensiveness of the experimental process, and to avoid modeling errors caused by incomplete experimental design. By automatically adjusting the parameter settings of the experimental equipment, the experimental conditions of each test point are ensured to be consistent, and the accuracy and repeatability of the experimental results are improved.

[0052] In an embodiment of the present specification, determining the deviation between the measured data and the simulation data further comprises: Step one, for the deviation of each parameter, confidence interval analysis is introduced to obtain the confidence interval analysis result; In this step, the confidence interval analysis refers to determining the fluctuation range of the deviation value by statistical methods. Specifically, the confidence interval of each parameter deviation can be calculated by normal distribution, which is used to judge whether the deviation has statistical significance.

[0053] Step two, according to the confidence interval analysis result, the deviation is classified to obtain accidental deviation and systematic deviation; In this step, accidental deviation refers to the deviation caused by random factors, which is specifically manifested as the random distribution of the deviation value within and outside the confidence interval, and is usually caused by measurement noise or transient interference. Random here refers to the fact that the deviation value does not have obvious rules at multiple time points or data segments, and is randomly distributed within and outside the confidence interval. Systematic deviation refers to the persistent deviation caused by model structure defects or parameter setting errors, which is specifically manifested as the deviation value continuously exceeding the confidence interval range, that is, continuously falling outside the confidence interval, and needs to be corrected through parameter sensitivity analysis or model structure optimization. Persistent here refers to the fact that the deviation value always shows certain consistent characteristics at multiple consecutive time points or data segments.

[0054] Step three, for the parameters with systematic deviation greater than the preset deviation threshold, analyze the reasons and propose improvement measures; In this step, the preset deviation threshold refers to the maximum value of the allowed systematic deviation, which can be determined according to the equipment accuracy requirement or industry standard, for example, set to 5% of the nominal value.

[0055] In this embodiment, specifically, in the bias analysis stage, first, the confidence interval calculation is performed on the bias sequence of each parameter in different time windows. If the bias value of a parameter exceeds the upper or lower limit of the confidence interval in continuous multiple time windows, it is determined as a system bias. For the parameters classified as system bias, further analysis is needed to find out the causes: for example, check whether the physical constraints of the parameter in the simulation model are consistent with the measured equipment, or verify whether there is a systematic error in the measurement of the parameter in the experiment. The improvement measures include adjusting the coefficients in the model parameter equation, optimizing the threshold setting in the control logic, or recalibrating the sensors of the experimental equipment. For accidental bias, it can be suppressed by increasing the number of data sampling or using filtering algorithm.

[0056] The embodiment can accurately identify key system biases that need to be prioritized, avoiding resource waste. The embodiment can effectively distinguish between systematic problems and random fluctuations in measured and simulated biases, optimize model parameters or experimental settings, reduce the number of invalid debugging times, and improve modeling efficiency and accuracy. Through the dynamic adjustment mechanism of the preset bias threshold, the method can adapt to new energy power modeling scenarios with different precision requirements, enhancing the universality of the method.

[0057] In an embodiment of the present specification, before determining the bias between the measured data and the simulation data, further comprising: Step one, the measured data and the simulation data are cut into multiple data segments according to the time window; In this step, time window cutting refers to dividing the continuously collected time series data into data segments with fixed or variable lengths, which can be realized by using the sliding window method, and the window length can be adaptively adjusted according to the dynamic response characteristics of the new energy power source.

[0058] Step two, calculate the similarity of the data segments in each window one by one; In this step, similarity calculation refers to measuring the matching degree of two data segments in shape, amplitude and time sequence characteristics, which can be realized by using dynamic time warping algorithm or cosine similarity method, and the abnormal data that significantly deviates from the matching threshold is identified.

[0059] Step three, mark the data segments with similarity lower than the matching threshold as abnormal segments, and backtrack the experimental operation parameters corresponding to the abnormal segments; In this step, abnormal segment marking refers to comparing the similarity calculation result with the preset threshold to automatically select the abnormal interval that needs to be focused on, which can be realized by setting a dynamic threshold mechanism, and the threshold can be dynamically adjusted according to the data quality.

[0060] Step four, based on the parameter distribution characteristics of the abnormal segments, automatically generate an experimental retest suggestion list, and preferentially retest the working condition points with the highest parameter combination frequency.

[0061] In this step, the parameter distribution feature analysis refers to statistical analysis and pattern recognition of the experimental parameters corresponding to the abnormal section. Specifically, a clustering algorithm can be used to identify high-frequency parameter combination regions, and the parameter space density distribution can be used to determine the priority of retesting. The retest suggestion list generation refers to automatically generating an optimized test plan based on the analysis results. Specifically, the sampling strategy can be dynamically adjusted based on historical data learning feedback mechanism to preferentially cover parameter sensitive regions.

[0062] In this embodiment, for example, the measured and simulated data streams are first divided into multiple data sections according to a preset time resolution, such as using a sliding window division method with an overlap rate of 50%. For each data section in the window, the morphological similarity score is calculated by the dynamic time warping algorithm, and when the score is below a preset threshold, it is automatically marked as an abnormal section. The experimental parameter combination corresponding to all abnormal sections is automatically extracted, and a density clustering algorithm is used to identify the abnormal aggregation region in the parameter space. Further, in combination with the physical constraint boundaries of the new energy power source, such as the safe working area of the power device and the control logic switching threshold, it is determined whether the abnormal region is in a sensitive interval with strong nonlinear response. For the high-sensitive regions identified, retest instructions containing multi-point dense sampling and variable resolution measurement are automatically generated, such as millisecond-level sampling in the voltage mutation interval and second-level sampling in the steady state interval. By recording the correlation between historical retest results and modeling accuracy, the abnormality judgment threshold and clustering weight parameters are dynamically optimized to form an adaptive guidance mechanism for parameter space exploration.

[0063] This embodiment realizes accurate positioning of abnormal intervals by introducing time window segmentation and dynamic similarity evaluation; establishes an intelligent retest strategy generation mechanism by combining parameter space clustering analysis and physical constraint conditions; continuously optimizes the abnormality detection and parameter exploration path using adaptive learning from historical data feedback, significantly improving the modeling efficiency. This embodiment effectively solves the technical problems of inaccurate abnormal data detection and lack of scientific basis for retest schemes in traditional modeling methods. Through time window segmentation and dynamic similarity evaluation, the identification accuracy of abnormal intervals is improved; based on parameter space clustering and physical constraint analysis, the optimization of retest resources is realized; with the help of adaptive learning mechanism, the ability of continuously improving the modeling path optimization is formed, and finally the accuracy and efficiency of multi-working-condition modeling of new energy power source are improved.

[0064] In an embodiment of the present specification, based on the parameter distribution characteristics of the abnormal section, an experimental retest suggestion list is automatically generated, including: Step one, cluster analysis is performed on the experimental operating parameters corresponding to all data marked as abnormal sections, and the key parameter combination region causing simulation and measurement deviation is identified; In this step, clustering analysis refers to the process of grouping similar feature abnormal data segments through unsupervised learning methods, which can be implemented using K-means algorithm or hierarchical clustering algorithm, to discover the deviation set in the parameter space. The key parameter combination region refers to the working condition interval where multiple experimental parameters jointly cause significant deviation between simulation and measured data. Through multi-dimensional parameter space projection identification, the parameter range that needs to be verified can be located.

[0065] Step two, combined with the physical constraint boundary of new energy power supply and the control logic sensitive area, judge whether the parameter combination region is in the working condition interval with strong nonlinear response or frequent control switching; In this step, the physical constraint boundary refers to the limitation conditions that new energy power supply must comply with in actual operation due to physical characteristics. These boundaries are usually determined by the physical characteristics, safety standards and operation specifications of the equipment to ensure that the equipment operates within a safe and effective range. For example, the voltage and current of the equipment cannot exceed its rated value or safety range; the output power of the equipment cannot exceed its rated power. The control logic sensitive area refers to the area where the change of certain parameters will cause the switching or significant change of control logic during the operation of new energy power supply. These areas are usually closely related to the control strategy of the equipment and may involve nonlinear control logic or complex feedback mechanism. For example, the change of certain parameters will cause the equipment to exhibit nonlinear characteristics; the change of certain parameters will cause the switching of control logic, such as from one operating mode to another.

[0066] The working condition interval with strong nonlinear response or frequent control switching refers to the area where new energy power supply exhibits strong nonlinear dynamics or frequent switching control mode under certain operating conditions. It can be determined by pre-defined control logic state machine and dynamic response threshold. Frequent switching can be understood as the switching frequency exceeding the pre-set switching frequency threshold.

[0067] Step three, if yes, it is determined as a high sensitive area, then add multi-point sampling instructions for the area in the retest suggestion list, and suggest adjusting the sampling time resolution to capture fast dynamic processes; In this step, multi-point sampling instructions refer to setting sampling points higher than the regular test density in the key parameter region, which can be generated using Latin hypercube sampling or adaptive mesh refinement method. Sampling time resolution adjustment refers to adjusting the data collection frequency according to the time scale characteristics of dynamic processes, such as shortening the sampling period from milliseconds to microseconds.

[0068] Step four, according to the learning feedback of historical retest results, dynamically update the abnormal judgment threshold and clustering weight configuration, form an adaptive guidance mechanism for parameter space exploration, and gradually approach the optimal modeling coverage path.

[0069] In this step, the learning feedback of historical retest results refers to inputting the deviation data after each retest into the machine learning model. Specifically, an online learning algorithm can be used to update the clustering weight and optimize the parameter space exploration path.

[0070] In this embodiment, specifically, when the similarity analysis of the measured data and the simulation data detects an abnormal section, the system first extracts the multi-dimensional data set corresponding to the experimental operation parameters. The abnormal data section is grouped according to the parameter combination characteristics by the clustering algorithm, and the area where the deviation is concentrated in the parameter space is identified. Then, combined with the physical limitation conditions of the new energy power supply and the control strategy characteristics, it is evaluated whether these areas are located in the sensitive interval of frequent control mode switching or nonlinear dynamic response. If it is determined to be a high sensitive area, a retest list containing dense sampling points and time resolution adjustment suggestions is automatically generated. For example, when the abnormal reactive current control parameter is detected in the low voltage ride-through process of the inverter, the system can set the sampling density of three times of the regular test points in the interval of 80% to 60% voltage drop, and adjust the sampling interval from 10ms to 2ms. At the same time, according to the convergence of the parameter identification result in the early retest, the weight coefficient of the clustering analysis is dynamically adjusted, for example, the weight proportion of the voltage fluctuation parameter is increased, to guide the subsequent experiment to preferentially cover the key parameter combination.

[0071] Through clustering analysis and dynamic weight adjustment, this embodiment can automatically identify parameter sensitive areas and continuously optimize the test strategy based on historical data, significantly improving the retest efficiency and modeling accuracy. This embodiment realizes intelligent analysis of abnormal data sections and automatic generation of retest strategies, effectively solving the problem of low efficiency and insufficient coverage of manual selection of retest points. This method ensures that key working condition combinations are verified preferentially through an adaptive exploration mechanism of the parameter space, and captures fast transient processes using a dynamic sampling strategy, thereby improving the accuracy and robustness of the new energy power supply model.

[0072] In an embodiment of the present specification, the plurality of algorithms can include a particle swarm optimization algorithm, a least squares method, and a BP neural network, etc.

[0073] In this embodiment, the particle swarm optimization algorithm is used for parameter identification, and the identification steps are as follows: Step one, design the objective function to minimize the error between the model output and the measured data; Step two, set the algorithm parameters, specifically including the number of particles corresponding to each inverter control parameter, the maximum number of iterations, the inertia weight, the learning factor, and the speed range; Step three, initialize the particle swarm, including the position and speed information of the particles, etc. Step four, for each particle, calculate its fitness value, i.e. the value of the objective function; specifically as shown in the following formula (1) and formula (2): (1); (2); Formula (1) represents particle velocity update, and formula (2) represents particle position update. Indicates the number of particle iterations; middle Indicates the particle's serial number. Represents a certain dimension of the current particle; , These are the particle's updated velocity and the particle's current velocity, respectively. , These represent the particle's updated position and its current position, respectively. , These represent individual extreme values ​​and global extreme values, respectively. , Indicates the learning factor; , This represents a random number between [0,1]. For particle update positions, please refer to... Figure 3 .

[0074] Step 5: Update the global optimal position and the historical optimal position of each particle; Step 6: Update the velocity and position of each particle based on the global best position and the historical best position; Step 7: Check the termination condition. If the maximum number of iterations is reached or the difference between the objective function values ​​of two consecutive iterations is less than the accuracy threshold, output the optimal solution; otherwise, return to step 4.

[0075] Parameter identification is performed using the traditional least squares method, based on n This experiment contains m+ A multiple linear regression model with one parameter to be solved can be represented by the following formula (3): (3); in, For the first b The model's response in this experiment. For the first b The first experiment d One independent variable, Let be the constant to be solved. for The corresponding regression coefficients to be solved, For the first b Random error under this experiment; The parameters that minimize the sum of squared random errors The estimated value This refers to the least squares estimation results corresponding to the parameters to be solved in the model, the estimated value vector. is shown in the following equation (4): (4). In equation (4), X is n · m a matrix of order Y is n a response vector of order

[0076] Take low penetration as an example: Under the grid voltage oriented vector control mode, considering various factors, the reference value of reactive current during fault The calculation formula is shown in the following equation (5): (5). In equation (5), K qU is the reactive current support coefficient; U is the per unit value of grid voltage amplitude; I q0 is the per unit value of pre-fault reactive current; K qI is the reactive current proportional coefficient; is the low voltage ride through reactive current basic value; I qmax is the maximum value of reactive current.

[0077] For the case where the reference value of reactive current does not reach the maximum limit, the multiple linear regression model without random error vector is shown in the following equation (6): (6). In equation (6), the last half of the subscript 1, 2, …, n represent different experimental conditions.

[0078] The calculation of active current reference value during fault can be roughly divided into the following cases according to different control schemes, as shown in equation (7): (7). In equation (7), P 0 is the per unit value of active power output by the inverter before fault; I max is the maximum value of inverter current; I d0 is the per unit value of pre-fault active current; K dI is the active current proportional coefficient; K dU is the active current support coefficient; is the low voltage ride through active current basic value.

[0079] For the third equation in the active current reference value calculation formula, the multiple linear regression model without random error vectors is shown in equation (8): (8); Substituting the processed experimental data into the above formula can obtain the respective corresponding multiple linear model without error vectors, and further solve the parameter vector in the reactive and active current reference value calculation formula and .

[0080] The BP neural network is used for parameter identification. The training of the BP neural network is alternately performed through forward propagation and back propagation. The BP neural network generally includes an input layer, a hidden layer and an output layer. The input layer is composed of multiple nodes, and each node represents an input variable. The hidden layer is located between the input layer and the output layer and can have multiple hidden layers. Each hidden layer is composed of multiple nodes, and the nodes are connected through weights. The output layer is composed of one or more nodes, and each node represents an output variable. The identification steps of the BP neural network are as follows: Step one, input sample data and pass it to the first layer of the network, i.e. the input layer. Each input variable corresponds to an input neuron. Step two, pass the input data through the weighted sum and nonlinear transformation of the input layer to the next layer of the network. Step three, repeat step two until the data is passed to the last layer of the network, i.e. the output layer, and calculate the output result of the network. Step four, compare the output result with the actual label and calculate the error. Step five, according to the error value and chain rule, calculate the gradient of each neuron from the output layer to the input layer, i.e. the partial derivative of the error with respect to the weight and bias. Step six, according to the gradient information calculated, adjust the weight and bias of each neuron to reduce the error value and improve the accuracy of parameter identification.

[0081] In an embodiment of the present specification, the average deviation of the steady-state interval F 1 is calculated as shown in equation (9): (9); In the formula, X S is the model simulation data unit value of the electrical quantity to be evaluated, X M is the unit value of the experimental data of the electrical quantity to be evaluated, K S_Start , K S_EndThe first and last serial numbers of the model simulation data within the error interval are calculated, K M_Start , K M_End The first and last serial numbers of the experimental data within the error interval are calculated. The first data point, that is, the serial number of the data point.

[0082] The average deviation of the transient interval F 2 can be calculated according to the above formula (9), which will not be repeated here.

[0083] The maximum deviation of the steady-state interval F 3 is shown in the following formula (10): (10); In some other embodiments of the present specification, after using multiple algorithms to identify the control parameters of the new energy power supply based on the measured data, and before selecting the identification result with the smallest deviation, the following steps are further included: Step 1: For the execution process of each identification algorithm, extract its convergence trajectory, parameter sensitivity change trend and residual evolution law to generate the corresponding identification process quality index; In this step, the convergence trajectory refers to the path of the parameter estimation value changing with time in the algorithm iteration process, which can be realized by recording the parameter value of each iteration and drawing the change curve, and is used to reflect the algorithm convergence stability. The parameter sensitivity change trend refers to the change law of the influence degree of the control parameter on the objective function, which can be realized by calculating the sensitivity coefficient matrix using the parameter perturbation method, and is used to identify the key parameters. The residual evolution law refers to the change characteristics of the error between the model output and the measured data with the iteration number, which can be realized by calculating the change trend of the root mean square value of the residual, and is used to evaluate the convergence quality of the algorithm. The process quality index refers to a quantitative evaluation system that comprehensively evaluates the convergence speed, parameter sensitivity and residual convergence, which can be realized by weighted calculation of multi-dimensional indexes, and is used to comprehensively evaluate the performance of the algorithm.

[0084] Step 2: Based on the process quality index, combined with the signal-to-noise ratio and dynamic excitation strength of the measured data under the corresponding working condition, evaluate the local credibility of each identification result in different parameter dimensions; In this step, the local credibility refers to the credibility of the identification result in a specific parameter dimension, which can be realized by confidence interval analysis and error propagation calculation, and is used to guide parameter fusion.

[0085] Step 3: For the key control parameters, if the identification results of multiple algorithms are significantly discrete, then according to the local credibility, the results are weighted and fused to generate a comprehensive identification parameter set. In this step, weighted fusion refers to the integration of parameters by assigning weights based on the credibility of different algorithms. Specifically, the credibility index can be normalized and used as the weighting coefficient to improve the overall reliability of the parameter set.

[0086] Step 4: Incorporate the comprehensive identification parameter set and the results of each individual algorithm into the subsequent simulation verification stage. If the simulation deviation is better than the results of all individual algorithms, then use it as the final modeling result; otherwise, select the result of the individual algorithm with the smallest deviation.

[0087] In this embodiment, by introducing quality indicators for the parameter identification process and comprehensively evaluating and optimizing the identification results, the accuracy and reliability of multi-condition modeling of new energy power sources are further improved. Specifically, it not only evaluates the execution process of each algorithm and extracts quality indicators such as convergence trajectory, parameter sensitivity change trend, and residual evolution law, but also quantitatively evaluates the local credibility of the identification results by combining the signal-to-noise ratio and dynamic excitation intensity of measured data. For key control parameters, when the identification results of multiple algorithms are significantly discrete, this method weights and fuses the results according to local credibility to generate a comprehensive identification parameter set, which is then incorporated into the simulation verification stage. If the simulation deviation of the comprehensive identification parameter set is better than that of all individual algorithm results, it is used as the final modeling result; otherwise, the single algorithm result with the smallest deviation is selected. This process not only improves the flexibility and adaptability of modeling but also ensures the accuracy and reliability of the final model under different operating conditions through a scientific evaluation and optimization mechanism, providing more solid theoretical support and practical guidance for the system design, operation control, and fault diagnosis of new energy power sources.

[0088] Based on the same general inventive concept, this invention also protects a multi-condition modeling device for new energy power sources based on identification algorithms and automated processes, such as... Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of the new energy power source multi-condition modeling device based on identification algorithms and automated processes provided in an embodiment of the present invention. The following describes the new energy power source multi-condition modeling device based on identification algorithms and automated processes provided by the present invention. The new energy power source multi-condition modeling device based on identification algorithms and automated processes described below can be referred to in correspondence with the new energy power source multi-condition modeling method based on identification algorithms and automated processes described above.

[0089] The new energy power source multi-condition modeling device based on identification algorithms and automated processes includes a power determination module 201, a parameter configuration module 202, a measured data module 203, a measured identification module 204, a simulation data module 205, and a measured modeling module 206.

[0090] The power determination module 201 is used to determine the power level of the measured experiment of the new energy power source to be identified; The parameter configuration module 202 is configured to select a measured experiment type of the new energy power supply based on the power level, and automatically configure corresponding experiment operation parameters; The measured data module 203 is configured to run the measured experiment of the new energy power supply based on the measured experiment type, and record measured data; The measured identification module 204 is configured to use multiple algorithms to respectively identify the control parameters of the new energy power supply based on the measured data, and obtain respective identification results; The simulation data module 205 is configured to configure the same operation parameters as the measured experiment in a new energy power supply simulation environment based on the identification results, run a new energy power supply simulation experiment, and record simulation data; The measured modeling module 206 is configured to determine the deviation between the measured data and the simulation data, select the algorithm and its identification parameters corresponding to the identification result with the minimum deviation, and complete the measured modeling of the new energy power supply.

[0091] Figure 4 is a structural schematic diagram of an electronic device provided by an embodiment of the application.

[0092] As shown in Figure 4 , the electronic device can include a processor 410, a communications interface 420, a memory 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 can communicate with each other through the communications bus 440. The processor 410 can invoke logical instructions in the memory 430 to execute a new energy power supply multi-working condition modeling method based on an identification algorithm and an automatic process.

[0093] In addition, the logical instructions in the memory 430 described above can be implemented in the form of a software function unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the embodiments of the application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0094] In another aspect, the present application also provides a computer program product comprising a computer program, the computer program being stored in a non-transitory computer-readable storage medium, and the computer program being executable by a processor to enable a computer to perform the new energy power supply multi-working condition modeling method based on the identification algorithm and the automatic process provided by the above methods.

[0095] In yet another aspect, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, the computer program being executable by a processor to implement the new energy power supply multi-working condition modeling method based on the identification algorithm and the automatic process provided by the above methods.

[0096] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment scheme according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0097] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary universal hardware platforms, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0098] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A new energy power supply multi-working condition modeling method based on identification algorithm and automatic process, characterized in that, The method comprises the following steps: determining the power level of the new energy power supply to be identified; based on the power level, selecting the type of real-time experiment of the new energy power supply and automatically configuring the corresponding experimental operation parameters; based on the type of real-time experiment, running the real-time experiment of the new energy power supply and recording the real-time data; based on the real-time data, using multiple algorithms to identify the control parameters of the new energy power supply respectively to obtain respective identification results; based on the identification results, configuring the same operation parameters in the new energy power supply simulation environment as the real-time experiment, running the new energy power supply simulation experiment, and recording the simulation data; determining the deviation between the real-time data and the simulation data, selecting the algorithm and its identification parameters corresponding to the identification result with the smallest deviation, and completing the real-time modeling of the new energy power supply.

2. The new energy power supply multi-working condition modeling method based on recognition algorithm and automatic process according to claim 1, characterized in that, Based on the power level, the type of real-time experiment of the new energy power supply is selected, which comprises: if the power level is greater than a preset power threshold, a semi-physical real-time experiment is selected; if the power level is less than or equal to the preset power threshold, a full-physical real-time experiment is selected.

3. The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process according to claim 1, characterized in that, Determine the deviation between the real-time data and the simulation data, select the algorithm and its identification parameters corresponding to the identification result with the smallest deviation, and complete the real-time modeling of the new energy power supply, which comprises: compare the real-time data and the simulation data to calculate the deviation of each parameter between them; based on the deviation of each parameter, the total deviation is calculated by using the weighted average method; compare the total deviations corresponding to different identification algorithms, and select the identification algorithm with the smallest total deviation and its corresponding identification parameters; select the identification algorithm and its identification parameters as the final modeling result to complete the real-time modeling of the new energy power supply.

4. The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process according to claim 3, characterized in that, Comparing the real-time data and the simulation data, the deviation of each parameter between them is calculated, which comprises: when the running state is normal working condition, the running process of the new energy power supply is divided into steady state interval and transient state interval; calculate the average deviation and the maximum deviation in the steady state interval; calculate the average deviation in the transient state interval; when the running state is fault condition, the running process is divided into pre-disturbance stage, disturbance period stage and post-disturbance stage; for the pre-disturbance stage, calculate the average deviation according to the steady state; for the disturbance period stage, calculate the average deviation according to the transient state; for the post-disturbance stage, calculate the average deviation according to the steady state; wherein, the weight of the disturbance period stage is greater than the weight of the post-disturbance stage, and the weight of the pre-disturbance stage is the smallest.

5. The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process according to claim 1, characterized in that, Automatically configure the corresponding experimental operation parameters, which comprises: automatically determine the voltage level and fault ride-through type that adapt to the power level; generate an experimental test sequence according to the power level, the voltage level and the fault ride-through type; for each test point in the experimental test sequence, automatically adjust the parameter settings of the experimental equipment.

6. The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process according to claim 1, characterized in that, The determination of the deviation between the real-time data and the simulation data further comprises: for the deviation of each parameter, confidence interval analysis is introduced to obtain the confidence interval analysis result; according to the confidence interval analysis result, classify the deviation to obtain accidental deviation and systematic deviation; for the parameters with a systematic deviation greater than a preset deviation threshold, analyze the reasons and propose improvement measures.

7. The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process according to claim 6, characterized in that, According to the confidence interval analysis result, the deviation is classified to obtain accidental deviation and systematic deviation, which comprises: If the deviation value continuously falls outside the confidence interval, it is determined that the system deviation; If the deviation value is randomly distributed inside and outside the confidence interval, it is determined that the accidental deviation. 8.The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process according to claim 1, characterized in that, Before determining the deviation between the measured data and the simulation data, further comprising: The measured data and the simulation data are cut into multiple data segments according to a time window; The similarity of each data segment in each window is calculated one by one; Data segments with a similarity lower than a matching threshold are marked as abnormal segments, and the experimental running parameters corresponding to the abnormal segments are traced back; Based on the parameter distribution characteristics of the abnormal segments, an experimental retest suggestion list is automatically generated. 9.The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process according to claim 1, characterized in that, The plurality of algorithms includes at least two of the following algorithms: Particle swarm optimization algorithm, BP neural network and least squares method.

10. The new energy power supply multi-working condition modeling method based on identification algorithm and automatic process according to claim 9, characterized in that, Parameter identification using particle swarm optimization algorithm includes: Step one, design the objective function to minimize the error between the model output and the measured data; Step two, set the algorithm parameters, including the number of particles corresponding to each inverter control parameter, the maximum number of iterations, the inertia weight, the learning factor, and the speed range; Step three, initialize the particle swarm, including the position and speed information of the particles; Step four, for each particle, calculate its fitness value, i.e. the value of the objective function; Step five, update the global optimal position and the historical optimal position of each particle; Step six, update the speed and position of each particle based on the global optimal position and the historical optimal position; Step seven, check the termination condition. If the maximum number of iterations is reached or the difference between the values of the objective function of two consecutive iterations is less than the precision threshold, output the optimal solution, otherwise return to step four; Parameter identification using BP neural network includes: Step one, input sample data and pass it to the first layer of the network, i.e. the input layer, with each input variable corresponding to an input neuron; Step two, pass the input data through the weighted sum and nonlinear transformation of the input layer to the next layer of the network; Step three, repeat step two until the data is passed to the last layer of the network, i.e. the output layer, and the output result of the network is calculated; Step four, compare the output result with the actual label and calculate the error; Step five, based on the error value and the chain rule, calculate the gradient of each neuron from the output layer to the input layer; Step six, based on the calculated gradient information, adjust the weight and bias of each neuron to reduce the error value; Parameter identification using least squares method includes: Based on n experiments, a multiple linear regression model with m+1 parameters to be solved can be represented by the following formula: ； wherein, is the response of the model for the b th experiment, is the response of the model for the b th experiment, d is the is the constant to be solved, is the is the regression coefficient to be solved, is the random error for the b th experiment; Parameters that minimize a random error sum of squares function An estimate of The least squares estimate of the parameters of the model to be solved is the minimum value, and the estimate vector The following formula: ； wherein, X is n ∙ m a matrix of order n of argument, Y is n a response vector of order n.

Images