Power system high-precision modeling method based on dynamic parameter traceability

Abstract

The invention relates to the general technical field of artificial intelligence, in particular to a power system high-precision modeling method based on dynamic parameter traceablility, and the method comprises the steps: S1, collecting multi-source data of a power system, and S2, carrying out the time-space alignment processing of the multi-source data of the power system, according to the method, quality evaluation is carried out on multi-source data of a target power system through the step S3, difference value indexes are obtained through the step S4, the state of deviation times is judged through the step S5, the state of deviation coefficients of different source parameters is judged through the step S6, and the target power system multi-source data is obtained through the step S5. According to the method, the state of the difference cumulative number of times is judged through the step S7, so that the accuracy of dynamic parameter tracing is improved, and the accuracy of relay protection setting calculation is further improved.

Description

Technical Field

[0001] This invention relates to the field of general artificial intelligence technology, and in particular to a high-precision modeling method for power systems based on dynamic parameter tracing. Background Technology

[0002] Modern power plants and substations deploy SCADA, PMU, fault recording, and online monitoring systems, generating massive amounts of data across multiple time scales (from seconds to microseconds) and types (steady-state, transient, and environmental). However, this data is scattered across different systems with varying formats and time scales, forming "data silos." The lack of effective means to integrate and transform this data into useful information for models results in a significant waste of data resources. Traditional model parameters primarily rely on equipment nameplate and design values, which are considered constant. However, this static assumption introduces substantial model errors. Furthermore, after grid topology changes and equipment replacements, models are often not updated in a timely manner, leading to a gradual mismatch between the digital model and the physical system—a discrepancy between the model and reality. Model verification typically relies on field tests under a few specific operating conditions, which is costly, risky, and has limited coverage, failing to achieve continuous online verification across all operating conditions.

[0003] Chinese patent application CN119026472A discloses a modeling method, apparatus, computer program product, and modeling system for power systems. The method includes: acquiring relevant information about the power system; constructing an intelligent generative model; inputting the relevant information into the intelligent generative model to obtain virtual model data corresponding to the relevant information; and performing modeling based on the virtual model data corresponding to the relevant information to obtain a virtual model, wherein the virtual model is a virtual model of the power system. It is evident that this approach still suffers from the problem of insufficient model accuracy due to the lack of real-time updating and optimization of key dynamic parameter estimates, resulting in low accuracy in tracing dynamic parameters in the power system. Summary of the Invention

[0004] To address this issue, the present invention provides a high-precision modeling method for power systems based on dynamic parameter tracing, which overcomes the problem in existing technologies where the lack of real-time updates and optimizations of key dynamic parameter estimates leads to low accuracy in dynamic parameter tracing within power systems, resulting in insufficient model precision.

[0005] To achieve the above objectives, this invention provides a high-precision modeling method for power systems based on dynamic parameter tracing, the method comprising: Step S1: Collect multi-source data from the power system; Step S2: Perform spatiotemporal alignment processing on the multi-source data of the power system to obtain the multi-source data of the target power system; Step S3: Perform quality assessment on the multi-source data of the target power system to obtain multi-source data quality indicators, and clean and repair the multi-source data of the target power system according to the multi-source data quality indicators. Step S4: Perform data correction on the multi-source data of the target power system to obtain the estimated values ​​of key dynamic parameters, obtain the difference value index based on the estimated values ​​of key dynamic parameters, and perform parameter correction on the estimated values ​​of key dynamic parameters based on the difference value index. Step S5: Obtain the number of deviations based on the difference value index, and optimize the data correction process based on the number of deviations. Step S6: Construct a parameter knowledge graph based on the estimated values ​​of key dynamic parameters, and optimize the acquisition process of the difference value index based on the parameter knowledge graph; Step S7: Obtain the cumulative number of differences and update the data based on the cumulative number of differences in the data optimization process; Step S8: Generate a digital twin model of the electrical system based on multi-source data and key dynamic parameter estimates of the target power system.

[0006] Furthermore, in step S2, when performing spatiotemporal alignment processing on the multi-source data of the power system, the spatiotemporal alignment processing method is used to perform spatiotemporal alignment processing on the multi-source data of the power system. The spatiotemporal alignment processing method includes: Step A01: Construct the time alignment model; Step A02: Input the multi-source data of the power system into the time alignment model to obtain the time-unified multi-source data of the power system output by the time alignment model; Step A03: Input the time-unified power system multi-source data into the weighted least squares state estimation model to obtain the target power system multi-source data output by the spatial alignment model.

[0007] Further, in step S3, when performing quality assessment on the multi-source data of the target power system, a quality assessment model is constructed. The multi-source data of the target power system is input into the quality assessment model to obtain a set of quality assessment indicators output by the quality assessment model. The set of quality assessment indicators includes data freshness Di, data integrity Dq, and data consistency De. Based on the data freshness Di, data integrity Dq, data consistency De, data freshness coefficient w1, data integrity coefficient w2, and data consistency coefficient w3, the multi-source data quality indicator DIq is calculated. DIq is set as Di×w1+Dq×w2+De×w3, and w1+w2+w3=1, thus obtaining the multi-source data quality indicator DIq.

[0008] Further, in step S3, when cleaning and repairing the multi-source data of the target power system according to the multi-source data quality index, the multi-source data quality index DIq is compared with the preset multi-source data quality index DIq0. Based on the comparison result, the compliance status of the target multi-source data quality index is judged, and the multi-source data of the target power system is cleaned and repaired based on the judgment result, wherein: When DIq≥DIq0, the target multi-source data quality indicators are deemed to be up to standard, and no cleaning or repair is performed on the target power system multi-source data. When DIq < DIq0, the target multi-source data quality index is determined to be non-compliant, and the target power system multi-source data is cleaned and repaired: the target power system multi-source data is cleaned and repaired using the Kalman filter algorithm to obtain cleaned and repaired data, and the cleaned and repaired data is output as the target power system multi-source data.

[0009] Furthermore, in step S4, when performing data correction on the multi-source data of the target power system, a dynamic parameter estimation model is constructed, and the dynamic measurement data and environmental data in the multi-source data of the target power system are input into the dynamic parameter estimation model to obtain the key dynamic parameter estimate β output by the dynamic parameter estimation model; In step S4, when obtaining the difference index based on the estimated value of the key dynamic parameter, the difference index Et is calculated based on the measured value of the key dynamic parameter Mu and the estimated value of the key dynamic parameter β. Et is set to Mu-β to obtain the difference index Et.

[0010] Further, in step S4, when correcting the estimated value of the key dynamic parameter based on the difference value index, the difference value index Et is compared with the preset difference value index Et0. The state of the difference value index is judged based on the comparison result, and the estimated value of the key dynamic parameter is corrected based on the judgment result, wherein: When Et≤Et0, the state of the difference value index is determined to be unbiased, and no parameter correction is performed on the estimated value of the key dynamic parameter. When Et > Et0, the state of the difference value index is determined to be biased, and parameter correction is performed on the key dynamic parameter estimate: the key dynamic parameter estimate is regenerated to obtain the regenerated key dynamic parameter estimate, and the difference value index is re-acquired based on the regenerated key dynamic parameter estimate.

[0011] Further, in step S5, the number of deviations is obtained based on the difference value index, the number of deviations Pt is compared with the preset number of deviations Pt0, the state of the number of deviations is judged based on the comparison result, and the data correction process is optimized based on the judgment result, wherein: When Pt≤Pt0, the number of deviations is determined to be low, and no data optimization is performed during the data correction process. When Pt > Pt0, the number of deviations is determined to be high, and the data correction process is optimized using federated learning optimization methods.

[0012] Furthermore, in step S6, when constructing a parameter knowledge graph based on the estimated values ​​of key dynamic parameters and comparing the consistency of different source parameters of the same device based on the parameter knowledge graph, the power system entities are used as nodes of the graph database, and the entity association attributes are used as edges of the graph database to construct the parameter knowledge graph, thus obtaining the parameter knowledge graph. In step S6, when optimizing the acquisition process of the difference value index based on the parameter knowledge graph, the deviation coefficients Mt of different source parameters are obtained. These deviation coefficients Mt are then compared with preset deviation coefficients Mt0 of different source parameters. Based on the comparison results, the state of the deviation coefficients is determined, and the acquisition process of the difference value index is optimized based on the determination results. Specifically: When Mt≤Mt0, the state of the deviation coefficient of different source parameters is determined to be unbiased, and no index optimization is performed on the process of obtaining the difference value index. When Mt > Mt0, the state of the deviation coefficients of different source parameters is determined to be biased. The process of obtaining the difference value index is optimized by changing the process of obtaining the difference value index to input the measured key dynamic parameter value Mu and the estimated key dynamic parameter value β into the difference assessment model to obtain the optimized difference value index Et' output by the difference assessment model, and the optimized difference value index Et' is output as the difference value index Et.

[0013] Further, in step S7, when acquiring the cumulative number of differences and updating the data during the data optimization process based on the cumulative number of differences, the cumulative number of differences Ct is acquired through the system log. The cumulative number of differences Ct is compared with a preset cumulative number of differences Ct0. Based on the comparison result, the status of the cumulative number of differences is determined, and the data optimization process is updated based on the determination result. Wherein: When Ct≤Ct0, the cumulative number of differences is determined to be low, and no data update is performed during the data optimization process; When Ct > Ct0, the cumulative difference count is determined to be high. The data optimization process involves updating the data. The preset deviation count Pt0 is updated according to the data update coefficient gh to obtain the updated preset deviation count Pt1, where Pt1 = Pt0 × gh. The updated preset deviation count Pt1 is rounded to obtain the rounded preset deviation count Pt0'. The rounded preset deviation count Pt0' is output as the preset deviation count Pt0, and the deviation count Pt is compared with the preset deviation count Pt0 again.

[0014] Furthermore, in step S8, a digital twin model of the electrical system is generated based on multi-source data of the target power system and estimates of key dynamic parameters.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: the method collects multi-source data of the power system in step S1; the method further performs spatiotemporal alignment processing on the multi-source data of the power system in step S2 to resolve the differences in occurrence time and sampling frequency of the multi-source data of the power system, and maps it spatially to electrical nodes, which facilitates model calculation and provides a basis for dynamic parameter tracing; the method further performs quality assessment on the multi-source data of the target power system in step S3, converting the quality of the multi-source data of the target power system into specific evaluation values, so as to facilitate timely processing of the multi-source data of the target power system and improve the accuracy of the multi-source data of the target power system; the method further obtains the difference value index in step S4 to estimate the key dynamic parameters in the power system. The real-time accuracy of the calculation is numerically displayed. The method also judges the status of the number of deviations in step S5 to avoid inaccurate data correction due to excessively high deviation counts. The method also judges the status of the deviation coefficients of different source parameters in step S6 to reduce the impact of the deviation of the estimated values ​​of key dynamic parameters of the same equipment in different data sources on the difference value index. The method also judges the status of the cumulative number of differences in step S7 to reduce the accuracy of the determination of the status of the number of deviations by the cumulative number of differences, thereby improving the accuracy of dynamic parameter tracing. The method also generates a digital twin model of the electrical system in step S8 to further improve the reliability of relay protection setting calculation based on the high-precision digital twin model of the electrical system. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the high-precision power system modeling method based on dynamic parameter tracing in this embodiment. Figure 2 This is a flowchart illustrating the spatiotemporal alignment processing method in this embodiment. Detailed Implementation

[0017] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0018] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0019] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0020] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0021] Please see Figure 1 The diagram shown is a flowchart of the high-precision power system modeling method based on dynamic parameter tracing in this embodiment. The method includes: Step S1: Collect multi-source data from the power system; Step S2: Perform spatiotemporal alignment processing on the multi-source data of the power system to obtain the multi-source data of the target power system; Step S3: Perform quality assessment on the multi-source data of the target power system to obtain multi-source data quality indicators, and clean and repair the multi-source data of the target power system according to the multi-source data quality indicators. Step S4: Perform data correction on the multi-source data of the target power system to obtain the estimated values ​​of key dynamic parameters, obtain the difference value index based on the estimated values ​​of key dynamic parameters, and perform parameter correction on the estimated values ​​of key dynamic parameters based on the difference value index. Step S5: Obtain the number of deviations based on the difference value index, and optimize the data correction process based on the number of deviations. Step S6: Construct a parameter knowledge graph based on the estimated values ​​of key dynamic parameters, and optimize the acquisition process of the difference value index based on the parameter knowledge graph; Step S7: Obtain the cumulative number of differences and update the data based on the cumulative number of differences in the data optimization process; Step S8: Generate a digital twin model of the electrical system based on multi-source data and key dynamic parameter estimates of the target power system.

[0022] Specifically, the high-precision power system modeling method based on dynamic parameter tracing is applied to power generation and monitoring equipment, such as substations and power plants. This method improves the accuracy of dynamic parameter tracing by acquiring, updating, and optimizing multi-source data and key dynamic parameter estimates of the target power system, thereby enhancing the accuracy of relay protection setting calculations. Specifically, the method collects multi-source data of the power system in step S1. Step S2 performs spatiotemporal alignment processing on the multi-source data to address differences in occurrence time and sampling frequency, and spatially maps it to electrical nodes, facilitating model calculations and providing a foundation for dynamic parameter tracing. Step S3 further evaluates the quality of the multi-source data of the target power system, converting the quality into specific evaluation values ​​to facilitate timely processing and improvement of the target power system's multi-source data. To improve the accuracy of multi-source data in the target power system, the method further acquires the difference value index in step S4 to numerically represent the real-time accuracy of the estimated values ​​of key dynamic parameters in the power system. The method also judges the state of the number of deviations in step S5 to avoid inaccurate data correction due to excessively high deviation counts. Furthermore, the method judges the state of the deviation coefficients of different source parameters in step S6 to reduce the impact of deviations in the estimated values ​​of key dynamic parameters from different data sources on the difference value index. The method also judges the state of the cumulative number of differences in step S7 to reduce the accuracy of the determination of the state of the number of deviations based on the cumulative number of differences, thereby improving the accuracy of dynamic parameter tracing. Finally, the method generates a digital twin model of the electrical system in step S8 to further improve the reliability of relay protection setting calculations based on the high-precision digital twin model of the electrical system.

[0023] Specifically, in step S1, multi-source data of the power system is collected, including static parameters, steady-state measurement data, dynamic measurement data, transient fault data, and environmental data.

[0024] Specifically, the static parameters refer to the initial parameters of the equipment obtained from the equipment nameplate and equipment design drawings, such as resistance and reactance. This embodiment does not limit the specific method of collecting static parameters; those skilled in the art can freely choose according to actual needs. For example, static parameters can be collected through the equipment nameplate and design drawings. In this embodiment, steady-state measurement data is collected through a SCADA system. The SCADA system refers to the data acquisition and monitoring control system in the power grid. The steady-state measurement data includes active power, reactive power, real-time voltage, and real-time current. Active power refers to the power that actually does work and produces effects in the power system. Reactive power refers to the power necessary to establish and maintain the electric and magnetic fields, but which does not do work on average. Real-time voltage refers to the voltage at the current moment in the power system. The voltage value is given below, and the real-time current refers to the current value in the power system at the current moment. In this embodiment, dynamic measurement data is collected through a PMU, which is a synchronous phasor measurement unit in the power grid. The dynamic measurement data includes voltage phasors with precise time scales and current phasors with precise time scales. The transient fault data refers to the voltage and current waveforms obtained by the fault recording device. The environmental data includes line temperature, ambient temperature, and ambient humidity. The line temperature refers to the temperature of the conductive part of the cable collected by a temperature sensor installed in the power system. The ambient temperature refers to the temperature value of the environment around the cable collected by a temperature sensor installed in the power system. The ambient humidity refers to the humidity value of the environment around the cable collected by a humidity sensor installed in the power system.

[0025] Specifically, in step S1, multi-source data of the power system is collected to facilitate subsequent spatiotemporal alignment processing based on the multi-source data of the power system, thereby improving the accuracy and reliability of relay protection setting calculation.

[0026] Specifically, in step S2, when performing spatiotemporal alignment processing on the multi-source data of the power system, the spatiotemporal alignment processing method is used to perform spatiotemporal alignment processing on the multi-source data of the power system.

[0027] Specifically, in step S3, when assessing the quality of multi-source data of the target power system, a quality assessment model is constructed. The multi-source data of the target power system is input into the quality assessment model to obtain a set of quality assessment indicators output by the quality assessment model. The set of quality assessment indicators includes data freshness Di, data integrity Dq, and data consistency De. Based on the data freshness Di, data integrity Dq, data consistency De, data freshness coefficient w1, data integrity coefficient w2, and data consistency coefficient w3, the multi-source data quality indicator DIq is calculated. DIq is set as Di×w1+Dq×w2+De×w3, and w1+w2+w3=1, thus obtaining the multi-source data quality indicator DIq.

[0028] Specifically, the quality assessment model refers to a recurrent neural network model that uses multi-source data of the target power system as input data and a set of quality assessment indicators as output data. This embodiment does not limit the specific construction method of the quality assessment model; those skilled in the art can freely choose according to actual needs. For example, historical multi-source data of the target power system and its corresponding quality assessment indicator set can be used as a training set to train the recurrent neural network model to obtain the quality assessment model. Data freshness refers to the numerical value obtained from the quality assessment model that measures the latency of the multi-source data of the target power system; data completeness refers to the numerical value obtained from the quality assessment model that measures the completeness of the multi-source data of the target power system; and data consistency refers to the numerical value obtained from the quality assessment model that measures the consistency of the multi-source data of the target power system. The numerical values ​​obtained are used to measure the degree of conflict in multi-source data of the target power system. The data freshness coefficient refers to the value that measures the importance of data freshness in the multi-source data quality index. The data integrity coefficient refers to the value that measures the importance of data integrity in the multi-source data quality index. The data consistency coefficient refers to the value that measures the importance of data consistency in the multi-source data quality index. This embodiment does not limit the specific values ​​of the data freshness coefficient w1, data integrity coefficient w2, and data consistency coefficient w3. Those skilled in the art can freely choose according to actual needs. For example, if this embodiment focuses on using data integrity as the main parameter to measure the multi-source data of the target power system, then w1=0.3, w2=0.4, and w3=0.3 are set.

[0029] Specifically, in step S3, the quality of the multi-source data of the target power system is evaluated and converted into specific evaluation values ​​so as to facilitate timely processing of the multi-source data of the target power system and improve the accuracy of dynamic parameter tracing.

[0030] Specifically, in step S3, when cleaning and repairing the multi-source data of the target power system according to the multi-source data quality index, the multi-source data quality index DIq is compared with the preset multi-source data quality index DIq0. Based on the comparison result, the compliance status of the target multi-source data quality index is judged, and the multi-source data of the target power system is cleaned and repaired based on the judgment result. When DIq≥DIq0, the target multi-source data quality indicators are deemed to be up to standard, and no cleaning or repair is performed on the target power system multi-source data. When DIq < DIq0, the target multi-source data quality index is determined to be non-compliant, and the target power system multi-source data is cleaned and repaired: the target power system multi-source data is cleaned and repaired using the Kalman filter algorithm to obtain cleaned and repaired data, and the cleaned and repaired data is output as the target power system multi-source data.

[0031] Specifically, the preset multi-source data quality index refers to a preset value used to judge the compliance status of the target multi-source data quality index. This embodiment does not calculate the specific value of the preset multi-source data quality index DIq0. Those skilled in the art can freely choose according to actual needs. For example, based on historical experience, when DIq0 < 0.85, the multi-source data quality is likely to fail to reach the ideal accuracy. When DIq0 > 0.85, it is likely to lead to a large number of repeated cleaning and repair processes, causing system instability. Therefore, this embodiment sets DIq0 = 0.85. The compliance status of the target multi-source data quality index refers to the degree of compliance of the multi-source data quality index. The compliance status of the target multi-source data quality index includes compliance and non-compliance. The Kalman filter algorithm refers to the existing technology algorithm for repairing data delay, data missing and multi-source data conflict in the target power system multi-source data.

[0032] Specifically, in step S3, the compliance status of the target multi-source data quality indicators is judged. When the compliance status of the target multi-source data quality indicators is not met, the target power system multi-source data is cleaned and repaired in a timely manner to improve the quality of the target multi-source data in real time, thereby improving the accuracy of dynamic parameter traceability.

[0033] Specifically, in step S4, when performing data correction on the multi-source data of the target power system, a dynamic parameter estimation model is constructed, and the dynamic measurement data and environmental data in the multi-source data of the target power system are input into the dynamic parameter estimation model to obtain the key dynamic parameter estimate β output by the dynamic parameter estimation model. In step S4, when obtaining the difference index based on the estimated value of the key dynamic parameter, the difference index Et is calculated based on the measured value of the key dynamic parameter Mu and the estimated value of the key dynamic parameter β. Et is set to Mu-β to obtain the difference index Et.

[0034] Specifically, the dynamic parameter estimation model refers to a deep belief network model that takes dynamic measurement data and environmental data as input data and key dynamic parameter estimates as output data. This embodiment does not limit the specific construction method of the dynamic parameter estimation model. Those skilled in the art can freely choose according to actual needs. For example, historical dynamic measurement data and environmental data can be used as historical key parameter sets, and the historical key parameter sets and their corresponding key dynamic parameter estimates can be used as training sets to train the deep belief network model to obtain the dynamic parameter estimation model. The key dynamic parameter estimates refer to the temperature drift line impedance obtained according to the dynamic parameter estimation model, and the measured key dynamic parameter values ​​refer to the actual temperature drift line impedance. This embodiment does not limit the specific method of obtaining the measured key dynamic parameter values. Those skilled in the art can freely choose according to actual needs. For example, the measured key dynamic parameter values ​​can be obtained through an online identification method based on PMU measurements.

[0035] Specifically, in step S4, the difference value index is obtained to numerically represent the real-time accuracy of the estimated values ​​of key dynamic parameters in the power system, so that the estimated values ​​of key dynamic parameters can be corrected in a timely manner based on the difference value index.

[0036] Specifically, in step S4, when correcting the estimated value of the key dynamic parameter based on the difference value index, the difference value index Et is compared with the preset difference value index Et0. The state of the difference value index is judged based on the comparison result, and the estimated value of the key dynamic parameter is corrected based on the judgment result, wherein: When Et≤Et0, the state of the difference value index is determined to be unbiased, and no parameter correction is performed on the estimated value of the key dynamic parameter. When Et > Et0, the state of the difference value index is determined to be biased, and parameter correction is performed on the key dynamic parameter estimate: the key dynamic parameter estimate is regenerated to obtain the regenerated key dynamic parameter estimate, and the difference value index is re-acquired based on the regenerated key dynamic parameter estimate.

[0037] Specifically, the preset difference value index refers to a preset value for judging the state of the difference value index. This embodiment does not limit the specific value setting of the preset difference value index Et0. Those skilled in the art can freely choose according to actual needs. For example, this embodiment sets Et0=0.22 based on historical experience. The state of the difference value index refers to the deviation of the difference value index judged according to the difference value index and the preset difference value index. The state of the difference value index includes no deviation and deviation.

[0038] Specifically, in step S4, the state of the difference value index is judged. When the state of the difference value index is that there is a deviation, the estimated value of the key dynamic parameter is corrected in time to reduce the impact of the difference value index on the estimated value of the key dynamic parameter, thereby improving the accuracy of dynamic parameter tracing.

[0039] Specifically, in step S5, the number of deviations is obtained based on the difference value index, the number of deviations Pt is compared with the preset number of deviations Pt0, the status of the number of deviations is judged based on the comparison result, and the data correction process is optimized based on the judgment result, wherein: When Pt≤Pt0, the number of deviations is determined to be low, and no data optimization is performed during the data correction process. When Pt > Pt0, the number of deviations is determined to be high, and the data correction process is optimized using federated learning optimization methods.

[0040] Specifically, the number of deviations refers to the number of times the difference value index obtained through system logs within a preset time period is judged to have a deviation. This embodiment does not limit the specific length of the preset time, and those skilled in the art can freely choose it according to actual needs. For example, this embodiment sets the preset time to 24 hours. The preset number of deviations refers to the preset value for judging the state of the number of deviations. This embodiment does not limit the specific value of the preset number of deviations Pt0, and those skilled in the art can freely choose it according to actual needs. For example, this embodiment sets Pt0=5 times based on historical experience. The state of the number of deviations refers to the high or low frequency of the number of deviations judged based on the number of deviations and the preset number of deviations. The state of the number of deviations includes high frequency and low frequency. The federated learning optimization method refers to the data optimization method of the data correction process using the distributed machine learning paradigm of federated learning. For example, a local power plant uses local data to build a dynamic parameter estimation model to obtain an updated dynamic parameter estimation model, and uploads the updated dynamic parameter estimation model to the cloud for aggregation to obtain an aggregated dynamic parameter estimation model. The dynamic measurement data and environmental data from the multi-source data of the target power system are then re-input into the aggregated dynamic parameter estimation model.

[0041] Specifically, in step S5, by judging the state of the number of deviations, when the number of deviations is low, the data correction process is optimized in a timely manner to avoid the data correction process being inaccurate due to the number of deviations being too high, thereby improving the accuracy of relay protection setting calculation.

[0042] Specifically, in step S6, when constructing a parameter knowledge graph based on the estimated values ​​of key dynamic parameters and comparing the consistency of different source parameters of the same device based on the parameter knowledge graph, the power system entities are used as nodes of the graph database and the entity association attributes are used as edges of the graph database to construct the parameter knowledge graph, thus obtaining the parameter knowledge graph. In step S6, when optimizing the acquisition process of the difference value index based on the parameter knowledge graph, the deviation coefficients Mt of different source parameters are obtained. These deviation coefficients Mt are then compared with preset deviation coefficients Mt0 of different source parameters. Based on the comparison results, the state of the deviation coefficients is determined, and the acquisition process of the difference value index is optimized based on the determination results. Specifically: When Mt≤Mt0, the state of the deviation coefficient of different source parameters is determined to be unbiased, and no index optimization is performed on the process of obtaining the difference value index. When Mt > Mt0, the state of the deviation coefficients of different source parameters is determined to be biased. The process of obtaining the difference value index is optimized by changing the process of obtaining the difference value index to input the measured key dynamic parameter value Mu and the estimated key dynamic parameter value β into the difference assessment model to obtain the optimized difference value index Et' output by the difference assessment model, and the optimized difference value index Et' is output as the difference value index Et.

[0043] Specifically, the power system entities refer to related equipment and devices in the power system, such as "generators" and "transformers." The entity association attributes refer to the connection methods and association hierarchies between power system entities, such as "connected to" and "protected by." The graph database refers to a database used for storing, managing, and querying power system entities and entity association attributes. The different source parameter deviation coefficient refers to a numerical value that measures the degree of deviation between the estimated values ​​of key dynamic parameters of the same device in different data sources. This embodiment does not limit the specific method of obtaining the different source parameter deviation coefficient. Those skilled in the art can freely choose according to actual needs. For example, by constructing a different source parameter deviation estimation model and inputting the estimated values ​​of key dynamic parameters of the same device in different data sources into the different source parameter deviation estimation model, the different source parameter deviation coefficient output by the different source parameter deviation estimation model can be obtained. This embodiment does not limit the specific construction method of the different source parameter deviation estimation model. Those skilled in the art can freely choose according to actual needs. For example, by inputting the estimated values ​​of key dynamic parameters of the same device in different data sources in the past and their corresponding different source parameter deviations... The difference coefficient is used as a training set to train the machine learning model, resulting in a different source parameter deviation estimation model. The preset different source parameter deviation coefficient refers to a preset value for judging the state of the different source parameter deviation coefficient. This embodiment does not limit the specific value setting of the preset different source parameter deviation coefficient Mt0. Those skilled in the art can freely choose according to actual needs. For example, in this embodiment, Mt0 is set to 0.27 based on historical experience. The state of the different source parameter deviation coefficient refers to the deviation of the different source parameter deviation coefficient. The state of the different source parameter deviation coefficient includes no deviation and deviation. The difference evaluation model refers to a recurrent neural network model that takes the measured key dynamic parameter values ​​and key dynamic parameter estimates as input data and the optimized difference value index as output data. This embodiment does not limit the specific construction method of the difference evaluation model. Those skilled in the art can freely choose according to actual needs. For example, the measured key dynamic parameter values ​​and key dynamic parameter estimates that occurred in the past can be used as historical reference data, and the historical reference data and its corresponding optimized difference value index can be used as a training set to train the recurrent neural network model to obtain the difference evaluation model.

[0044] Specifically, in step S6, by judging the state of the deviation coefficients of different source parameters, when the state of the deviation coefficients of different source parameters is that there is a deviation, the process of obtaining the difference value index is optimized in a timely manner, so as to reduce the impact of the deviation of the estimated values ​​of key dynamic parameters of the same device in different data sources on the difference value index, thereby improving the accuracy of dynamic parameter tracing.

[0045] Specifically, in step S7, when acquiring the cumulative number of differences and updating the data during the data optimization process based on the cumulative number of differences, the cumulative number of differences Ct is acquired through the system log. The cumulative number of differences Ct is compared with a preset cumulative number of differences Ct0. Based on the comparison result, the status of the cumulative number of differences is determined, and the data optimization process is updated based on the determination result. Wherein: When Ct≤Ct0, the cumulative number of differences is determined to be low, and no data update is performed during the data optimization process; When Ct > Ct0, the cumulative difference count is determined to be high. The data optimization process involves updating the data. The preset deviation count Pt0 is updated according to the data update coefficient gh to obtain the updated preset deviation count Pt1, where Pt1 = Pt0 × gh. The updated preset deviation count Pt1 is rounded to obtain the rounded preset deviation count Pt0'. The rounded preset deviation count Pt0' is output as the preset deviation count Pt0, and the deviation count Pt is compared with the preset deviation count Pt0 again.

[0046] Specifically, the cumulative difference count refers to the number of times a deviation is determined based on the status of the deviation coefficients of different source parameters within a preset time period obtained from the system log. The preset cumulative difference count refers to a preset value for judging the status of the cumulative difference count. This embodiment does not limit the specific value of the preset cumulative difference count Ct0; those skilled in the art can freely choose according to actual needs. For example, this embodiment sets Pt0=3 times based on historical experience. The status of the cumulative difference count refers to the high and low counts of the cumulative difference count, which includes low and high counts. This embodiment does not limit the specific value of the data update coefficient gh; those skilled in the art can freely choose according to actual needs, as long as 0 < g. The requirement of h < 0.8 is sufficient. For example, in this embodiment, gh = 0.6 is set. In this embodiment, the range of values ​​for the data update coefficient gh is obtained through a coefficient evaluation experiment. The coefficient evaluation experiment refers to the experiment of repeatedly defining the boundary of the data update coefficient to determine the optimal range of values ​​for the data update coefficient. When gh > 0.8, the number of preset deviations increases significantly, and frequent data optimization leads to system instability. Therefore, the range of values ​​for the data update coefficient gh is set to 0 < gh < 0.8. The rounding process refers to the process of rounding the preset deviation number after the update to an integer. This embodiment does not limit the specific method of rounding. Those skilled in the art can freely choose according to actual needs, such as rounding the preset deviation number after the update to an integer.

[0047] Specifically, in step S7, by judging the state of the cumulative difference count, when the state of the cumulative difference count is high, the preset deviation count is reduced in time by updating the data coefficient, so as to reduce the accuracy of the determination of the state of the deviation count by the cumulative difference count, thereby improving the accuracy of dynamic parameter tracing and further improving the accuracy of relay protection setting calculation.

[0048] Specifically, in step S8, a digital twin model of the electrical system is generated based on multi-source data of the target power system and estimates of key dynamic parameters.

[0049] Specifically, this embodiment does not limit the specific generation method of the digital twin model of the electrical system. Those skilled in the art can freely choose according to actual needs. For example, the initial state simulation model can be constructed by using 3D MAX in the three-dimensional model building software based on multi-source data of the target power system, and the initial state simulation model can be physically simulated based on the estimated values ​​of key dynamic parameters to obtain the digital twin model of the electrical system.

[0050] Specifically, in step S8, a digital twin model of the electrical system is generated to further improve the reliability of relay protection setting calculations based on a high-precision digital twin model of the electrical system.

[0051] Please see Figure 2 As shown, this is a flowchart illustrating the spatiotemporal alignment processing method of this embodiment. The spatiotemporal alignment processing method includes: Step A01: Construct the time alignment model; Step A02: Input the multi-source data of the power system into the time alignment model to obtain the time-unified multi-source data of the power system output by the time alignment model; Step A03: Input the time-unified power system multi-source data into the weighted least squares state estimation model to obtain the target power system multi-source data output by the spatial alignment model.

[0052] Specifically, the time alignment model refers to a long short-term memory network model based on an attention mechanism that takes multi-source power system data as input data and time-unified multi-source power system data as output data. This embodiment does not limit the specific construction method of the time alignment model. Those skilled in the art can freely choose according to actual needs. For example, the long short-term memory network model based on an attention mechanism can be trained by using historical multi-source power system data and its corresponding time-unified multi-source power system data as training sets to obtain the time alignment model. The long short-term memory network model based on an attention mechanism refers to a network model composed of an attention mechanism and a long short-term memory network. The time-unified multi-source power system data refers to the multi-source power system data unified to the same time point obtained according to the time alignment model. The weighted least squares state estimation model refers to a mathematical model based on physical laws for spatial unification of time-unified multi-source power system data. Spatial unification refers to the process of mapping measurement data from different locations in the power system to a unified electrical node through topological association. The target multi-source power system data refers to the time-unified multi-source power system data mapped to a unified electrical node according to the weighted least squares state estimation model.

[0053] Specifically, in step S2, the multi-source data of the power system is spatiotemporally aligned to resolve the differences in occurrence time and sampling frequency of the multi-source data of the power system, and to map it to electrical nodes in space, which facilitates model calculation, provides a basis for dynamic parameter tracing, and thus improves the accuracy of relay protection setting calculation.

[0054] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A high-precision modeling method for power systems based on dynamic parameter tracing, characterized in that, The method includes: Step S1: Collect multi-source data from the power system; Step S2: Perform spatiotemporal alignment processing on the multi-source data of the power system to obtain the multi-source data of the target power system; Step S3: Perform quality assessment on the multi-source data of the target power system to obtain multi-source data quality indicators, and clean and repair the multi-source data of the target power system according to the multi-source data quality indicators. Step S4: Perform data correction on the multi-source data of the target power system to obtain the estimated values ​​of key dynamic parameters, obtain the difference value index based on the estimated values ​​of key dynamic parameters, and perform parameter correction on the estimated values ​​of key dynamic parameters based on the difference value index. Step S5: Obtain the number of deviations based on the difference value index, and optimize the data correction process based on the number of deviations. Step S6: Construct a parameter knowledge graph based on the estimated values ​​of key dynamic parameters, and optimize the acquisition process of the difference value index based on the parameter knowledge graph; Step S7: Obtain the cumulative number of differences and update the data based on the cumulative number of differences in the data optimization process; Step S8: Generate a digital twin model of the electrical system based on multi-source data and key dynamic parameter estimates of the target power system.

2. The high-precision power system modeling method based on dynamic parameter tracing according to claim 1, characterized in that, In step S2, when performing spatiotemporal alignment processing on the multi-source data of the power system, the spatiotemporal alignment processing method is used to perform spatiotemporal alignment processing on the multi-source data of the power system. The spatiotemporal alignment processing method includes: Step A01: Construct the time alignment model; Step A02: Input the multi-source data of the power system into the time alignment model to obtain the time-unified multi-source data of the power system output by the time alignment model; Step A03: Input the time-unified power system multi-source data into the weighted least squares state estimation model to obtain the target power system multi-source data output by the spatial alignment model.

3. The high-precision power system modeling method based on dynamic parameter tracing according to claim 2, characterized in that, In step S3, when performing quality assessment on the multi-source data of the target power system, a quality assessment model is constructed. The multi-source data of the target power system is input into the quality assessment model to obtain a set of quality assessment indicators output by the quality assessment model. The set of quality assessment indicators includes data freshness Di, data integrity Dq, and data consistency De. Based on the data freshness Di, data integrity Dq, data consistency De, data freshness coefficient w1, data integrity coefficient w2, and data consistency coefficient w3, the multi-source data quality indicator DIq is calculated. DIq is set as Di×w1+Dq×w2+De×w3, and w1+w2+w3=1, thus obtaining the multi-source data quality indicator DIq.

4. The high-precision power system modeling method based on dynamic parameter tracing according to claim 3, characterized in that, In step S3, when cleaning and repairing the multi-source data of the target power system according to the multi-source data quality index, the multi-source data quality index DIq is compared with the preset multi-source data quality index DIq0. Based on the comparison result, the compliance status of the target multi-source data quality index is judged, and the multi-source data of the target power system is cleaned and repaired based on the judgment result. Specifically: When DIq≥DIq0, the target multi-source data quality indicators are deemed to be up to standard, and no cleaning or repair is performed on the target power system multi-source data. When DIq < DIq0, the target multi-source data quality index is determined to be non-compliant, and the target power system multi-source data is cleaned and repaired: the target power system multi-source data is cleaned and repaired using the Kalman filter algorithm to obtain cleaned and repaired data, and the cleaned and repaired data is output as the target power system multi-source data.

5. The high-precision power system modeling method based on dynamic parameter tracing according to claim 4, characterized in that, In step S4, when performing data correction on the multi-source data of the target power system, a dynamic parameter estimation model is constructed, and the dynamic measurement data and environmental data in the multi-source data of the target power system are input into the dynamic parameter estimation model to obtain the key dynamic parameter estimate β output by the dynamic parameter estimation model. In step S4, when obtaining the difference index based on the estimated value of the key dynamic parameter, the difference index Et is calculated based on the measured value of the key dynamic parameter Mu and the estimated value of the key dynamic parameter β. Et is set to Mu-β to obtain the difference index Et.

6. The high-precision power system modeling method based on dynamic parameter tracing according to claim 5, characterized in that, In step S4, when correcting the estimated value of the key dynamic parameter based on the difference value index, the difference value index Et is compared with the preset difference value index Et0. The state of the difference value index is judged based on the comparison result, and the estimated value of the key dynamic parameter is corrected based on the judgment result, wherein: When Et≤Et0, the state of the difference value index is determined to be unbiased, and no parameter correction is performed on the estimated value of the key dynamic parameter. When Et > Et0, the state of the difference value index is determined to be biased, and parameter correction is performed on the key dynamic parameter estimate: the key dynamic parameter estimate is regenerated to obtain the regenerated key dynamic parameter estimate, and the difference value index is re-acquired based on the regenerated key dynamic parameter estimate.

7. The high-precision power system modeling method based on dynamic parameter tracing according to claim 6, characterized in that, In step S5, the number of deviations is obtained based on the difference value index, the number of deviations Pt is compared with the preset number of deviations Pt0, the status of the number of deviations is judged based on the comparison result, and the data correction process is optimized based on the judgment result, wherein: When Pt≤Pt0, the number of deviations is determined to be low, and no data optimization is performed during the data correction process. When Pt > Pt0, the number of deviations is determined to be high, and the data correction process is optimized using federated learning optimization methods.

8. The high-precision power system modeling method based on dynamic parameter tracing according to claim 7, characterized in that, In step S6, a parameter knowledge graph is constructed based on the estimated values ​​of key dynamic parameters. When comparing the consistency of different source parameters of the same device based on the parameter knowledge graph, the power system entities are used as nodes of the graph database, and the entity association attributes are used as edges of the graph database to construct the parameter knowledge graph, thus obtaining the parameter knowledge graph. In step S6, when optimizing the acquisition process of the difference value index based on the parameter knowledge graph, the deviation coefficients Mt of different source parameters are obtained. These deviation coefficients Mt are then compared with preset deviation coefficients Mt0 of different source parameters. Based on the comparison results, the state of the deviation coefficients is determined, and the acquisition process of the difference value index is optimized based on the determination results. Specifically: When Mt≤Mt0, the state of the deviation coefficient of different source parameters is determined to be unbiased, and no index optimization is performed on the process of obtaining the difference value index. When Mt > Mt0, the state of the deviation coefficients of different source parameters is determined to be biased. The process of obtaining the difference value index is optimized by changing the process of obtaining the difference value index to input the measured key dynamic parameter value Mu and the estimated key dynamic parameter value β into the difference assessment model to obtain the optimized difference value index Et' output by the difference assessment model, and the optimized difference value index Et' is output as the difference value index Et.

9. The high-precision power system modeling method based on dynamic parameter tracing according to claim 8, characterized in that, In step S7, when acquiring the cumulative number of differences and updating the data during the data optimization process based on the cumulative number of differences, the cumulative number of differences Ct is acquired through the system log. The cumulative number of differences Ct is compared with a preset cumulative number of differences Ct0. Based on the comparison result, the status of the cumulative number of differences is determined, and the data optimization process is updated based on the determination result. Wherein: When Ct≤Ct0, the cumulative number of differences is determined to be low, and no data update is performed during the data optimization process; When Ct > Ct0, the cumulative difference count is determined to be high. The data optimization process involves updating the data. The preset deviation count Pt0 is updated according to the data update coefficient gh to obtain the updated preset deviation count Pt1, where Pt1 = Pt0 × gh. The updated preset deviation count Pt1 is rounded to obtain the rounded preset deviation count Pt0'. The rounded preset deviation count Pt0' is output as the preset deviation count Pt0, and the deviation count Pt is compared with the preset deviation count Pt0 again.

10. The high-precision power system modeling method based on dynamic parameter tracing according to claim 9, characterized in that, In step S8, a digital twin model of the electrical system is generated based on multi-source data of the target power system and estimates of key dynamic parameters.

Images