Virtual power plant energy scheduling method and system under local data abnormal condition

Abstract

The invention discloses a virtual power plant energy scheduling method and system under a local data abnormal condition, and the method comprises the steps: recognizing a data type and an incidence relation related to abnormal data through an incidence matrix after the abnormal data of a virtual power plant are collected and recognized; and calculating an abnormal data correction value based on the association relationship function and the real-time association data, and performing weighted average on the abnormal data correction value and the original abnormal data to obtain correction data. And then calculating an abnormal data time difference and a correlation deviation degree, and distributing a credibility weight for the corrected data. And finally, calculating the correlation degree between the scheduling parameter of the preset scheduling scheme and the weighted correction value, and selecting the scheme with the highest correlation degree as a virtual power plant resource scheduling scheme. By implementing the technical scheme provided by the invention, the scheduling accuracy in a data exception state is improved.

Description

Technical Field

[0001] This application relates to the field of smart grids, and particularly to a virtual power plant energy scheduling method and system under local data anomaly conditions. Background Art

[0002] At present, with the rapid development of distributed energy technologies and the wide application of virtual power plants in modern energy systems, virtual power plants, as a technology that aggregates distributed energy, energy storage devices, and flexible loads for coordinated management, have become an important part of smart grids. By uniformly scheduling and optimizing the utilization of resources, virtual power plants can not only improve energy utilization efficiency but also effectively address the challenges brought by the volatility of renewable energy, playing an important role in promoting energy transformation and achieving the "dual carbon" goal. Therefore, ensuring the stability and accuracy of the virtual power plant scheduling system is the key to realizing its technical value.

[0003] In related technologies, the power generation data of distributed energy and user load data are usually monitored in real time, the supply-demand balance state is calculated based on the received real-time data, and a scheduling plan is further generated to allocate power generation resources and energy storage resources. For example, by accessing sensors to collect the power output data of distributed energy and combining the load demand data at the user end, the supply-demand difference of the current power system is analyzed, and then resource allocation and load adjustment are performed according to the analysis results. This method relies on the integrity and accuracy of real-time data to ensure that the scheduling plan can accurately reflect the current state of the power system.

[0004] However, in related technologies, when the power generation data of distributed energy or user load data are abnormal due to communication interruption, network delay, or interface failure, the system may generate an incorrect scheduling plan based on the incorrect data, resulting in the disruption of the supply-demand balance, possible scheduling interruption or unreasonable resource allocation, and reduced scheduling accuracy in the state of data anomaly, which will further lead to problems such as energy waste or insufficient power supply for critical loads. Summary of the Invention

[0005] This application provides a virtual power plant energy scheduling method and system under local data anomaly conditions, which is used to improve the scheduling accuracy in the state of data anomaly.

[0006] In the first aspect of this application, a virtual power plant energy scheduling method under local data anomaly conditions is provided. The method includes: Collect the real-time power generation data of distributed energy sources and the real-time user load data in the virtual power plant; determine whether there are abnormal data in the real-time power generation data or the real-time user load data; if so, identify the associated data type and association relationship related to the type of abnormal data according to the association matrix; obtain the real-time associated data corresponding to the associated data type; calculate the correction value of the abnormal data based on the association relationship function and the real-time associated data in the association relationship; perform a weighted average calculation on the correction value and the abnormal data to obtain the corrected data; calculate the time difference of the abnormal data and the association deviation degree, and assign a credibility weight to the corrected data; calculate the association degree between the scheduling parameters of each resource scheduling plan in the preset scheduling plan table and the weighted corrected value; sort all the association degrees to obtain the association degree ranking; use the resource scheduling plan with the highest ranking in the association degree ranking as the virtual power plant resource scheduling plan.

[0007] In the above embodiment, by introducing the association matrix and real-time associated data to calculate the correction value in the case of data anomalies, and assigning a credibility weight to the corrected data, the virtual power plant can still achieve accurate energy scheduling and resource allocation under data anomaly conditions. In this process, the association matrix is used to identify the data types and real-time data related to the abnormal data, and the abnormal data is corrected by combining weighted average and credibility assignment to ensure the reliability and accuracy of the data. Subsequently, by sorting the association degrees between the corrected data and the scheduling plans, the optimal scheduling plan is selected, enabling the system to avoid the interference of abnormal data and prevent resource waste and supply-demand imbalance caused by incorrect scheduling. It effectively improves the accuracy and stability of scheduling under the condition of data anomalies, ensures the stable power supply of critical loads, and significantly improves the energy utilization efficiency and the operation quality of the virtual power plant.

[0008] Combined with some embodiments of the first aspect, in some embodiments, calculating the correction value of the abnormal data based on the association relationship function and the real-time associated data in the association relationship specifically includes: Determine whether the abnormal time characteristic of the abnormal data is in a preset critical period; if so, extract the correlation change pattern of the same time characteristic as the abnormal time characteristic from the preset historical database, and establish a time-varying correlation coefficient set; according to the time-varying correlation coefficient set, perform a time-series correction on the static correlation coefficients in the association matrix to generate a dynamic association matrix adapted to the abnormal time characteristic; based on the dynamic association matrix and the associated data, construct a multiple linear regression equation system; solve the multiple linear regression equation system to obtain the correction value of the abnormal data.

[0009] In the above embodiments, by determining whether the abnormal time characteristics of the abnormal data are within a preset critical time period, extracting the correlation change pattern of the same type of time characteristics from the historical database, establishing a time-varying correlation coefficient set, and performing time-series correction on the static correlation coefficients in the correlation matrix to generate a dynamic correlation matrix, the correlation matrix can dynamically adapt to the time characteristics of the abnormal data and reflect the time-series correlation changes between the data. Based on the dynamic correlation matrix, a multiple linear regression equation system is constructed and the correction value is solved, so that the correction value can more accurately reflect the true state of the abnormal data. The limitations of the static correlation coefficient under complex time characteristics are effectively reduced, the accuracy of data correction is improved, it is ensured that the abnormal data can be reasonably corrected under different time conditions, the error accumulation caused by ignoring the time characteristics is avoided, and the accuracy of scheduling in the state of data abnormality is improved.

[0010] Combined with some embodiments of the first aspect, in some embodiments, after calculating the time difference and correlation deviation degree of the abnormal data and assigning credibility weights to the corrected data, it further includes: Quantitatively evaluate the credibility weights; according to the preset score interval threshold, divide the comprehensive credibility score into a high credibility interval, a medium credibility interval, and a low credibility interval; when the credibility weight is in the low credibility interval, select a backup data source to access according to the type of abnormal data; fuse the data of the backup data source with the corrected data to generate enhanced corrected data; quantitatively evaluate the enhanced corrected data; in the case where the enhanced corrected data is in the low credibility interval, increase the number of accessed backup data sources until the credibility weight is increased to the medium credibility interval or the backup data sources are exhausted.

[0011] In the above embodiments, in the case of low credibility, the data correction process can dynamically introduce more reliable data sources, gradually improve the credibility of the corrected data, solve the limitations of a single data correction method in the case of data abnormality or insufficient data sources, ensure that the system can still generate reliable corrected data under low credibility conditions, and avoid problems such as scheduling errors or unreasonable resource allocation caused by insufficient data quality. Through multiple rounds of fusion evaluation and incremental access of backup data, the accuracy and stability of the corrected data are ensured, high-quality data support is provided for the virtual power plant scheduling, and the anti-abnormality ability and operation reliability of the system are further improved.

[0012] Combined with some embodiments of the first aspect, in some embodiments, if so, then identify the associated data type and association relationship related to the type of abnormal data according to the correlation matrix, specifically including: Analyze the abnormal type and degree of abnormal data; set a dynamic search radius in the association matrix according to the degree of abnormality; extract the first-level association nodes from the association matrix according to the dynamic search radius; expand to the second-level association nodes when the scale of the first-level association nodes is less than the preset node scale threshold; assign an impact factor to each node in the effective association resource set based on resource type similarity, historical correlation strength and current operating status; measure the real-time measurement values of the nodes in the effective association resource set as real-time association data, and use them together with the association coefficient in the association matrix as the real-time association relationship.

[0013] In the above embodiment, by assigning influence factors to nodes based on the resource type similarity, historical correlation strength and current operating status of the nodes, and measuring the real-time measurement values of valid associated resources as real-time associated data, the association analysis can dynamically adapt to different abnormality levels and data structures, accurately capture the real-time association relationship related to the abnormal data, and ensure that the extracted associated data has higher relevance and timeliness, thereby providing a more scientific and reliable basis for abnormal data correction and improving the accuracy of scheduling under abnormal data conditions.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, after allocating an impact factor to each node in the valid associated resource set based on resource type similarity, historical correlation strength, and current operating status, the method further includes: Detect whether there is an abnormal node cluster in the effective associated resource set; if so, determine whether there is a regional abnormal event; if so, isolate the resources in the abnormal event area of the abnormal event from the virtual power plant scheduling range.

[0015] In the above embodiments, by identifying and isolating abnormal areas, we prevent erroneous data or resource failures caused by regional anomalies from interfering with normal scheduling, ensuring that the system can maintain supply and demand balance and operational reliability under complex fault conditions. By accurately locating abnormal areas and isolating abnormal resources, the risk resistance and emergency response capabilities of virtual power plants are significantly improved, providing a safe and stable operating environment for abnormal handling and scheduling optimization.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after the abnormal time area of the abnormal event is isolated from the virtual power plant scheduling scope, the method further includes: Mobilize backup power generation and energy storage resources from areas other than the abnormal event area; release the isolation state when it is detected that data communication in the abnormal event area has returned to normal or the abnormal event has ended.

[0017] In the above embodiments, by using backup resources to make up for the lack of resources in the abnormal area, and at the same time lifting the isolation state when it is detected that the data communication in the abnormal event area returns to normal or the event ends, the system can quickly restore the integrity and coordination of scheduling, ensuring the continuity of energy supply during abnormal events, effectively avoiding power supply interruption or large-scale system imbalance caused by the failure of resources in the abnormal area, and reducing the long-term impact of abnormal events on the overall operation efficiency and reliability, thereby ensuring the stable operation of the virtual power plant and the efficient utilization of energy resources.

[0018] In combination with some embodiments of the first aspect, in some embodiments, if, after analyzing the abnormal type and degree of the abnormal data, it further includes: When the abnormal data does not exist or the abnormal degree is mild, execute the resource scheduling plan at a preset first scheduling time interval; when the abnormal degree is moderate, execute the resource scheduling plan at a preset second scheduling time interval; when the abnormal degree is severe, execute the resource scheduling plan at a preset third scheduling time interval.

[0019] In the above embodiments, the response ability of the scheduling process to abnormal data is improved by setting time intervals, enabling the resource scheduling plan to more efficiently adapt to different degrees of abnormal situations. When the abnormal degree is relatively light, it reduces the waste of resources caused by frequent scheduling. At the same time, when the abnormal degree is relatively high, it improves the system's rapid response ability to abnormal changes through a more intensive scheduling frequency, avoiding problems such as resource allocation imbalance or supply-demand imbalance caused by delayed scheduling. This enables the system to maintain the accuracy and flexibility of scheduling under different abnormal conditions, effectively improving the operation efficiency and overall stability.

[0020] In a second aspect, an embodiment of the present application provides a virtual power plant energy scheduling system under local data abnormal conditions. The virtual power plant energy scheduling method under local data abnormal conditions includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, and the computer program code includes computer instructions. The one or more processors call the computer instructions to enable the virtual power plant energy scheduling system under local data abnormal conditions to execute the methods described in the first aspect and any possible implementation manner in the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer program product containing instructions. When the above computer program product runs on a virtual power plant energy scheduling system under local data abnormal conditions, it enables the virtual power plant energy scheduling system under local data abnormal conditions to execute the methods described in the first aspect and any possible implementation manner in the first aspect.

[0022] Fourthly, an embodiment of the present application provides a computer-readable storage medium, including instructions, which, when running on a virtual power plant energy scheduling system under local data anomaly conditions, cause the virtual power plant energy scheduling system under the local data anomaly conditions to execute the method described in the first aspect and any possible implementation manner in the first aspect.

[0023] It can be understood that the virtual power plant energy scheduling system under local data anomaly conditions provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the virtual power plant energy scheduling method under local data anomaly conditions provided by the embodiments of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, and will not be elaborated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. In the present application, by introducing an association matrix and calculating correction values for real-time associated data in the case of data anomalies, and assigning credibility weights to the corrected data, the virtual power plant can still achieve precise energy scheduling and resource allocation under data anomaly conditions. In this process, the association matrix is used to identify data types and real-time data related to the abnormal data, and the abnormal data is corrected by combining weighted average and credibility assignment to ensure the reliability and accuracy of the data. Subsequently, by sorting the correlation degrees between the corrected data and the scheduling scheme, the optimal scheduling scheme is selected, enabling the system to avoid interference from abnormal data and prevent resource waste and supply-demand imbalance caused by incorrect scheduling. This effectively improves the accuracy and stability of scheduling under data anomaly conditions, ensures stable power supply for critical loads, and significantly improves energy utilization efficiency and the operation quality of the virtual power plant.

[0025] 2. The present application determines whether the abnormal time characteristics of the abnormal data are in a preset critical period, extracts the correlation change pattern of the same time characteristics from the historical database, establishes a time-varying correlation coefficient set, and performs time-series correction on the static correlation coefficients in the association matrix to generate a dynamic association matrix, enabling the association matrix to dynamically adapt to the time characteristics of the abnormal data and reflect the time-series correlation changes between the data. Based on the dynamic association matrix, a multiple linear regression equation system is constructed and solved to obtain correction values, enabling the correction values to more accurately reflect the true state of the abnormal data. This effectively reduces the limitations of static correlation coefficients under complex time characteristics, improves the accuracy of data correction, ensures that abnormal data can be reasonably corrected under different time conditions, avoids error accumulation caused by ignoring time characteristics, and improves the accuracy of scheduling under data anomaly conditions.

[0026] 3. In this application, an influence factor is assigned to a node based on the resource type similarity, historical correlation strength, and current operating state of the node, and the real-time measurement value of the effectively associated resources is measured as real-time association data, enabling the association analysis to dynamically adapt to different degrees of anomalies and data structures, accurately capture the real-time association relationships related to the abnormal data, ensure that the extracted association data has higher relevance and timeliness, thereby providing a more scientific and reliable basis for the correction of abnormal data, and improving the accuracy of scheduling in the state of data anomalies. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 FIG. is a schematic flowchart of a virtual power plant energy scheduling method under local data anomaly conditions in an embodiment of this application; Figure 2 FIG. is another schematic flowchart of a virtual power plant energy scheduling method under local data anomaly conditions in an embodiment of this application; Figure 3 FIG. is an exemplary hardware structure schematic diagram of a virtual power plant energy scheduling system under local data anomaly conditions in an embodiment of this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] The terms used in the following embodiments of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. As used in the specification and appended claims of this application, the singular forms "a", "an", "the", "above", "said", "this" are also intended to include the plural forms unless the context clearly dictates otherwise. It should also be understood that the term "and / or" used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0029] Hereinafter, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as implying or suggesting relative importance or implicitly indicating the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the embodiments of this application, unless otherwise specified, the meaning of "a plurality" is two or more.

[0030] In the related art, the power generation data of distributed energy and the user load data are usually monitored in real time, and the supply-demand balance state is analyzed based on the received real-time data, and a scheduling plan is generated to allocate power generation resources and energy storage resources. This method relies on the integrity and accuracy of real-time data to ensure that the scheduling plan can accurately reflect the actual state of the power system. However, when the distributed energy power generation data or user load data is abnormal due to communication interruption, network delay or interface failure, the system may generate an incorrect scheduling plan based on the incorrect data, resulting in the destruction of the supply-demand balance, such as scheduling interruption, unreasonable resource allocation or insufficient power supply for critical loads, and ultimately leading to problems such as energy waste or reduced system operation efficiency.

[0031] In the embodiment of the present application, aiming at the problem of inaccurate scheduling caused by the above data abnormality, a virtual power plant energy scheduling method based on an association matrix and data correction is proposed. By collecting the real-time power generation data and user load data in the virtual power plant, it is first determined whether the data is abnormal; if an abnormality is detected, the data types and association relationships associated with the abnormal data are identified using the association matrix, and the correction value is calculated through the real-time associated data and the association relationship function to perform weighted correction on the abnormal data. At the same time, further combining the time difference of the abnormal data and the association deviation degree, a credibility weight is assigned to the corrected data to improve the reliability of the corrected data. Finally, by calculating the association degree between the corrected data and the scheduling plan, the optimal resource scheduling plan is selected. This method can efficiently correct abnormal data under the condition of data abnormality and optimize the scheduling plan, significantly improving the stability and accuracy of the virtual power plant scheduling, avoiding problems such as energy waste or insufficient power supply for critical loads caused by abnormal data, and providing a strong guarantee for the safe and efficient operation of the virtual power plant.

[0032] Figure 1 FIG. is a schematic flow chart of using the virtual power plant energy scheduling method under the condition of local data abnormality in the embodiment of the present application, including the following steps: S101. Collect the real-time power generation data of distributed energy and the real-time user load data in the virtual power plant; Specifically, the real-time power generation data of distributed energy (such as photovoltaic power generation, wind power generation, energy storage devices, etc.) is collected through deployed smart meters, sensors and data acquisition terminals. The smart meter is responsible for monitoring parameters such as the output power, voltage, and current of the power generation equipment, and the sensor can monitor environmental conditions (such as light intensity, wind speed, temperature, etc.) to assist in analyzing the power generation state. These devices are connected to the energy management system (EMS) in the virtual power plant energy scheduling system under the condition of local data abnormality, and the collected data is transmitted to the database of the virtual power plant energy scheduling system under the condition of local data abnormality through communication protocols (such as Modbus, IEC 61850, etc.).

[0033] For user load data, smart meters are installed on the user side to monitor in real time data such as the user's power consumption, power factor, and load change trend. These smart meters upload the data to the data center of the virtual power plant energy dispatching system under local data abnormal conditions through technologies such as local area network, wireless communication (such as LoRa, NB-IoT), or power line carrier communication (PLC).

[0034] The data center stores and preprocesses the collected real-time data (such as data formatting, noise filtering, and timestamp calibration) to ensure data integrity and consistency.

[0035] S102. Determine whether there is abnormal data in the real-time generation data or real-time user load data; If so, execute the following step S103; If not, return to execute the above step S101; Specifically, first, the collected data is preliminarily screened and cleaned. By removing obviously incorrect data points (such as negative values, missing values, or data with discontinuous timestamps), the basic validity of the input data is ensured. On this basis, the data is judged using a preset threshold range. The setting of the threshold range can be based on fixed upper and lower limits or dynamically adjusted according to time periods, weather conditions, etc.

[0036] For example, the fluctuation range of photovoltaic power generation data is smaller during the noon on sunny days, while a larger fluctuation is allowed during sunrise and sunset. In addition, statistical methods can also be used. By calculating key indicators such as the mean and standard deviation of the data, it is judged whether the current data deviates from the normal range. If there is abnormal data, execute the following step S103 to further process the abnormal data; if there is no abnormal data, return to execute S101 to continue collecting data.

[0037] S103. Identify the associated data types and association relationships related to the types of abnormal data according to the association matrix; Specifically, by establishing a mutual correlation model between distributed energy and user load data, the potential impact range or source of abnormal data is analyzed. The correlation matrix is a mathematical model used to describe the correlation strength between various data types within the system (such as photovoltaic power generation, wind power generation, energy storage status, user load, etc.). Each element in the matrix represents the degree of correlation between two types of data. First, the construction of the correlation matrix can be achieved through statistical analysis of historical data and physical modeling. Calculate the correlation coefficient between photovoltaic power generation and light intensity through time series correlation analysis, or calculate the relationship between wind speed and wind power generation through physical property modeling. After completing the correlation matrix, when a certain data is determined to be abnormal, the system will query the correlation matrix, identify the correlation between the abnormal data and other data types, and extract the real-time data of the relevant data types (such as light intensity, wind speed) and their correlation relationships.

[0038] The correlation relationship can be static (fixed correlation coefficient) or dynamic (real-time updated correlation coefficient). The static correlation matrix is applicable to scenarios where the relationship between data is stable, while the dynamic correlation matrix adjusts the correlation according to real-time environmental factors (such as time, weather conditions). In addition, in complex scenarios, the system can also adopt machine learning algorithms to mine non-linear correlation relationships through multi-dimensional data analysis and dynamically update the matrix.

[0039] S104. Obtain the real-time correlation data corresponding to the correlated data types; First, environmental variables, energy status, or user load are monitored in real time through collection devices such as sensors or smart meters. The light intensity of the photovoltaic power station is collected by light sensors, and the user load data is recorded by smart meters. The collected data is transmitted to the central management system of the virtual power plant through wireless communication (such as NB-IoT, LoRa) or wired communication (such as fiber optic communication). When abnormal data is identified, the system finds the data types related to the abnormal data according to the correlation matrix and issues a data request to obtain these correlation data in real time from the collection end.

[0040] During the data acquisition process, the system needs to perform time synchronization and integrity verification on the collected data. Time synchronization is usually achieved through GPS timestamps or the Network Time Protocol (NTP) to ensure that all data has a consistent time reference. During transmission, a verification algorithm (such as CRC verification) is used to verify the integrity of the data to avoid judgment deviation caused by data loss or errors.

[0041] In some embodiments of the present application, when a communication interruption or sensor failure occurs, the system obtains associated data from redundant data sources through a backup mechanism. These backup data sources can be sensor devices in neighboring areas, historical data models, or third-party services (such as meteorological services). When real-time data cannot be directly obtained, the system can combine the association matrix and the historical data model to calculate the value of the current associated data, ensuring the continuity and integrity of the data.

[0042] S105. Calculate a correction value for the abnormal data based on the association relationship function and the real-time associated data in the association relationship; Specifically, the association relationship function is a mathematical model that describes the relationship between data and can be a linear function, a non-linear function, or a prediction model based on machine learning.

[0043] First, the system determines the functional relationship between the abnormal data and the associated data type through the association matrix. Subsequently, the system substitutes the real-time associated data (such as light intensity, wind speed, etc.) into the association relationship function to calculate the theoretical value, and uses this theoretical value to replace or correct the abnormal data.

[0044] The implementation of the association relationship function can be divided into three methods: a physical model-based method, a statistical model-based method, and a machine learning-based non-linear model method. For data with clear physical laws, the association relationship function can be directly defined by physical formulas; for scenarios where the relationship is not clear but can be statistically fitted (such as the relationship between photovoltaic power generation and light intensity, temperature), an association function is established through a multiple regression model, and then the real-time associated data is substituted to calculate the correction value; for scenarios with complex non-linear relationships, a machine learning model (such as a neural network, support vector machine) can be used to train a prediction model in combination with historical data, and the correction value of the abnormal data is predicted by inputting the real-time associated data.

[0045] In some embodiments of the present application, the steps of calculating the correction value can be divided into the following steps: When the abnormal time characteristic of the abnormal data is in the preset critical period, extract the correlation change pattern of the same type of time characteristic as the abnormal time characteristic from the preset historical database, and establish a time-varying correlation coefficient set. When the abnormal time characteristic of the abnormal data is in the preset critical period (such as the twilight transition period of photovoltaic power generation, the weather system transition period of a wind farm, or the load peak-valley conversion period), due to the rapid change of environmental conditions, the correlation between data shows dynamic characteristics. The system extracts the correlation change pattern under the same time characteristic from the historical database and establishes a time-varying correlation coefficient set to correct the abnormal data. First, the system identifies whether the time characteristic of the abnormal data belongs to the critical period and extracts data under similar time characteristics or environmental conditions from the historical database. For example, during the photovoltaic twilight transition period, extract the light intensity and power generation data under the same conditions in history. Subsequently, the system uses a sliding window, time series model, or regression analysis to calculate the dynamic correlation between the associated data, form a time-varying correlation coefficient set, and dynamically adjust the association model to adapt to the complex changes in the critical period. Finally, based on the time-varying correlation coefficient set, correct the abnormal data. For example, reduce the dependence on light intensity during the photovoltaic twilight period, or combine the data of adjacent wind turbines to correct the power during the wind power weather system transition period, so that the system can accurately describe the data relationship during the critical period and improve the accuracy and stability of abnormal data correction.

[0046] According to the time-varying correlation coefficient set, perform temporal correction on the static correlation coefficients in the correlation matrix to generate a dynamic correlation matrix that adapts to the abnormal time characteristic. First, the system extracts the time-varying correlation coefficient set under similar conditions to the current abnormal time characteristic from the historical database. These time-varying correlation coefficients reflect the law of the correlation between data changing over time and are usually calculated through sliding window analysis, time series models (such as ARIMA, LSTM), or dynamic regression methods. Next, the system corrects the static correlation coefficients according to the time-varying correlation coefficients. The correction methods are divided into two types: one is the weighted update strategy, which combines the static correlation coefficients and the time-varying correlation coefficients according to the weight factor, and is applicable to scenarios where the static correlation still has certain reference value; the other is to directly replace the static correlation coefficients with the time-varying correlation coefficients, which is applicable to scenarios where the dynamic characteristics completely dominate. After the correction is completed, the system generates a dynamic correlation matrix, which can accurately reflect the dynamic correlation of the data during the current critical period.

[0047] During the twilight transition period of a photovoltaic power station, the correlation between photovoltaic power generation and light intensity will weaken due to the rapid change of the solar altitude angle. After the system identifies that the current period belongs to the twilight transition period, it extracts the time-varying correlation coefficient set of the twilight transition period from the historical database and performs weighted correction on the static correlation matrix to generate a dynamic correlation matrix, so that the corrected matrix can accurately describe the dynamic relationship between photovoltaic power and light intensity under the current conditions, thereby assisting in the correction of abnormal data.

[0048] Based on the dynamic correlation matrix and correlation data, a multiple linear regression equation system is constructed. The system uses the time-varying correlation coefficients in the correlation matrix as regression coefficients, and combines the correlation data to construct a multiple linear regression equation system. Through each group of time-varying correlation coefficients in the dynamic correlation matrix, the system dynamically adjusts the coefficients of the regression model, thereby updating the regression equation system in real time and being able to adapt to the dynamic changes of data over time.

[0049] Solve the multiple linear regression equation system to obtain the correction value for the abnormal data. Substitute the correlation data into the multiple linear regression model for solution. Common solution methods include direct least squares method (OLS, Ordinary Least Squares) or matrix solution method. If the error term of the equation system can be ignored (i.e., the model error is small), it can be directly solved in the form of a matrix equation. Through calculation, the theoretical value, that is, the predicted value of the abnormal data to be corrected, is obtained.

[0050] In some embodiments of the present application, if there is data missing or large error, the system can adopt a variety of optimization techniques. For example, adding regularization methods (such as ridge regression or Lasso regression) to prevent model overfitting, or adopting iterative optimization algorithms (such as gradient descent method) to improve the solution efficiency. When there are sensor failures or anomalies in the correlation data, the system can use historical correlation data or adjacent device data for supplementation to enhance the stability of the equation system.

[0051] S106. Perform weighted average calculation on the correction value and the abnormal data to obtain the corrected data; Specifically, the calculation formula for weighted average is: corrected data = first weight coefficient × correction value obtained by solving the multiple linear regression equation system + second weight coefficient × original abnormal data, and the weight coefficients satisfy the first weight coefficient + second weight coefficient = 1. The allocation of weights is dynamically adjusted according to the deviation degree of the abnormal data and the reliability of the correction model, so as to flexibly cope with different abnormal scenarios.

[0052] When the deviation of the abnormal data is small (such as only slight fluctuations), a higher weight can be assigned to the abnormal data (i.e., the second weight coefficient > the first weight coefficient) to retain more characteristics of the original data and prevent overcorrection. When the deviation of the abnormal data is large (such as significantly exceeding the normal range), the weight of the correction value needs to be increased (i.e., the first weight coefficient > the second weight coefficient) to ensure that the final corrected data is closer to the theoretical value. In addition, there are various ways to allocate weights: the static weight method is suitable for scenarios where the abnormal characteristics are relatively stable, and preset fixed weight values; the dynamic weight method adjusts the weights in real time according to the degree of deviation of the abnormal data from the normal range; the adaptive weight method uses machine learning algorithms to learn the optimal weights and is suitable for complex scenarios.

[0053] S107. Calculate the time difference of abnormal data and the associated deviation degree, and assign credibility weights to the corrected data; Specifically, the time difference of abnormal data is the time interval between the detection time of the abnormal data and the current processing time, and the associated deviation degree is the ratio of the difference between the abnormal data and the corrected value to the abnormal data.

[0054] The time difference refers to the time interval between the detection time of the abnormal data and the current processing time. The larger the time difference, the lower the timeliness of the abnormal data and the smaller the reference value. Therefore, the weight in the corrected data should be reduced. The associated deviation degree is the relative deviation between the abnormal data and the corrected value. The larger the deviation degree, the greater the difference between the abnormal data and the corrected value, and the lower the credibility of the abnormal data. The weight of the corrected value should be increased accordingly. Finally, the system dynamically adjusts the weight allocation of the abnormal data and the corrected value using the time difference and the associated deviation degree to ensure the reliability of the corrected data.

[0055] The weight allocation of the time difference can adopt a linear method (the weight decreases uniformly with the time difference), a non-linear method (such as exponential decay of the weight), or the weight can be automatically adjusted by machine learning methods. When the deviation degree is small, the weight of the abnormal data is large; while when the deviation degree is large or the time difference is large, the weight of the corrected value is large to improve the credibility of the corrected data.

[0056] In some embodiments of the present application, after the above steps, it is also possible to perform a quantitative evaluation of the credibility weight. If the credibility is relatively low, steps of adding a backup data source for fusion processing of the corrected data to enhance the credibility can be performed.

[0057] Perform a quantitative evaluation of the credibility weight. Performing a quantitative evaluation of the credibility weight is a process of converting the time difference of the abnormal data into a time decay coefficient, converting the associated deviation degree into a deviation coefficient, and calculating a comprehensive credibility score by combining weighted products, thereby clearly dividing the credibility of the abnormal data and the corrected value. The time decay coefficient is used to quantify the impact of the time difference on the credibility, the deviation coefficient is used to quantify the impact of the associated deviation degree on the credibility, and the comprehensive credibility score is calculated by the weighted product of the time decay coefficient and the deviation coefficient.

[0058] According to the preset score interval threshold, the comprehensive credibility score is divided into a high credibility interval, a medium credibility interval, and a low credibility interval. When the credibility weight is in the low credibility interval, a backup data source is selected for access according to the type of abnormal data. First, the system identifies the type of abnormal data and dynamically selects a suitable backup data source for supplementation according to the specific type of abnormal data. The backup data sources mainly include the associated resource data of adjacent area virtual power plants, the regional power grid dispatching information, and the third-party meteorological service data. The associated resource data of adjacent area virtual power plants is applicable to scenarios with strong spatial proximity, such as accessing the power generation data of adjacent photovoltaic power plants or wind farms as a reference; the regional power grid dispatching information is applicable to abnormal data related to load demand or power distribution and can provide dispatching plans or load prediction data; the third-party meteorological service data is applicable to abnormal data closely related to meteorological conditions, such as using real-time illumination intensity, wind speed, and temperature data to correct the abnormal values of photovoltaic or wind power.

[0059] The data of the backup data source and the corrected data are fused to generate enhanced corrected data. First, the system preprocesses the backup data source and the corrected data to ensure consistency in time scale, data format, and dimension. For example, the backup data is standardized through interpolation or complementation methods. Then, the system assigns weights to the backup data and the corrected data, and the weights are determined by the credibility of the two. When the credibility of the backup data is high, the weight of the backup data dominates; when the corrected data is more reliable, the weight of the corrected data is appropriately increased.

[0060] In the data fusion process, a suitable fusion method can be selected according to the actual scenario. The simple weighted average method is applicable to scenarios where the backup data and the corrected data have a strong correlation; the Kalman filter method is applicable to time series data fusion and can dynamically balance the data weights and smooth the noise; the Bayesian update method is applicable to scenarios with high uncertainty and can generate more reliable results by dynamically adjusting the weights; for multi-dimensional complex scenarios, machine learning methods (such as neural networks or random forests) can be used to train a model through historical data to predict the optimal fusion result.

[0061] The enhanced corrected data is quantitatively evaluated. The quantitative evaluation of the enhanced corrected data includes three main links: deviation analysis, confidence calculation, and consistency verification. First, the error between the enhanced corrected data and the actual observed value or the high-credibility reference value is quantified through deviation analysis. Second, confidence calculation is performed to evaluate the reliability of the enhanced corrected data. Usually, combined with the deviation calculation result, the higher the confidence, the stronger the credibility of the data. Finally, the relevance between the enhanced corrected data and other related data (such as adjacent area data or time series data) is checked through consistency verification.

[0062] When the enhanced correction data is in the low confidence interval, increase the number of accesses to the backup data sources until the confidence weight is increased to the medium confidence interval or the backup data sources are exhausted. First, prioritize the backup data sources and set the access order according to the relevance, reliability, or timeliness of the data. The system sequentially accesses the backup data sources with higher priorities, gradually integrates the newly accessed data, and recalculates the confidence weight of the enhanced correction data with the existing correction data. The improvement standard of the confidence weight can be achieved through a dynamic evaluation formula. When the confidence reaches the medium confidence interval or all backup data sources have been accessed, stop the increment operation.

[0063] S108. Calculate the scheme correlation degree between the scheduling parameters of each resource scheduling scheme in the preset scheduling scheme table and the weighted correction value; Specifically, first, the weighted correction value is calculated from the correction data and the confidence weight. Specifically, the confidence weight is allocated according to the time difference and correlation deviation degree of the abnormal data. The time difference reflects the timeliness of the abnormal data, while the correlation deviation degree reflects the difference between the abnormal data and the correction value. Under the weighted effect of the confidence weight, the correction data obtains a weighted correction value that comprehensively considers the correction accuracy and reliability. Second, data preprocessing needs to be performed on the scheduling parameters and the weighted correction value. By standardization, the data is mapped to the same dimension range to ensure the comparability between different parameters. Then, according to the matching degree between the scheduling parameters and the weighted correction value, the deviation method, correlation coefficient method, or weighted similarity method is used to calculate the scheme correlation degree.

[0064] The deviation method evaluates the difference by calculating the mean absolute percentage error (MAPE) between the scheduling parameters and the weighted correction value. The smaller the error, the higher the correlation degree; while the correlation coefficient method measures the linear relationship between the two by calculating the Pearson correlation coefficient. The closer the correlation coefficient is to 1, the higher the correlation degree. In addition, the weighted similarity method calculates the similarity between the scheme and the correction value by assigning weights to different parameters and using the weighted Euclidean distance or weighted cosine similarity.

[0065] S109. Sort all the scheme correlation degrees to obtain the scheme correlation degree ranking; Specifically, first, the system extracts the correlation degree scores of each scheduling scheme and organizes them into a unified list or array to ensure consistent data formats. Then, select a suitable sorting algorithm according to the data volume and scenario requirements. For scenarios with a small number of schemes, basic algorithms such as bubble sort or insertion sort can be used; while in scenarios with a large number of schemes, advanced algorithms such as quick sort or merge sort can be selected to improve the sorting efficiency. During the sorting process, the system arranges the schemes in descending order of the correlation degree, ensuring that the scheme with the highest score is ranked first. If there are multiple scheme correlation degree scores that are the same, secondary indicators (such as scheduling cost, resource utilization rate, or response time) are introduced for multi-dimensional sorting to further refine the priorities.

[0066] S110. Use the resource scheduling plan with the highest ranking in the plan correlation degree ranking as the virtual power plant resource scheduling plan.

[0067] Specifically, after completing the calculation and ranking of all plan correlation degrees, the system extracts the plan with the highest ranking from the ranking results, and this plan is preferentially selected as the final scheduling plan of the virtual power plant.

[0068] If there are multiple plans with the same correlation degree score, the system can introduce secondary indicators (such as scheduling cost, response time, or resource utilization rate) for further screening. For example, in the case of the same correlation degree, the plan with the lowest scheduling cost can be preferentially selected as the final plan. If the secondary indicators still cannot distinguish, a suitable plan can be selected through manual review or based on historical data experience. In addition, in some scenarios with high real-time requirements or large data fluctuations, the system may dynamically adjust the scheduling parameters of the plan with the highest ranking, by fine-tuning power output or energy storage allocation parameters to make it more in line with actual requirements, thereby further optimizing the scheduling effect.

[0069] In the above embodiment, by introducing a correlation matrix and real-time correlation data to calculate the correction value in the case of data anomalies, and assigning credibility weights to the corrected data, the virtual power plant can still achieve accurate energy scheduling and resource allocation under abnormal data conditions. In this process, the correlation matrix is used to identify the data types and real-time data related to the abnormal data, and the abnormal data is corrected by combining weighted average and credibility assignment to ensure the reliability and accuracy of the data. Subsequently, through the correlation degree ranking of the corrected data and the scheduling plan, the optimal scheduling plan is selected, enabling the system to avoid the interference of abnormal data and prevent resource waste and supply-demand imbalance caused by incorrect scheduling. It effectively improves the accuracy and stability of scheduling in the state of data anomalies, ensures the stable power supply of critical loads, and significantly improves the energy utilization efficiency and the operation quality of the virtual power plant.

[0070] In some other embodiments of the present application, during the operation of a virtual power plant, the real-time collected distributed energy generation data or user load data may have abnormal data due to environmental changes, equipment failures, or communication anomalies. These abnormal data may lead to unbalanced resource scheduling or a decrease in system operation efficiency. By using the virtual power plant energy scheduling method under local data anomaly conditions provided in the present application, when abnormal data is detected, the search radius of the correlation matrix can be dynamically adjusted by analyzing the type and degree of the anomaly, associated nodes can be flexibly extracted, and influence factors can be assigned to effective associated resources based on resource type similarity, historical correlation, and current operating status. Further, a real-time association relationship is established by combining real-time measurement values and correlation coefficients, providing accurate association support for subsequent correction of abnormal data, thereby effectively improving the reliability and accuracy of data correction, and ensuring the stable operation of the system and the high efficiency of resource scheduling under abnormal conditions.

[0071] As Figure 2 shown, it is another schematic flowchart of the virtual power plant energy scheduling method under local data anomaly conditions provided by the embodiments of the present application, including the following steps: S201. Collect the real-time generation data of distributed energy and the real-time user load data in the virtual power plant; S202. Determine whether there is abnormal data in the real-time generation data or the real-time user load data; If so, execute the following step S203; If not, return to execute the above step S201; S203. Analyze the type and degree of the anomaly of the abnormal data; Specifically, after identifying the abnormal data, classify the type of the anomaly. An outlier refers to an individual data point deviating from the normal value range; a trend anomaly usually shows that the overall trend of the data deviates from the expectation, such as detecting the trend of the data using time series analysis; a structural anomaly shows a change in the overall form of the data distribution. Then, evaluate the degree of the anomaly according to the deviation degree of the abnormal data from the normal range. A mild anomaly is usually that the data slightly deviates from the normal range but does not affect the overall analysis result; a moderate anomaly may have a certain impact on the analysis result; and a severe anomaly may lead to the distortion of the analysis result or abnormal system operation. Common methods include calculating the percentage of the deviation value or the relative error, and setting a threshold interval.

[0072] In some embodiments of the present application, the resource scheduling plan will be adjusted according to the degree of the anomaly of the abnormal data.

[0073] When there is no abnormal data or the abnormality is mild, the resource scheduling plan is executed at the preset first scheduling time interval. First, the input data is analyzed by the data anomaly detection module to determine the type and severity of the anomaly. If the test result shows that there is no abnormality in the data, or the abnormality is mild, the system enters the normal scheduling mode and performs resource scheduling according to the preset scheduling cycle. At this time, the preset first scheduling time interval (such as 5 minutes or 15 minutes) is used as the basic time unit of the system scheduling plan. The scheduling module will allocate resources according to the established scheduling plan and perform specific operations such as power control and load distribution.

[0074] When the degree of abnormality is moderate, the resource scheduling plan is executed at the preset second scheduling time interval. First, the data is analyzed through the abnormality detection module to confirm that the degree of abnormality is moderate. For example, the error of the data deviation from the normal value is between 10% and 30%, or the power fluctuation exceeds the load stability range but does not pose a serious threat to the operation of the system. After confirming the degree of abnormality, the system automatically switches to the preset second scheduling time interval (such as shortening from 15 minutes to 5 minutes or 10 minutes), and re-executes the resource scheduling plan based on this. Shortening the scheduling time interval allows the system to adjust resource allocation more frequently, such as optimizing load distribution or energy storage charging and discharging operations in real time to cope with larger fluctuations that may be caused by moderate abnormalities.

[0075] When the abnormality is severe, the resource scheduling plan is executed at the preset third scheduling time interval. The abnormality detection module determines that the abnormality is severe (such as the data deviates from the normal value by more than 30%, or the system load fluctuation seriously affects the power balance), and the system automatically switches to the preset third scheduling time interval (such as 30 minutes or 1 hour) to reduce the scheduling frequency and reduce system pressure. Unlike moderate abnormalities, frequent scheduling in severe abnormalities may not effectively alleviate the problem, but may aggravate the instability of the system. Therefore, extending the scheduling time interval can buy more adjustment time for the system, and improve the reliability of scheduling by implementing more comprehensive abnormal correction measures and scheduling optimization strategies. For example, the system can concentrate resources to prioritize abnormal areas, or restore the system to a stable state through large-scale load reduction, energy storage rescheduling and other means.

[0076] S204, setting a dynamic search radius in the association matrix according to the degree of abnormality; Specifically, first, the search radius is dynamically set according to the degree of anomaly: when the degree of anomaly is relatively mild, the system sets a smaller search radius and focuses on high-precision correlation search in the local area; when the degree of anomaly is moderate or severe, the search radius is gradually expanded to cover a larger data range, thereby capturing more potential correlation relationships and adapting to the interference of abnormal data on the overall distribution of the system.

[0077] The setting of the dynamic search radius can be achieved through a distance metric formula or a weight adjustment strategy. For example, the Euclidean distance threshold or similarity weight of data points in the association matrix can be adjusted according to the degree of anomaly. In the case of mild anomalies, a smaller distance threshold is set to search only for data points with a very high degree of association; in the case of moderate anomalies, the distance threshold is expanded to accommodate more potentially relevant points; in the case of severe anomalies, the search radius is further expanded, and a correction algorithm is combined to process low-correlation data.

[0078] S205. Extract first-level associated nodes from the association matrix according to the dynamic search radius; Specifically, first, in the association matrix, each row or column represents the association relationship between a node and other nodes, and its value is usually the degree of association (such as a correlation coefficient or similarity). The dynamic search radius is set according to the degree of anomaly, and the threshold range of the degree of association is defined. In the case of mild anomalies, a smaller search radius is set to extract only highly correlated nodes; in the case of moderate and severe anomalies, the search radius is expanded to allow the extraction of moderately associated nodes to cope with anomaly interference. Then, for the target node, all nodes with a degree of association within the search radius are screened out from the association matrix, sorted in descending order of the degree of association, and the node with the highest degree of association is extracted as the first-level associated node set.

[0079] S206. When the scale of the first-level associated nodes is smaller than the preset node scale threshold, expand to the second-level associated nodes; Specifically, first, calculate the scale of the first-level associated nodes through the association matrix and compare it with the preset threshold. If the number of first-level associated nodes is less than the threshold, the expansion mechanism is triggered, and the associated node sets of the first-level associated nodes (i.e., the second-level associated nodes) are searched in turn. When expanding, nodes are filtered according to the dynamic search radius or the set degree-of-association threshold, and only the nodes that meet the conditions are retained in the second-level associated node set. After the expansion is completed, the scale of the associated nodes is re-counted. If it is still less than the threshold, the expansion can continue iteratively to deeper levels of associated nodes, or the maximum depth of the expansion can be limited according to system requirements. To optimize the expansion result, the system also sorts the expanded node set and preferentially retains nodes with high importance according to the weight or degree of association of the nodes to ensure that the expanded node set has a high analysis value.

[0080] S207. Assign influence factors to each node in the effective associated resource set based on resource type similarity, historical correlation strength, and the current operating state; Specifically, when allocating the impact factor, the system normalizes three indicators: resource type similarity, historical correlation strength, and current operating state, and comprehensively calculates the impact factor through a weighted model. In case of anomalies, the weight of the current operating state can be increased and the weight of historical correlation can be decreased, so as to dynamically adapt to real-time changes. Finally, the system allocates an impact factor to each node in the effective associated resource set and stores it in the association matrix or weight vector, providing an accurate basis for subsequent optimal scheduling.

[0081] In some embodiments of the present application, when it is detected that there is an abnormal node cluster in the effective associated resource set, it is determined whether there is a regional abnormal event. First, through real-time monitoring of the effective associated resource set and abnormal detection algorithms, nodes with abnormal operating states are marked, such as abnormal power output, communication delay, excessive load, or equipment failure. Subsequently, the system further analyzes these abnormal nodes to identify whether there is an abnormal node cluster. The definition of an abnormal node cluster needs to meet the following conditions: the geographical locations of the nodes are relatively concentrated (such as in the same substation, the same area, or within a communication network), the functional attributes are similar (such as the same type of equipment or the same operating task), and the occurrence of anomalies has time correlation (such as anomalies occurring synchronously within a short period of time). Once an abnormal node cluster is determined, the system will further determine whether it is caused by a regional abnormal event. The basis for judging a regional abnormal event includes geographical distribution characteristics, time synchronization, anomaly type consistency, and external event verification, etc. Geographical distribution characteristics refer to whether the abnormal node cluster is concentrated in the same area. If the geographical distance between nodes is relatively close, it is more likely to be caused by a regional event. Time synchronization refers to whether the anomalies occur simultaneously or within a short period of time. High time correlation is also an important feature of a regional event. Anomaly type consistency refers to whether the manifestations of abnormal nodes are similar, such as all being communication interruptions, equipment failures, or output fluctuations. In addition, by verifying whether events such as extreme weather, communication interruptions, or power grid failures have occurred in the region through external data sources (such as weather data, communication network status, power grid operation records, etc.), the existence of a regional abnormal event can be further confirmed.

[0082] In the presence of regional abnormal events, isolate the resources in the abnormal event area of the abnormal event from the dispatching scope of the virtual power plant. First, through real-time monitoring and abnormal detection algorithms, identify the scope of the abnormal area. The determination of the abnormal area is based on the geographical distribution, functional attributes of the affected nodes, and external event verification (such as weather warnings, communication network status, etc.). Usually, the resource nodes in the abnormal area exhibit abnormal behaviors such as power output fluctuations, communication interruptions, or equipment failures, and their geographical locations are close or their functional attributes are similar. Then, the system analyzes the status of the abnormal nodes to determine whether isolation is required. The isolation conditions include the unavailability of resource nodes (such as communication failure or power output abnormality exceeding the threshold), whether the abnormal event affects the resources in other areas, and whether the continued participation of the abnormal nodes in dispatching will increase the system dispatching risk. When the isolation conditions are met, perform isolation operations on the resources in the abnormal area. Specifically, it includes stopping the transmission of dispatching instructions for abnormal resources to avoid dispatching failures caused by communication interruptions or equipment failures; removing the resources in the abnormal area from the dispatching optimization model and recalculating the dispatching plan of the virtual power plant to ensure the efficient operation of the remaining normal resources; marking and continuously monitoring the resources in the abnormal area, and gradually restoring them to the dispatching scope after the abnormal event is lifted. In addition, to cope with the resource gap caused by isolation, the system calls standby resources, adjusts the load distribution in other areas, or activates energy storage systems through a dynamic dispatching adjustment mechanism to ensure the supply-demand balance of the overall system.

[0083] Dispatch standby power generation and energy storage resources from areas other than the abnormal event area. First, the system evaluates the resource status in the abnormal area through real-time monitoring and prediction models to determine the scale of resource failure and the power gap. This gap is usually represented by the difference between the available power and the actual demand, and combines the load prediction model to evaluate the demand change trend during the abnormal state. Subsequently, screen available standby resources from other areas except the abnormal area, including standby power generation equipment (such as gas turbines or distributed photovoltaic power generation), energy storage systems (such as battery energy storage), and flexible load resources (such as load shedding schemes). The screening process is based on the power availability, response speed, current status of the resources, and the feasibility of the dispatching path (such as the load condition of the transmission line). After the screening of standby resources is completed, adopt a dynamic dispatching strategy to determine the optimal resource combination. The dispatching strategy is usually based on optimization algorithms (such as linear programming or genetic algorithms), and the goal is to minimize the dispatching cost, reduce power fluctuations, and meet the supply-demand balance requirements as much as possible. At the same time, consider the response time and transmission loss of standby resources to ensure fast and efficient dispatching. After determining the dispatching plan, the system sends a call instruction to the standby resources to start the power output of the power generation equipment or release the energy storage device. In addition, the system may combine demand response technologies to further reduce the load pressure by reducing the electricity demand of non-critical loads.

[0084] In the process of mobilizing backup resources, the power balance status, backup resource operation status and the development of abnormal events are continuously monitored. If the scope of the abnormal event expands or the backup resources are insufficient, the system will optimize the scheduling plan in real time, call other available resources or upgrade the emergency response level.

[0085] When it is detected that the data communication in the abnormal event area has returned to normal or the abnormal event has ended, the isolation state is lifted. First, the system monitors the communication status, equipment operation and external environment in the abnormal area in real time to confirm that the communication has returned to normal (such as network delay and packet loss rate return to normal range), the equipment status is stable (such as power output and load are normal), and the triggering factors of the abnormal event (such as extreme weather or power grid failure) have been lifted. Subsequently, the system determines whether the conditions for lifting the isolation are met, and verifies the status of each resource node in the area one by one, including equipment self-test, power output test, and scheduling instruction responsiveness test. After the verification is completed, the system updates the scheduling model, re-includes the resources that have returned to normal in the scheduling range, and issues scheduling instructions to these resources. At the same time, it dynamically adjusts the allocation of standby resources in other areas to optimize the overall scheduling plan.

[0086] S208, measuring the real-time measurement value of the node in the effective association resource set as real-time association data, and using it together with the association coefficient in the association matrix as the real-time association relationship; Specifically, first, the system collects real-time operating data of nodes in the effective associated resource concentration through monitoring equipment, such as power output, load level, voltage and frequency, etc. These data can reflect the operating status of the nodes in real time. Then, the association coefficient stored in the association matrix is used as a static characteristic to represent the historical association strength between nodes. These association coefficients are usually calculated through statistical analysis of historical data, time series modeling or graph network analysis, and can reflect the synergy, correlation or coupling relationship between nodes. Subsequently, the real-time measurement value is combined with the association coefficient. Common methods include weighted models (such as linear weighting) or prediction models based on machine learning. The choice of weights can be dynamically adjusted according to the volatility of real-time data or system requirements.

[0087] This calculation method can be implemented in a variety of ways: for scenarios with small fluctuations in real-time data, a fixed weighting method can be used to linearly combine historical correlation coefficients and real-time data; for complex scenarios with large fluctuations in real-time data, a dynamic weighting strategy can be used, or a machine learning model (such as a neural network or regression model) can be used to capture more subtle correlation changes. In addition, in cases where data uncertainty is high, fuzzy logic methods can be used to fuzzify the uncertainty of real-time measurements and correlation coefficients, and then calculate real-time correlation relationships based on fuzzy reasoning.

[0088] S209, obtaining real-time associated data corresponding to the associated data type; S210. Calculate the correction value of the abnormal data based on the correlation function in the association relationship and the real-time association data; S211. Perform a weighted average calculation on the correction value and the abnormal data to obtain the corrected data; S212. Calculate the time difference of the abnormal data and the association deviation degree, and assign a credibility weight to the corrected data; S213. Calculate the association degree between the scheduling parameters of each resource scheduling plan in the preset scheduling plan table and the weighted correction value; S214. Sort all the association degrees of the plans to obtain the sorted association degree of the plans; S215. Use the resource scheduling plan with the highest ranking in the sorted association degree of the plans as the virtual power plant resource scheduling plan.

[0089] Steps S201 - S202, S209 - S215 are similar to Figure 1 Steps S101 - S102, S104 - S115 in the embodiment shown, and reference can be made to the descriptions in Steps S101 - S102, S104 - S115, which will not be elaborated here.

[0090] In the above embodiment, the association range and node scale are dynamically adjusted, enabling the system to flexibly obtain resource information highly relevant to the abnormal data for different degrees of abnormality, ensuring a more accurate and comprehensive data correction basis. By introducing real-time association data and historical association relationships, not only the depth and accuracy of abnormal data analysis are improved, but also the real-time response ability of the virtual power plant under complex abnormal conditions is effectively enhanced, providing reliable data support for subsequent data correction and scheduling optimization. This processing method significantly reduces the negative impact of abnormal data on the overall operation of the system, and improves the stability and scheduling accuracy of the virtual power plant operation.

[0091] Next, the virtual power plant energy scheduling system 300 under the exemplary local data abnormality condition provided by the embodiment of the present application is introduced. Figure 3 It is an exemplary hardware structure diagram of the virtual power plant energy scheduling system 300 under the local data abnormality condition provided by the embodiment of the present application.

[0092] In some embodiments, the virtual power plant energy dispatching system 300 under the condition of local data anomaly is a computer device or the virtual power plant energy dispatching system 300 under the condition of local data anomaly includes a computer device. The computer device includes a processor, a memory, and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data. The network interface of the computer device is used to communicate with other external terminals or servers through a network connection. In some embodiments, the network interface can be a wired network interface, and in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, it implements the method in the embodiments of the present application.

[0093] Those skilled in the art can understand that Figure 3 the structure shown in is only a block diagram of some structures related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0094] As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

[0095] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as "if..." or "after..." or "in response to determining..." or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if detecting (the stated condition or event)" can be interpreted as "if determining..." or "in response to determining..." or "when detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)".

[0096] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as coaxial cable, optical fiber, digital subscriber line) or wirelessly (such as infrared, wireless, microwave, etc.). The computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that includes one or more integrated available media. The available medium may be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid-state drive), etc.

[0097] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware with a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it may include the processes of the above method embodiments. The foregoing storage medium includes various media that can store program codes, such as ROM or random access memory RAM, magnetic disks, or optical discs.

Claims

1. A virtual power plant energy scheduling method under local data anomaly conditions, characterized in that Including: Collecting the real-time power generation data of distributed energy and the real-time user load data in the virtual power plant; Judging whether there is abnormal data in the real-time power generation data or the real-time user load data; If so, identifying the associated data type and the association relationship associated with the type of the abnormal data according to the association matrix; Obtaining the real-time associated data corresponding to the associated data type; Calculating a correction value of the abnormal data based on the association relationship function in the association relationship and the real-time associated data; Performing a weighted average calculation on the correction value and the abnormal data to obtain corrected data; Calculating the abnormal data time difference and the association deviation degree, and assigning a credibility weight to the corrected data; the abnormal data time difference is the time interval between the detection time of the abnormal data and the current processing time; The association deviation degree is the ratio of the difference between the abnormal data and the correction value to the abnormal data; Calculating the scheme association degree between the scheduling parameters of each resource scheduling scheme in the preset scheduling scheme table and the weighted correction value, where the weighted correction value is calculated by the corrected data and the credibility weight; Sorting all the scheme association degrees to obtain a scheme association degree sorting; Taking the resource scheduling scheme with the highest sorting in the scheme association degree sorting as the virtual power plant resource scheduling scheme.

2. The method according to claim 1, characterized in that, The calculating a correction value of the abnormal data based on the association relationship function in the association relationship and the real-time associated data specifically includes: Judging whether the abnormal time characteristic of the abnormal data is in a preset critical period; the critical period includes the dawn-dusk transition period of photovoltaic power generation, the weather system transition period of the wind farm, and the load peak-valley conversion period; If so, extracting the correlation change pattern of the same type of time characteristic as the abnormal time characteristic from the preset historical database to establish a time-varying correlation coefficient set; Performing a time series correction on the static correlation coefficient in the association matrix according to the time-varying correlation coefficient set to generate a dynamic association matrix adapted to the abnormal time characteristic; Constructing a multiple linear regression equation set based on the dynamic association matrix and the associated data; the regression coefficients of the multiple linear regression equation set are determined by the time-varying correlation coefficients of the dynamic association matrix; Solving the multiple linear regression equation set to obtain a correction value of the abnormal data.

3. The method according to claim 1, wherein After calculating the abnormal data time difference and the association deviation degree and assigning a credibility weight to the corrected data, it further includes: Performing a quantitative evaluation on the credibility weight; the quantitative evaluation is to convert the abnormal data time difference into a time decay coefficient, convert the association deviation degree into a deviation coefficient, and calculate a comprehensive credibility score through weighted product; Dividing the comprehensive credibility score into a high credibility interval, a medium credibility interval, and a low credibility interval according to a preset score interval threshold; When the credibility weight is in the low credibility interval, selecting a backup data source to access according to the type of the abnormal data; the backup data source includes the associated resource data of the adjacent area virtual power plant, the regional power grid dispatching information, and the third-party meteorological service data; Performing a fusion process on the data of the backup data source and the corrected data to generate enhanced corrected data; Performing quantitative evaluation on the enhanced correction data; When the enhanced correction data is in a low credibility interval, the number of accesses to the backup data source is increased incrementally until the credibility weight is increased to a medium credibility interval or the backup data source is exhausted.

4. The method according to claim 1, characterized in that The step of identifying the associated data types and associated relationships associated with the types of the abnormal data according to the association matrix specifically includes: Analyze the abnormal type and abnormal degree of the abnormal data; the abnormal degree includes mild, moderate and severe; According to the abnormality level, a dynamic search radius is set in the association matrix; Extracting first-level association nodes from the association matrix according to the dynamic search radius; When the scale of the first-level associated node is smaller than a preset node scale threshold, the nodes are expanded to the second-level associated nodes; the second-level associated nodes are nodes other than the first node that are associated with the first-level associated node; Based on resource type similarity, historical correlation strength and current operation status, an impact factor is assigned to each node in a valid associated resource set; the valid associated resource set includes the first-level associated node and the second-level associated node; The real-time measurement values of the nodes in the effective association resource set are measured as real-time association data, and are used together with the association coefficients in the association matrix as real-time association relationships.

5. The method according to claim 4, wherein After allocating the impact factor to each node in the valid associated resource set based on resource type similarity, historical correlation strength and current operation status, the method further includes: Detecting whether there is an abnormal node cluster in the effective associated resource set; the abnormal node cluster is a plurality of resource nodes with similar geographical locations or functions that have abnormalities at the same time; If yes, determine whether there is a regional abnormal event; the regional abnormal event includes a local communication network interruption, a regional extreme weather event or a power grid failure; If so, the resources in the abnormal event area of the abnormal event are isolated from the virtual power plant scheduling scope.

6. The method according to claim 5, characterized in that After isolating the resources in the abnormal time zone of the abnormal event from the virtual power plant scheduling scope, the method further includes: mobilizing backup generation and storage resources from areas other than the area of the abnormal event; When it is detected that data communication in the abnormal event area returns to normal or the abnormal event ends, the isolation state is released.

7. The method according to claim 4, characterized in that After analyzing the abnormal type and abnormal degree of the abnormal data, the method further includes: When the abnormal data does not exist or the abnormality is mild, executing the resource scheduling plan at a preset first scheduling time interval; When the abnormality level is moderate, executing the resource scheduling plan at a preset second scheduling time interval; When the abnormality level is severe, the resource scheduling plan is executed at a preset third scheduling time interval; the third scheduling time interval is greater than the second scheduling time interval, and the second scheduling time interval is greater than the first scheduling time interval.

8. A virtual power plant energy scheduling system under local data anomaly conditions, characterized in that, The virtual power plant energy scheduling system under the condition of local data anomaly includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the virtual power plant energy scheduling system under the condition of local data anomaly to execute the method described in any one of claims 1-7.

9. A computer program product comprising instructions, characterized in that, When the computer program product runs on the virtual power plant energy scheduling system under the condition of local data anomaly, it enables the virtual power plant energy scheduling system under the condition of local data anomaly to execute the method described in any one of claims 1-7.

10. A computer-readable storage medium, comprising instructions, characterized in that, When the instruction runs on the virtual power plant energy scheduling system under the condition of local data anomaly, it enables the virtual power plant energy scheduling system under the condition of local data anomaly to execute the method described in any one of claims 1-7.

Images