Electric energy quality situation deduction model evaluation method, system, equipment and medium

Abstract

The invention belongs to the technical field of power systems, and discloses a power quality situation deduction model evaluation method, system and device and a medium. The method comprises the following steps: acquiring power quality monitoring data, and forming a data set containing reference data and to-be-predicted data; constructing a deep learning situation deduction model based on the obtained data set, and constructing a performance evaluation framework of the deep learning situation deduction model based on a Taylor graph; performing skill score calculation on the performance indexes of the obtained performance evaluation framework to obtain skill score values, sorting the skill score values, and determining a deep learning situation deduction model with the highest score; and based on the deep learning situation deduction model with the highest score, evaluating the influence of the distributed source load power fluctuation on the power quality index of each node in the power distribution network system. According to the invention, accurate reverse traceability of the power quality problem and accurate positioning of the key source load are realized, and scientific decision support is provided for prevention and control and responsibility division of the power quality problem of the power distribution network.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a method, system, equipment and medium for evaluating power quality status projection models. Background Technology

[0002] In recent years, deep learning technology has made significant progress in the field of time series data prediction, providing a new technical approach for power quality status prediction in distribution networks. However, existing research has significant shortcomings in model evaluation methods: most studies only use single indicators such as mean relative error and root mean square error to evaluate model performance. While this evaluation method can reflect the degree of deviation between predicted and actual values, it cannot comprehensively assess the consistency between predicted and actual curves in terms of statistical characteristics such as correlation and standard deviation, and it is difficult to accurately measure the model's ability to capture the trend and discrete characteristics of power quality changes. Especially in application scenarios such as power quality status prediction in distribution networks, which are sensitive to the shape of time series curves, the limitations of traditional indicator evaluation methods are even more prominent.

[0003] Taylor charts have been widely used in the multidimensional evaluation of predictive models, particularly in meteorological models. They provide a comprehensive statistical chart that simultaneously displays the correlation coefficient, standard deviation ratio, and centered root mean square error between model predictions and observed values ​​in a single graph. This allows for the simultaneous characterization of the trend, dispersion, and absolute value differences between the predicted curve and the actual curve. Furthermore, most power quality prediction models employ time-series forecasting, failing to capture the multivariate nonlinear regression relationship between distributed source load power fluctuations and power quality issues. This results in prediction models being limited to power quality exceedance alarms. Moreover, they cannot accurately identify the key source loads causing the problems during the alarm period.

[0004] Existing power quality traceability technologies mostly rely on complete power system structural parameters and detailed measurement data to determine responsibility. The calculation process usually requires vector measurement technology of voltage and current, which is not supported by most electricity meters in the existing power grid. Summary of the Invention

[0005] This invention provides a method, system, device, and medium for evaluating power quality status projection models to solve the aforementioned problems in the prior art.

[0006] According to a first aspect of the present invention, a method for evaluating a power quality situation projection model is provided.

[0007] The power quality situation projection model evaluation method includes:

[0008] Acquire power quality monitoring data of key nodes in the distribution network and form a dataset containing reference data and data to be predicted;

[0009] Based on the obtained dataset, a deep learning situation inference model is constructed, and a performance evaluation framework for the deep learning situation inference model is built based on the Taylor diagram.

[0010] The performance metrics of the obtained performance evaluation framework are used to calculate skill scores, and the skill scores are ranked to determine the deep learning situation inference model with the highest score.

[0011] Based on the deep learning situational inference model with the highest score, the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system is assessed.

[0012] In one embodiment, the power quality monitoring data includes: voltage data, current data, power factor data, and harmonic content data.

[0013] In one embodiment, forming a dataset containing reference data and data to be predicted includes: performing data preprocessing on the obtained power quality monitoring data, and constructing a dataset containing reference data and data to be predicted based on the preprocessed power quality monitoring data; the data preprocessing includes: noise filtering, outlier removal, missing value imputation, time series alignment, data normalization, and standardized data processing.

[0014] In one embodiment, the deep learning situational inference model takes the distribution network topology, load characteristics, and power generation output as input features and power quality indicators as output targets.

[0015] In one embodiment, the performance evaluation framework uses a polar coordinate system to display the statistical relationship between the prediction results of the deep learning situational inference model and the actual monitoring data; the performance indicators of the performance evaluation framework include: correlation coefficient, standard deviation, and root mean square error.

[0016] In one embodiment, the impact assessment of distributed source-load power fluctuations on power quality indicators of each node in the distribution network system based on the highest-scoring deep learning situational inference model includes: assessing the impact of distributed source-load power fluctuations on power quality indicators of each node in the distribution network system from four dimensions: peak value, mean value, volatility, and cumulative effect of disturbance impact, based on the highest-scoring deep learning situational inference model.

[0017] In one embodiment, the power quality situation inference model evaluation method further includes: ranking distributed source loads based on their influence levels according to the influence assessment results, and identifying key source loads that cause power quality problems during periods when power quality exceeds limits.

[0018] According to a second aspect of the present invention, a power quality status prediction model evaluation system is provided.

[0019] The power quality situation prediction model evaluation system includes:

[0020] The data processing module is used to acquire power quality monitoring data of key nodes in the distribution network and form a dataset containing reference data and data to be predicted.

[0021] The model processing module is used to build a deep learning situation inference model based on the obtained dataset, and to build a performance evaluation framework for the deep learning situation inference model based on the Taylor diagram.

[0022] The model determination module is used to calculate skill scores for the performance indicators of the obtained performance evaluation framework, obtain skill score values, sort the skill score values, and determine the deep learning situation inference model with the highest score.

[0023] The model evaluation module is used to assess the impact of distributed source-load power fluctuations on power quality indicators of each node in the distribution network system based on the deep learning situational inference model with the highest score.

[0024] In one embodiment, the power quality monitoring data includes: voltage data, current data, power factor data, and harmonic content data.

[0025] In one embodiment, when forming a dataset containing reference data and data to be predicted, the data processing module performs data preprocessing on the obtained power quality monitoring data, and constructs a dataset containing reference data and data to be predicted based on the preprocessed power quality monitoring data; the data preprocessing includes: noise reduction filtering, outlier removal, missing value imputation, time series alignment, data normalization, and standardized data processing.

[0026] In one embodiment, the deep learning situational inference model takes the distribution network topology, load characteristics, and power generation output as input features and power quality indicators as output targets.

[0027] In one embodiment, the performance evaluation framework uses a polar coordinate system to display the statistical relationship between the prediction results of the deep learning situational inference model and the actual monitoring data; the performance indicators of the performance evaluation framework include: correlation coefficient, standard deviation, and root mean square error.

[0028] In one embodiment, when the model evaluation module assesses the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system based on the deep learning situational inference model with the highest score, it assesses the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system from four dimensions: peak value, mean value, volatility, and cumulative effect of the disturbance.

[0029] In one embodiment, the power quality situation inference model evaluation system further includes: a source load tracing module, used to sort distributed source loads based on the influence level according to the influence assessment results, and identify the key source loads that cause power quality problems during periods of power quality exceeding the limit.

[0030] According to a third aspect of the present invention, a computer device is provided.

[0031] In one embodiment, the computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0032] According to a fourth aspect of the present invention, a computer-readable storage medium is provided.

[0033] In one embodiment, a computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the above method.

[0034] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0035] This invention constructs a multi-dimensional evaluation system for deep learning models based on Taylor diagrams and a source-load disturbance correlation strength identification mechanism. It overcomes the limitations of traditional power quality prediction model evaluation methods that rely on a single indicator and cannot comprehensively reflect prediction performance, as well as the technical bottleneck of existing source tracing technologies that depend on complete system parameters and vector measurement data. Through the innovative application of Taylor diagram multi-dimensional evaluation technology, this method achieves comprehensive performance evaluation of deep learning models across multiple dimensions such as correlation, standard deviation, and root mean square error, establishing a scientific model optimization mechanism. The source-load disturbance correlation strength identification technology constructed through multivariate sensitivity difference analysis can accurately identify key source loads causing power quality problems without requiring complex grid parameters and vector measurements, realizing a functional expansion from prediction and alarm to precise source tracing. This method provides strong technical support for the scientific selection of power quality situation projection models for distribution networks and the allocation of responsibility for power quality problems, significantly improving the accuracy and targeting of power quality management and control in distribution networks.

[0036] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit the invention. Attached Figure Description

[0037] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0038] Figure 1This is a flowchart illustrating a power quality situation deduction model evaluation method according to an exemplary embodiment;

[0039] Figure 2 This is a structural block diagram of a power quality situation inference model evaluation system according to an exemplary embodiment;

[0040] Figure 3 This is a block diagram illustrating the optimization and reverse tracing logic of the power quality status deduction model for a distribution network, based on an exemplary embodiment.

[0041] Figure 4 This is a schematic diagram illustrating model optimization and reverse tracing analysis dimensions according to an exemplary embodiment;

[0042] Figure 5 This is a schematic diagram of the network architecture of four deep learning inference models according to an exemplary embodiment;

[0043] Figure 6 This is a schematic diagram comparing the inferred curves and the actual curves of four inference models according to an exemplary embodiment;

[0044] Figure 7 This is a schematic diagram illustrating an accuracy evaluation based on a Taylor diagram, according to an exemplary embodiment.

[0045] Figure 8 This is a schematic diagram of the time-varying characteristic curve of a key source load over 24 hours, according to an exemplary embodiment.

[0046] Figure 9 This is a schematic diagram illustrating the statistical distribution of critical source loads during the over-limit period, according to an exemplary embodiment.

[0047] Figure 10 This is a schematic diagram of the structure of a computer device according to an exemplary embodiment. Detailed Implementation

[0048] The following description and accompanying drawings fully illustrate specific embodiments described herein to enable those skilled in the art to practice them. Some embodiments may include or substitute parts and features of other embodiments. The scope of the embodiments herein encompasses the entire scope of the claims and all available equivalents thereof. Throughout this document, the terms “first,” “second,” etc., are used only to distinguish one element from another without requiring or implying any actual relationship or order between the elements. Indeed, a first element can also be referred to as a second element, and vice versa. Furthermore, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a structure, apparatus, or device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a structure, apparatus, or device. Without further limitation, an element defined by the phrase “comprising one…” does not exclude the presence of other identical elements in the structure, apparatus, or device that includes said element. The various embodiments described herein are presented in a progressive manner, with each embodiment focusing on its differences from other embodiments; similar or identical parts between embodiments can be referred to interchangeably.

[0049] The terms "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer" used in this document to indicate orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings. They are used solely for the convenience of describing the document and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In the description herein, unless otherwise specified and limited, the terms "installed," "connected," and "linked" should be interpreted broadly. For example, they can refer to mechanical or electrical connections, or internal connections between two elements; they can be direct connections or indirect connections through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms according to the specific circumstances.

[0050] In this document, unless otherwise stated, the term "multiple" means two or more.

[0051] In this article, the character " / " indicates that the objects before and after it are in an "or" relationship. For example, A / B means: A or B.

[0052] In this article, the term "and / or" describes an association between objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or A and B.

[0053] It should be understood that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order constraint on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the diagram may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0054] The modules in the apparatus or system of this application can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0055] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0056] Figure 1 An embodiment of the power quality situation projection model evaluation method of the present invention is shown.

[0057] In this optional embodiment, the power quality situation projection model evaluation method includes:

[0058] Step S101: Obtain power quality monitoring data of key nodes in the distribution network and form a dataset containing reference data and data to be predicted;

[0059] Step S102: Based on the obtained dataset, construct a deep learning situation inference model, and construct a performance evaluation framework for the deep learning situation inference model based on the Taylor diagram.

[0060] Step S103: Calculate skill scores for the performance indicators of the obtained performance evaluation framework, obtain skill score values, sort the skill score values, and determine the deep learning situation inference model with the highest score.

[0061] Step S104: Based on the deep learning situational inference model with the highest score, assess the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system.

[0062] Figure 2 An embodiment of a power quality status prediction model evaluation system according to the present invention is shown.

[0063] In this optional embodiment, the power quality situation projection model evaluation system includes:

[0064] The data processing module 201 is used to acquire power quality monitoring data of key nodes in the distribution network and form a dataset containing reference data and data to be predicted.

[0065] The model processing module 202 is used to construct a deep learning situation inference model based on the obtained dataset, and to construct a performance evaluation framework for the deep learning situation inference model based on the Taylor diagram.

[0066] The model determination module 203 is used to calculate the skill scores of the performance indicators of the obtained performance evaluation framework, obtain skill score values, sort the skill score values, and determine the deep learning situation inference model with the highest score.

[0067] The model evaluation module 204 is used to evaluate the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system based on the deep learning situational inference model with the highest score.

[0068] In practical applications, this invention mainly includes the following steps:

[0069] (1) Data Acquisition and Standardization Preprocessing of Power Quality Monitoring Data in Distribution Networks. Raw power quality data (including voltage, current, power factor, and harmonic content) at key nodes of the distribution network were collected. Preprocessing operations such as noise reduction filtering, outlier removal, and missing value filling were performed on the monitoring data. Through time series alignment, data normalization, and standardization, a standardized multidimensional dataset containing reference data and prediction data from the model to be evaluated was constructed to ensure data quality and consistency, providing a reliable data foundation for subsequent model evaluation and analysis.

[0070] (2) Construction and training of deep learning situation prediction model. Based on preprocessed historical monitoring data, a deep neural network model is constructed with distribution network topology, load characteristics, and power generation output as input features and power quality indicators as output targets. A nonlinear mapping relationship between source load power and power quality indicators is established to achieve accurate prediction of the power quality situation of the distribution network.

[0071] (3) A multidimensional statistical evaluation system for deep learning models based on Taylor charts. A multidimensional evaluation framework for model performance based on Taylor charts is constructed. This framework uses a polar coordinate system to display the statistical relationship between model prediction results and actual monitoring data. In the Taylor chart, the angular axis represents the correlation coefficient between predicted and measured values, and the radial axis represents the standard deviation of predicted values. The polar coordinate position can simultaneously reflect three key statistical indicators: correlation coefficient, standard deviation ratio, and root mean square error. A reference arc for standard deviation, a reference line for correlation coefficient, and a contour line for root mean square error are established to form a complete visualization evaluation space for model performance.

[0072] (4) Design and Implementation of a Comprehensive Skills Scoring Mechanism. A comprehensive skills scoring algorithm is designed to quantitatively evaluate the predictive performance of deep learning models. This algorithm uses a weighted average method to comprehensively calculate the correlation coefficient, standard deviation score, and root mean square error score to obtain a skills score in the range of 0-1. Among them, the correlation coefficient directly reflects the prediction accuracy, the standard deviation score is calculated through the degree of bias, and the root mean square error score is calculated through the normalized root mean square error.

[0073] (5) Key source load tracing technology for power quality problems. Based on the nonlinear mapping relationship between source load power and power quality indicators, the influence of power fluctuations of each distributed source load on the power quality indicators of each node in the system is evaluated from four dimensions: peak value, mean value, volatility, and cumulative effect of disturbance. Distributed source loads are sorted according to their influence level to identify key source loads that cause power quality problems during periods of power quality exceeding limits.

[0074] To facilitate a better understanding of the above-mentioned technical solutions of the present invention, the following specific examples will be used to further illustrate the above-mentioned technical solutions of the present invention.

[0075] For annual historical monitoring data of photovoltaic users in a rural power distribution network, a situational inference model from power information to voltage deviation information is constructed based on four deep learning models. The accuracy of the inference results of the four models is then evaluated based on Taylor diagrams. Finally, the optimal model is selected for reverse tracing. The overall logic is as follows: Figure 3 As shown. In model optimization and reverse tracing techniques, the analytical dimensions considered are as follows: Figure 4 As shown. The specific steps are as follows:

[0076] The source load and power quality data obtained from regional monitoring are cleaned and supplemented to construct training and testing sets. The dimensions B (input information such as power) and C (output information such as power quality) within the region are identified.

[0077] The information transmission mode of this power system is transformed into a regression model with B input and C output. Four deep learning models, FCNN, CNN-LSTM, Transformer and LSTM, are used to construct the network structure, build the deep learning model, and train it until the model converges.

[0078] Export the PTH files after the four models have converged during training, and perform validation on the test set. Calculate the correlation, standard deviation, and root mean square error between the projected and actual curves. The formula for calculating the correlation between the projected and actual data is:

[0079]

[0080] In the formula, R is the Pearson correlation coefficient, and x i Let y be the i-th element of the real data. i For the i-th element of the deduced data, The average of the actual data. is the average value of the extrapolated data, and n is the total number of elements in the test set.

[0081] Standard deviation σ of real data and projected data x and σ y The formula for calculation is:

[0082]

[0083] The root mean square error E between the extrapolated data and the real data is... ' The formula for calculation is:

[0084]

[0085] Based on root mean square error and standard deviation of real data σ x σ, the standard deviation of the extrapolated data y The correlation coefficient R between Pearson and Taylor plots indicates the inference performance of four types of deep learning models.

[0086] Further calculation of the ordinary root mean square error E between the projected data and the actual data:

[0087]

[0088] The following scoring formula is used to comprehensively evaluate the performance of the inference model:

[0089]

[0090] In the formula, P represents the performance evaluation score of the simulation model, q1 is the correlation coefficient score weight, q2 is the standard deviation score weight, and q3 is the root mean square error score weight. In power quality situation simulation, it is first necessary to accurately predict outliers to prevent equipment damage; secondly, it is necessary to accurately capture changing trends to provide early warnings; and finally, it is necessary to reasonably reflect the degree of variation for state assessment. Therefore, the correlation coefficient score weight q1, the standard deviation score weight q2, and the root mean square error score weight q3 are set to 0.3, 0.1, and 0.6, respectively.

[0091] Based on the highest-scoring power quality model, a multi-dimensional source-load disturbance intensity sensitivity analysis is conducted. Taking voltage deviation as an example, a comprehensive evaluation is performed from four dimensions: peak value, mean value, fluctuation, and cumulative effect of the disturbance.

[0092] First, a ±5% rate of change is performed on each source-load disturbance term with a step size of 1%, and the degree of voltage deviation at each node is quantitatively analyzed within this range. The relative rate of change ΔV of the source-load disturbance term b with respect to the voltage deviation at node c is then determined at a rate of change k. bck Defined as:

[0093]

[0094] In the formula, V b,c,k For node c, the voltage deviation caused by the distributed source load b at a rate of change k is V. c,m Let be the voltage deviation of node c under basic operating conditions. Based on this definition, the impact of source load disturbances is evaluated from the following four dimensions: First, through the extreme value impact index M... b,c The ultimate impact capability of distributed source load b on node c is evaluated, and this index reflects the system's response under the worst-case scenario. Secondly, an average impact index μ is introduced. b,c The overall impact level of source-load disturbances is assessed, where K is the total number of sampling points. k is the sampling point number, starting from -5% and increasing by 1% to end at 5%, for a total of 11 points; third, the standard deviation index σ is used. b,c Assess the severity of voltage fluctuations caused by source-load disturbances; finally, use the curve integral area index A. bc Assess the cumulative impact of source load disturbances. The calculation formula is as follows:

[0095]

[0096]

[0097] To comprehensively assess the impact of source-load disturbances, a comprehensive sensitivity index S was constructed. b,c This indicator combines the evaluation results from the four dimensions mentioned above using a weighted approach. Maximum Impact M b,cThe weight assigned 0.4 reflects the strong focus on the system's security boundary; the average impact μ b,c A weight of 0.3 reflects the importance of considering the quality of the system's steady-state operation; the standard deviation index σ b,c Area index of curve integral A b,c Weights of 0.2 and 0.1 are assigned respectively to supplement the assessment of the dynamic characteristics and persistent effects of source load disturbances. The calculation formula is as follows:

[0098] S b,c =0.4M b,c +0.3μ b,c +0.2σ b,c +0.1A b,c

[0099] Based on the above sensitivity evaluation system, the comprehensive sensitivity index of each source load disturbance is calculated for each node and sorted to identify the source loads that have a dominant influence on the voltage quality of the node.

[0100] Taking the power and voltage data of 27 photovoltaic users in a rural power distribution network as an example, with a data time density of 15 minutes, a network structure is constructed based on four deep learning models: FCNN, CNN-LSTM, Transformer, and LSTM. Figure 5 As shown. After the models converge on the training set, the performance of the four models on the test set and the voltage deviation exceeding the limit are as follows. Figure 6 As shown. A 7% voltage deviation threshold, based on national standards, was used. The relative errors of the four types of curves are similar; therefore, the accuracy differences among the four types of models cannot be effectively assessed solely from the perspective of image trends and average relative errors.

[0101] After applying the evaluation method of this invention, the spatial relationships of the four deep learning models on the Taylor graph are as follows: Figure 7As shown in the figure, the FCNN model is located furthest from the reference point, with a standard deviation of 0.632, which is significantly lower than the reference value and the lowest among the four models. However, its correlation coefficient reaches 0.861, the highest among the four models. In contrast, the LSTM model is located closer to the reference point, and its standard deviation is closer to the statistical characteristics of the measured data, but its correlation coefficient is slightly lower. The Transformer model exhibits the best overall performance in the Taylor plot, being closest to the reference point, with a standard deviation that closely matches the measured data, a high correlation coefficient, and a relatively low root mean square error. The CNN-LSTM hybrid model performs between LSTM and Transformer; although its accuracy is not as good as the Transformer model, it performs well in terms of standard deviation matching. The proposed method reveals the differences among different models in terms of prediction accuracy, trend fitting ability, and data dispersion modeling, providing a multi-dimensional quantitative evaluation method for users to select the best model based on application scenarios.

[0102] The Transformer model with the highest score was selected for source load perturbation intensity sensitivity analysis. The time-varying characteristics of the top two key perturbation source loads over a 24-hour period on the simulation date are as follows: Figure 8 As shown, its time-varying pattern is affected by photovoltaic (PV) output. During the night and early morning hours, PV users stop generating power, the volatility of voltage deviation disturbance sources decreases, and the location of the nodes dominating power quality during these periods remains relatively fixed. The top two most influential nodes remain constant. Around noon, distributed PV output increases, the volatility of voltage deviation disturbance sources intensifies, and the dominant nodes affecting power quality begin to switch frequently. During periods of frequent voltage exceedances, the duration distribution of the top two most influential nodes is shown in the figure. Figure 9 As shown, distributed photovoltaic panels 26 and 27 were the key factors that triggered this voltage exceedance issue.

[0103] Based on the above technical solutions, this invention comprehensively considers three key indicators: standard deviation, correlation coefficient, and root mean square error. It constructs a scoring system based on power quality prediction requirements and the Taylor diagram mechanism. Numerical examples verify the effectiveness of the proposed method in revealing the differentiated performance of different models in terms of prediction accuracy, trend fitting ability, and data dispersion modeling. The evaluation framework proposed in this invention provides a scientific basis for the selection and optimization of power quality situation prediction models for distribution networks, and offers a general evaluation method for engineering prediction problems sensitive to the shape of time-series curves. Secondly, addressing the issue of the varying source-load correlation strength that causes local power quality problems, this invention considers the sensitivity differences among multivariate variables in deep learning prediction models and proposes a source-load disturbance correlation strength identification technology. This achieves a quantitative assessment of the causal correlation strength between source loads and weak nodes in the disturbance propagation path, identifies the key source loads causing local power quality problems, and realizes accurate reverse tracing of power quality problems and accurate location of key source loads, providing scientific decision support for the prevention, control, and responsibility allocation of power quality problems in distribution networks.

[0104] Figure 10 An embodiment of a computer device according to the present invention is shown. The computer device may be a server, and includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores static and dynamic information data. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the above-described method embodiment.

[0105] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0106] The present invention provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0107] In addition, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0108] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0109] This invention is not limited to the structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this invention is limited only by the appended claims.

Claims

1. A method for evaluating a power quality situation projection model, characterized in that, include: Acquire power quality monitoring data of key nodes in the distribution network and form a dataset containing reference data and data to be predicted; Based on the obtained dataset, a deep learning situation inference model is constructed, and a performance evaluation framework for the deep learning situation inference model is built based on the Taylor diagram. The performance metrics of the obtained performance evaluation framework are used to calculate skill scores, and the skill scores are ranked to determine the deep learning situation inference model with the highest score. Based on the deep learning situational inference model with the highest score, the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system is assessed.

2. The power quality situation deduction model evaluation method according to claim 1, characterized in that, The power quality monitoring data includes: voltage data, current data, power factor data, and harmonic content data.

3. The power quality situation projection model evaluation method according to claim 1, characterized in that, The dataset consisting of reference data and data to be predicted includes: The obtained power quality monitoring data is preprocessed, and a dataset containing reference data and data to be predicted is constructed based on the preprocessed power quality monitoring data. Data preprocessing includes: noise reduction filtering, outlier removal, missing value imputation, time series alignment, data normalization, and data standardization.

4. The power quality situation deduction model evaluation method according to claim 1, characterized in that, The deep learning situational inference model takes the distribution network topology, load characteristics, and power generation output as input features, and power quality indicators as output targets.

5. The power quality situation deduction model evaluation method according to claim 1, characterized in that, The performance evaluation framework uses a polar coordinate system to display the statistical relationship between the prediction results of the deep learning situational inference model and the actual monitoring data; The performance metrics of the performance evaluation framework include: correlation coefficient, standard deviation, and root mean square error.

6. The power quality situation projection model evaluation method according to claim 1, characterized in that, Based on the highest-scoring deep learning situational inference model, the impact assessment of distributed source-load power fluctuations on power quality indicators of various nodes in the distribution network system includes: Based on the highest-scoring deep learning situational inference model, the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system is assessed from four dimensions: peak value, mean value, volatility, and cumulative effect of disturbance.

7. The power quality situation deduction model evaluation method according to claim 1, characterized in that, Also includes: Based on the impact assessment results, distributed source loads are ranked according to their impact level to identify key source loads that cause power quality problems during periods when power quality exceeds limits.

8. A power quality situation prediction model evaluation system, characterized in that, include: The data processing module is used to acquire power quality monitoring data of key nodes in the distribution network and form a dataset containing reference data and data to be predicted. The model processing module is used to build a deep learning situation inference model based on the obtained dataset, and to build a performance evaluation framework for the deep learning situation inference model based on the Taylor diagram. The model determination module is used to calculate skill scores for the performance indicators of the obtained performance evaluation framework, obtain skill score values, sort the skill score values, and determine the deep learning situation inference model with the highest score. The model evaluation module is used to assess the impact of distributed source-load power fluctuations on power quality indicators of each node in the distribution network system based on the deep learning situational inference model with the highest score.

9. The power quality situation projection model evaluation system according to claim 8, characterized in that, The power quality monitoring data includes: voltage data, current data, power factor data, and harmonic content data.

10. The power quality situation projection model evaluation system according to claim 8, characterized in that, When forming a dataset containing reference data and data to be predicted, the data processing module performs data preprocessing on the obtained power quality monitoring data, and constructs a dataset containing reference data and data to be predicted based on the preprocessed power quality monitoring data. Data preprocessing includes: noise reduction filtering, outlier removal, missing value imputation, time series alignment, data normalization, and data standardization.

11. The power quality situation projection model evaluation system according to claim 8, characterized in that, The deep learning situational inference model takes the distribution network topology, load characteristics, and power generation output as input features, and power quality indicators as output targets.

12. The power quality situation projection model evaluation system according to claim 8, characterized in that, The performance evaluation framework uses a polar coordinate system to display the statistical relationship between the prediction results of the deep learning situational inference model and the actual monitoring data; The performance metrics of the performance evaluation framework include: correlation coefficient, standard deviation, and root mean square error.

13. The power quality situation projection model evaluation system according to claim 8, characterized in that, When the model evaluation module assesses the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system based on the deep learning situational inference model with the highest score, it evaluates the impact of distributed source-load power fluctuations on the power quality indicators of each node in the distribution network system from four dimensions: peak value, mean value, volatility and cumulative effect of the disturbance.

14. The power quality situation projection model evaluation system according to claim 8, characterized in that, Also includes: The source-load tracing module is used to sort distributed source loads based on their impact level according to the impact assessment results, and identify the key source loads that cause power quality problems during periods when power quality exceeds the limit.

15. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images