A virtual power plant anomaly detection system and its anomaly detection method

Abstract

The present invention discloses a virtual power plant anomaly detection system and an anomaly detection method thereof, relating to the technical field of anomaly detection. By calculating the similarity coefficient between the energy storage devices of every two virtual power plants, and grouping the energy storage devices in the virtual power plant according to the similarity coefficient, several power plant groups are obtained, and the difference between the similarity coefficients of any two virtual power plant energy storage devices in each power plant group is less than a preset maximum difference; and calculating the priority detection coefficient for each power plant group; the monitoring platform performs anomaly detection on the energy storage devices of each power plant group in the order from large to small of the priority detection coefficient; when performing anomaly detection on the operation data of the energy storage devices in the virtual power plant through the monitoring platform, it can reduce the complexity of detection, reduce the influence of chaos or misjudgment on the anomaly detection results of the energy storage devices in the virtual power plant by the monitoring platform, and ensure the accuracy of the anomaly detection results of the monitoring platform.

Description

Technical Field

[0001] The present invention relates to the technical field of anomaly detection, and particularly to an anomaly detection system for a virtual power plant and an anomaly detection method therefor. Background Art

[0002] A virtual power plant is an aggregate composed of multiple distributed energy resources (such as solar energy, wind power generation, energy storage devices, electric vehicle charging piles, etc.). Through intelligent management and dispatching, it realizes the optimal allocation and utilization of electric power. In particular, for the energy storage devices of a virtual power plant, whether the energy storage devices are abnormal directly affects the flexibility and response speed of the virtual power plant dispatching and the stability of the entire power grid dispatching. Therefore, it is necessary to perform anomaly detection on the energy storage devices in the virtual power plant to ensure the stability of the power grid dispatching. In the detection of the energy storage devices in the virtual power plant, the operation data of the energy storage devices of multiple power plants are uniformly collected and uploaded to a monitoring platform for centralized management and monitoring. The monitoring platform uses advanced anomaly detection technologies and combines the real-time collected data to determine whether there are any abnormal situations. The anomaly detection methods mainly include threshold-based methods, statistical analysis-based methods, machine learning algorithms (such as K-means clustering, support vector machines, etc.), and time series analysis-based models (such as LSTM), etc. These methods can identify the possible abnormal conditions of the energy storage devices in the virtual power plant during operation, provide early warning information, and help power plants and relevant operation and maintenance personnel take quick and effective countermeasures to ensure the stability of power supply and the safe operation of the power grid.

[0003] However, when performing anomaly detection on the operation data of the energy storage devices in the virtual power plant through the monitoring platform, generally, the operation data of the energy storage devices of each virtual power plant are simultaneously subjected to anomaly detection. The data is huge and highly complex, which easily leads to confusion or misjudgment in the anomaly detection results of the energy storage devices in the virtual power plant by the monitoring platform, resulting in low accuracy of the anomaly detection results of the monitoring platform. Summary of the Invention

[0004] The object of the present invention is to solve the above-mentioned problems and provide an anomaly detection system for a virtual power plant and an anomaly detection method therefor.

[0005] In the first aspect of the implementation of the present invention, an anomaly detection method for a virtual power plant is first proposed. The method includes:

[0006] For the energy storage devices of each virtual power plant, obtain the device types and working environments of the energy storage devices in every two virtual power plants, and calculate the similarity coefficient of the corresponding two virtual power plants according to the device types and working environments.

[0007] Group the energy storage devices in the virtual power plants according to the similarity coefficients between every two virtual power plants, obtaining several power plant groups; and the difference in similarity coefficients between any two virtual power plant energy storage devices in each power plant group is less than a preset maximum difference;

[0008] For each power plant group, obtain the historical number of anomalies and historical discharge times of the energy storage devices in each virtual power plant within the group, and calculate the priority detection coefficient for each power plant group;

[0009] The monitoring platform performs anomaly detection on the energy storage devices of each power plant group in descending order of the priority detection coefficient.

[0010] Optionally, the steps of obtaining the device types and working environments of the energy storage devices in every two virtual power plants and calculating the similarity value corresponding to the two virtual power plants according to the device types and working environments are as follows:

[0011] Obtain the types of device types of the energy storage devices in each virtual power plant, and form a set corresponding to the set of device types of the energy storage devices;

[0012] And calculate the Jaccard similarity between the two virtual power plants according to the sets of device types of the energy storage devices in the two virtual power plants;

[0013] In the sets of device types of the energy storage devices in the two virtual power plants, record the device types with the same type as the common types, and record the minimum number of such devices in the energy storage devices of the two virtual power plants corresponding to each common type as the common quantity of the common type;

[0014] Add up the common types of all types, and divide the sum by the total number of the energy storage devices in the two virtual power plants to obtain the quantity similarity between the two virtual power plants;

[0015] Obtain the device similarity value of the two virtual power plants according to the Jaccard similarity and quantity similarity between the two virtual power plants:

[0016] Calculate the similarity coefficient corresponding to the two virtual power plants according to the device similarity value and the working environment.

[0017] Optionally, the steps of calculating the similarity coefficient corresponding to the two virtual power plants according to the device similarity value and the working environment are as follows:

[0018] For the energy storage devices in the two virtual power plants, obtain the environmental data of the environments where the two energy storage devices are located within a preset time period, and preprocess the environmental data of the environments where the energy storage devices in the two virtual power plants are located;

[0019] Construct a distance matrix, where each element is the distance between an environmental data point of the environment where the energy storage device is located in one virtual power plant and an environmental data point of the environment where the energy storage device is located in another virtual power plant;

[0020] Use the dynamic programming algorithm to calculate the shortest path in the distance matrix to find the best match between the environmental data sequences of the environments where the energy storage devices are located in two virtual power plants;

[0021] Align the environmental data sequences of the environments where the energy storage devices are located in two virtual power plants according to the calculated shortest path;

[0022] According to the aligned environmental data sequences of the environments where the energy storage devices are located in two virtual power plants, calculate the similarity between them, and use the calculated similarity as the environmental similarity value of the corresponding energy storage devices in the two virtual power plants;

[0023] Calculate the similarity coefficient of the corresponding two virtual power plants according to the device similarity value and the environmental similarity value.

[0024] Optionally, the steps to calculate the similarity coefficient of the corresponding two virtual power plants according to the device similarity value and the environmental similarity value are as follows:

[0025] In the formula, is the similarity coefficient of the two virtual power plants, , are the device similarity value and the environmental similarity value respectively, are respectively , preset proportionality coefficients of and both are greater than 0.

[0026] Optionally, for each power plant group, the steps to obtain the historical abnormal times and historical discharge times of the energy storage devices in each virtual power plant in each power plant group and calculate the priority detection coefficient of each power plant group are as follows:

[0027] Obtain the total number of times of detecting abnormalities in the historical detection records of the energy storage devices in each virtual power plant within each power plant group, and divide the total number of times of abnormalities by the total working time of the energy storage devices to obtain the abnormal frequency of the energy storage devices;

[0028] Obtain the capacity of the energy storage device in each virtual power plant, and multiply the capacity of the energy storage device by the corresponding abnormal frequency to obtain the weighted abnormal value of the energy storage device in the corresponding virtual power plant; Add up the weighted abnormal values of the energy storage devices in each virtual power plant within each power plant group to obtain the total weighted abnormal value of the corresponding power plant group;

[0029] Calculate the priority detection coefficient of each power plant group according to the total weighted abnormal value of each power plant group and the historical discharge times of the energy storage devices in each virtual power plant.

[0030] Optionally, the steps for calculating the priority detection coefficient of each power plant group according to the weighted total abnormal value of each power plant group and the historical discharge times of the energy storage devices in each virtual power plant are as follows:

[0031] Obtain the historical discharge times of the energy storage devices in each virtual power plant of each power plant group, and obtain the actual discharge power and the power before discharge for each discharge. Divide the actual discharge power by the power before discharge to obtain the discharge depth; and calculate the average discharge depth of the energy storage devices in each virtual power plant as the average discharge depth;

[0032] According to the discharge depth of each energy storage device in each virtual power plant each time, calculate the standard deviation of the corresponding discharge depth, and use the standard deviation as the discharge depth instability value of the energy storage device in each virtual power plant;

[0033] Obtain the discharge frequency of the energy storage devices in each virtual power plant, normalize the discharge frequency to obtain the discharge frequency value, and calculate the discharge influence coefficient of each power plant group according to the discharge frequency value, the average discharge depth, and the discharge depth instability value. The calculation formula is: , where is the discharge influence coefficient of the power plant group, represents the sequence number of the energy storage device in the virtual power plant in the power plant group, represents the total number of energy storage devices in the virtual power plant in the power plant group; , and respectively represent the average discharge depth, the discharge depth instability value, and the discharge frequency value; , , are respectively , and preset weight values, and , , are all greater than 0;

[0034] Calculate the priority detection coefficient of each power plant group according to the weighted total abnormal value and the discharge influence coefficient of each power plant group.

[0035] Optionally, the steps for calculating the priority detection coefficient of each power plant group according to the weighted total abnormal value and the discharge influence coefficient of each power plant group are as follows:

[0036] where is the priority detection coefficient of each power plant group, , are respectively the weighted total abnormal value and the discharge influence coefficient of each power plant group, are respectively , The preset proportionality coefficient, and both are greater than 0.

[0037] In the second aspect of the implementation of the present invention, a virtual power plant anomaly detection system is proposed. The system includes:

[0038] Similarity module: For the energy storage devices of each virtual power plant, obtain the device types and working environments of the energy storage devices in every two virtual power plants, and calculate the similarity coefficient of the corresponding two virtual power plants according to the device types and working environments;

[0039] Grouping module: Group the energy storage devices in the virtual power plant according to the similarity coefficients of every two virtual power plants to obtain several power plant groups; and the difference in similarity coefficients between any two virtual power plant energy storage devices in each power plant group is less than the preset maximum difference;

[0040] Priority detection module: For each power plant group, obtain the historical anomaly times and historical discharge times of the energy storage devices in each virtual power plant in the power plant group, and calculate the priority detection coefficient of each power plant group;

[0041] Anomaly detection module: The monitoring platform performs anomaly detection on the energy storage devices of each power plant group in the order from largest to smallest of the priority detection coefficients.

[0042] Advantages of the present invention:

[0043] The present invention proposes a virtual power plant anomaly detection system and its anomaly detection method. By calculating the similarity coefficients between the energy storage devices of every two virtual power plants, and grouping the energy storage devices in the virtual power plant according to the similarity coefficients of every two virtual power plants to obtain several power plant groups, and the difference in similarity coefficients between any two virtual power plant energy storage devices in each power plant group is less than the preset maximum difference; and calculating the priority detection coefficient for each power plant group; the monitoring platform performs anomaly detection on the energy storage devices of each power plant group in the order from largest to smallest of the priority detection coefficients; in this way, when performing anomaly detection on the operation data of the energy storage devices in the virtual power plant through the monitoring platform, it is possible to reasonably perform anomaly detection on the operation data of the energy storage devices of each virtual power plant in sequence, reduce the complexity of detection, reduce the influence of chaos or misjudgment on the anomaly detection results of the energy storage devices in the virtual power plant by the monitoring platform, and ensure the accuracy of the anomaly detection results of the monitoring platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The following further describes the present invention with reference to the drawings.

[0045] Figure 1 It is a flowchart of a virtual power plant anomaly detection method;

[0046] Figure 2 It is a framework diagram of a virtual power plant anomaly detection system. Specific embodiments

[0047] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0048] All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0049] The embodiments of the present invention provide a method for detecting anomalies in a virtual power plant. Refer to Figure 1 , Figure 1 which is a flowchart of a method for detecting anomalies in a virtual power plant provided by the embodiments of the present invention. The method includes the following steps:

[0050] For the energy storage devices of each virtual power plant, obtain the device types and working environments of the energy storage devices in every two virtual power plants, and calculate the similarity coefficients of the corresponding two virtual power plants according to the device types and working environments;

[0051] Group the energy storage devices in the virtual power plant according to the similarity coefficients of every two virtual power plants to obtain several power plant groups; and the difference in similarity coefficients between any two virtual power plant energy storage devices in each power plant group is less than a preset maximum difference;

[0052] For each power plant group, obtain the historical anomaly times and historical discharge times of the energy storage devices in each virtual power plant in the power plant group, and calculate the priority detection coefficient of each power plant group;

[0053] The monitoring platform performs anomaly detection on the energy storage devices of each power plant group in descending order of the priority detection coefficient.

[0054] Based on the method for detecting anomalies in a virtual power plant provided by the embodiments of the present invention, in the above manner, when the monitoring platform performs anomaly detection on the operation data of the energy storage devices in the virtual power plant, it can reasonably perform anomaly detection on the operation data of the energy storage devices of each virtual power plant in sequence, reduce the complexity of detection, reduce the impact of confusion or misjudgment on the anomaly detection results of the monitoring platform for the energy storage devices in the virtual power plant, and ensure the accuracy of the anomaly detection results of the monitoring platform.

[0055] In one embodiment, the step of obtaining the device types and working environments of the energy storage devices in every two virtual power plants and calculating the similarity values of the corresponding two virtual power plants according to the device types and working environments is as follows:

[0056] Obtain the types of the energy storage devices in each virtual power plant, and form a set corresponding to the set of the device types of the energy storage devices;

[0057] And calculate the Jaccard similarity between the two virtual power plants according to the sets of the device types of the energy storage devices in the two virtual power plants;

[0058] In the sets of the device types of the energy storage devices in the two virtual power plants, record the device types with the same type as the common types, and record the minimum quantity of the devices in the energy storage devices of the two virtual power plants corresponding to each common type as the common quantity of the common type;

[0059] Add up the common types of all types, and divide the sum by the total number of the quantities of the energy storage devices in the two virtual power plants to obtain the quantity similarity between the two virtual power plants;

[0060] Obtain the device similarity value of the two virtual power plants according to the Jaccard similarity and the quantity similarity between the two virtual power plants:

[0061] Calculate the similarity coefficient corresponding to the two virtual power plants according to the device similarity value and the working environment.

[0062] It should be noted that the device types of the energy storage devices in the virtual power plant can include various different types of energy storage devices, such as lithium batteries, lead-acid batteries, supercapacitors, etc.; these device types reflect different technologies and applications of the energy storage system, and have different energy storage and release characteristics; in the above calculation process, the data involved mainly includes the types and quantities of the energy storage devices in each virtual power plant, which are usually obtained from the device installation records in the monitoring platform of the virtual power plant; the Jaccard similarity is a statistic used to measure the similarity between two sets, defined as the size of the intersection of the two sets divided by the size of the union of the two sets; the similarity between them is evaluated by comparing the ratio of the common elements and all elements in the two sets; specifically in the example of the energy storage devices in the virtual power plant, the Jaccard similarity can be used to measure the similarity degree of the energy storage device types in the two virtual power plants.

[0063] The steps to obtain the device similarity value of the two virtual power plants according to the Jaccard similarity and the quantity similarity between the two virtual power plants are as follows: , where is the device similarity value of the two virtual power plants, , are the Jaccard similarity and the quantity similarity respectively, are respectively , 's preset proportionality coefficients, and are all greater than 0; in addition, is set by professionals according to the actual situation. Generally, The sum is 1. For example, they can be 0.55 and 0.45 respectively, or other numbers, which are not specifically limited.

[0064] It should be noted that when the similarity value of the equipment of two virtual power plants is larger, when they are jointly uploaded to the monitoring platform for anomaly detection, the detection result of the monitoring platform is more accurate. The reason is that when the similarity value of the equipment of two virtual power plants is larger, it means that the types and quantities of their energy storage equipment are relatively close, which to a certain extent indicates that the equipment of these two power plants has high consistency in operation mode, performance and response characteristics. Therefore, when the monitoring platform performs anomaly detection on these devices, the detection result will be more accurate. The reason is that a higher similarity value can help the monitoring platform better understand the normal operation mode and expected behavior of the devices, and then can more accurately identify which data deviate from the normal range; if the equipment characteristics of two virtual power plants are quite different, the detection system may need to make complex adjustments and optimizations in multiple dimensions, increasing the risk of misjudgment or missed judgment. When the equipment similarity value is high, the monitoring platform can rely on the historical data and typical behaviors of similar equipment groups for more refined analysis, thereby improving the reliability and accuracy of anomaly detection. In short, high similarity provides a stable benchmark for anomaly detection, improving the response speed of the system and the accuracy of fault diagnosis.

[0065] In one embodiment, the steps of calculating the similarity coefficient corresponding to two virtual power plants according to the equipment similarity value and the working environment are as follows:

[0066] For the energy storage equipment in two virtual power plants, obtain the environmental data of the environments where the two energy storage devices are located within a preset time period, and preprocess the environmental data of the environments where the energy storage equipment in the two virtual power plants are located;

[0067] Construct a distance matrix, where each element is the distance between the environmental data points of the environment where one energy storage device in a virtual power plant is located and the environmental data points of the environment where the other energy storage device in the other virtual power plant is located;

[0068] Use the dynamic programming algorithm to calculate the shortest path in the distance matrix to find the best match between the environmental data sequences of the environments where the energy storage equipment in the two virtual power plants are located;

[0069] According to the calculated shortest path, align the environmental data sequences of the environments where the energy storage equipment in the two virtual power plants are located;

[0070] According to the aligned environmental data sequences of the environments where the energy storage equipment in the two virtual power plants are located, calculate the similarity between them, and use the calculated similarity as the environmental similarity value of the energy storage equipment in the corresponding two virtual power plants;

[0071] Calculate the similarity coefficient of the corresponding two virtual power plants according to the device similarity value and the environment similarity value.

[0072] It should be noted that the preset time period is set by professionals according to the actual situation, and no specific limitation and elaboration are made here; for each energy storage device of the two virtual power plants, collect the environmental data of the environment where it is located within the preset time period. The relevant meteorological data can be obtained through the external meteorological data providing department where the energy storage device is located. These environmental data include temperature, humidity, air pressure, wind speed, etc. These factors will affect the working state and performance of the energy storage device, and thus affect the anomaly detection of the device; in order to make the data comparable effectively, the environmental data needs to be preprocessed, which includes denoising, standardization or normalization processing to ensure the comparability between different data sources; after the environmental data is preprocessed, a distance matrix is constructed, where each element represents the distance between the environmental data points of two energy storage devices. To calculate the "distance" between environmental data points, Euclidean distance, Manhattan distance or other distance metrics suitable for the type of environmental data can be used; after the distance matrix is constructed, a dynamic programming algorithm is used to calculate the shortest path in the distance matrix, with the aim of finding the best match between the environmental data sequences of the environments where the energy storage devices in the two virtual power plants are located; the dynamic time warping algorithm can effectively handle the non-linear alignment problem in time series data, ensuring that even when there is a time misalignment, the minimum matching error between the two sets of data can be found; the shortest path obtained by dynamic programming can align the environmental data sequences of the energy storage devices of the two virtual power plants. This alignment process enables the environmental data points of the two devices at different times to correspond, reducing the error caused by time or data misalignment; finally, according to the aligned environmental data sequences, the environmental similarity between the energy storage devices of the two virtual power plants can be calculated; common calculation methods include calculating the root mean square error (RMSE), Pearson correlation coefficient, etc. between the aligned sequences. These metrics can reflect the similarity degree of the environments where the two energy storage devices are located. For example, the closer the Pearson correlation coefficient is to 1, the higher the environmental similarity between the two devices.

[0073] It should be noted that when the environmental similarity value between two virtual power plants is larger, when they are jointly uploaded to the monitoring platform for anomaly detection, the detection result of the monitoring platform is more accurate. The reason is as follows: A high environmental similarity value means that the energy storage devices of these two virtual power plants operate under similar environmental conditions and may face similar external influencing factors, such as temperature changes, humidity fluctuations, or voltage fluctuations. Therefore, the monitoring platform can perform more accurate anomaly prediction and identification based on these similar environmental characteristics. Environmental factors often have an important impact on the performance and stability of energy storage devices. When two devices operate in a similar environment, the patterns and frequencies of their anomalies may also be similar, which enables the monitoring platform to quickly identify potential anomalies through comparison and model training, avoiding misjudgment or missed judgment caused by device differences or inconsistent environmental factors. Therefore, the higher the environmental similarity, the more similar data features the monitoring platform can utilize during anomaly detection, thereby improving the detection accuracy and response speed.

[0074] In one embodiment, the steps of calculating the similarity coefficient corresponding to two virtual power plants according to the device similarity value and the environmental similarity value are as follows:

[0075] ; where is the similarity coefficient of the two virtual power plants, and are the device similarity value and the environmental similarity value respectively, are respectively and 's preset proportionality coefficients, and are both greater than 0;

[0076] It should be noted that is set by professionals according to the actual situation. Generally, 's sum is 1. For example can be 0.5, 0.5 respectively, or other numbers, and specific values are not limited.

[0077] In one embodiment, the energy storage devices in the virtual power plant are grouped according to the similarity coefficient of every two virtual power plants, obtaining several power plant groups; and the difference in the similarity coefficient between any two energy storage devices in each power plant group is less than the preset maximum difference;

[0078] Grouping the energy storage devices in the virtual power plant according to the similarity coefficient of every two virtual power plants is actually quantifying the similarity between energy storage devices through similarity, so as to effectively perform subsequent anomaly detection. The core goal of this step is to gather energy storage devices with higher similarity together, so that when the monitoring platform performs anomaly detection, it can centrally process devices with similar environmental and device characteristics, thereby improving the detection accuracy and efficiency.

[0079] In the specific grouping process, first, the similarity coefficient between every two virtual power plants is calculated to measure the similarity of the equipment and environment between them. If the similarity coefficient between two virtual power plants is less than a preset maximum difference (threshold), then these two virtual power plants and their energy storage devices are grouped into the same group. This preset threshold (maximum difference) is used to ensure sufficient similarity between the virtual power plants within each group, thus avoiding misclassifying energy storage devices with too large differences into the same group. Specifically, if the difference in the similarity coefficients of two virtual power plants is large, it indicates that there are significant differences in the equipment types, working environments, etc. of their energy storage devices, and such devices should not be placed in the same group for processing. On the contrary, if their similarity coefficients are small, it can be considered that their energy storage devices are very similar in certain characteristics, and combining them into the same group is convenient for centralized monitoring and detection.

[0080] In one implementation method, the advantage of such grouping is that by grouping energy storage devices with high similarity into the same group, the management efficiency of the monitoring platform for the energy storage devices of each virtual power plant can be effectively improved. First, the grouped devices, under similar environments and equipment characteristics, can reduce errors and inaccurate judgments caused by equipment differences, ensuring that the monitoring system can more accurately identify potential faults and abnormal conditions. Second, through grouping, the monitoring platform can formulate corresponding detection strategies and priorities according to the characteristics of different groups, avoiding undifferentiated full-scale detection, which not only improves the detection accuracy but also saves computing and processing resources. In addition, grouping helps to respond more quickly to potential risks. For example, when an abnormality occurs in the energy storage devices within a certain group, the root cause of the problem can be quickly determined based on the common characteristics of the group, improving the timeliness of emergency response. Generally speaking, such a grouping mechanism makes the monitoring platform more flexible and efficient when facing a large number and variety of energy storage devices, and can improve the overall efficiency of management and operation and maintenance while ensuring safety.

[0081] In one embodiment, for each power plant group, the steps of obtaining the historical number of abnormal occurrences and historical discharge times of the energy storage devices in each virtual power plant within each power plant group and calculating the priority detection coefficient of each power plant group are as follows:

[0082] Obtain the total number of times of detected abnormalities in the historical detection records of the energy storage devices of each virtual power plant within each power plant group, and divide the total number of abnormal occurrences by the total working time of the energy storage devices to obtain the abnormal frequency of the energy storage devices;

[0083] Obtain the capacity of the energy storage devices in each virtual power plant, and multiply the capacity of the energy storage devices by the corresponding abnormal frequency to obtain the weighted abnormal value of the energy storage devices in the corresponding virtual power plant; Add up the weighted abnormal values of the energy storage devices in each virtual power plant within each power plant group to obtain the total weighted abnormal value of the corresponding power plant group;

[0084] Calculate the priority detection coefficient for each power plant group based on the weighted total abnormal value of each power plant group and the historical discharge times of energy storage devices in each virtual power plant.

[0085] It should be noted that the total number of abnormalities in the historical detection records is usually obtained through the monitoring system or maintenance records. The monitoring platform will record the faults and abnormalities of each energy storage device, including downtime, performance degradation, etc. When obtaining the total working hours of the energy storage device, it can be extracted through the operation records of the device, the life cycle log, or the working hour data accumulated in the energy management system. The capacity of the energy storage device is usually obtained through the technical specifications of the device or the device management system, and this data is usually provided by the device manufacturer and recorded in the system. Finally, the calculation of the weighted abnormal value needs to combine the actual operation data of the device with the detection history, and the capacity and abnormal frequency are weighted through an algorithm. All these data are automatically collected, summarized, and processed through an integrated energy management platform or monitoring system.

[0086] It should be noted that the larger the weighted total abnormal value of each power plant group, the higher the order of the virtual power plant in the corresponding power plant group for abnormal detection through the monitoring platform. The reason is that when the weighted total abnormal value of each power plant group is larger, it indicates that the energy storage devices in the virtual power plants within this power plant group have a higher frequency of abnormalities, and the capacities of these energy storage devices are larger. Therefore, their importance and influence during normal operation are also greater. The weighted total abnormal value not only reflects the failure frequency of the device but also takes into account the capacity of the device, which means that the energy storage devices of this power plant may have a more significant impact on grid stability and the scheduling flexibility of the virtual power plant. Due to the relatively high potential failure risks of these devices and the possible large fluctuations in power supply when a failure occurs, the monitoring platform will prioritize the abnormal detection of power plant groups with higher weighted total abnormal values, so as to ensure that problems can be discovered as early as possible and measures can be taken in a timely manner to avoid serious impacts on grid stability. Through such sequential detection, the accuracy and response speed of detection can be effectively improved, and potential power failures caused by ignoring high-risk devices can be reduced.

[0087] In one implementation, the benefit of analyzing the weighted total abnormal value of each power plant group for the abnormal detection result of the energy storage device of the virtual power plant by the monitoring platform is as follows: The weighted total abnormal value of each power plant group can provide an objective and comprehensive basis for the monitoring platform to evaluate the priority of abnormal detection. By combining the abnormal frequency and capacity of the energy storage device, the platform can not only identify the devices with higher failure rates, but also consider the importance of the devices, thus effectively improving the efficiency and accuracy of abnormal detection. Specifically, when the weighted total abnormal value is large, it indicates that the abnormal risk of the energy storage devices in the power plant group is high and may have a greater impact on the stability of the power system. Prioritizing the detection of these devices helps the platform to discover potential fault problems in the first time and avoid power grid failures or energy supply interruptions caused by neglecting high-risk devices. Therefore, this calculation method of the weighted total abnormal value can significantly improve the response speed of the monitoring platform, give early warnings of possible abnormalities, and ensure the more stable and reliable operation of the power system.

[0088] In one embodiment, the steps of calculating the priority detection coefficient of each power plant group according to the weighted total abnormal value of each power plant group and the historical discharge times of the energy storage devices in each virtual power plant are as follows:

[0089] Obtain the historical discharge times of the energy storage devices in each virtual power plant of each power plant group, and obtain the actual discharge power and the power before discharge for each discharge. Divide the actual discharge power by the power before discharge to obtain the discharge depth; and calculate the average discharge depth of the energy storage devices in each virtual power plant as the average discharge depth;

[0090] According to the discharge depth of each energy storage device in each virtual power plant each time, calculate the standard deviation of the corresponding discharge depth, and use the standard deviation as the discharge depth instability value of the energy storage devices in each virtual power plant;

[0091] Obtain the discharge frequency of the energy storage devices in each virtual power plant, normalize the discharge frequency to obtain the discharge frequency value, and calculate the discharge influence coefficient of each power plant group according to the discharge frequency value, the average discharge depth, and the discharge depth instability value. The calculation formula is: , where is the discharge influence coefficient of the power plant group, represents the sequence number of the energy storage device in the virtual power plant in the power plant group, represents the total number of energy storage devices in the virtual power plant in the power plant group; , and respectively represent the average discharge depth, the discharge depth instability value, and the discharge frequency value; , , are respectively , and Preset weight values, and , , are all greater than 0;

[0092] Calculate the priority detection coefficient for each power plant group according to the weighted total abnormal value and discharge influence coefficient of each power plant group.

[0093] It should be noted that , , are set by professionals according to the actual situation. Generally, , , sum to 1. For example , , can be 0.3, 0.3, 0.4 respectively, or other numbers, and there is no specific limit.

[0094] It should be noted that in the above calculation process, the main data acquisition methods involved are first to obtain the historical discharge times of the energy storage devices in each virtual power plant through the control system or monitoring platform of the energy storage devices. These systems will record the detailed information of each discharge, including the start and end times of the discharge and the discharge power. The actual discharge power and the power before discharge can be monitored in real time through the battery management system (BMS) of the energy storage device. These data provide the battery state information of the device during the discharge process. The calculation of the discharge depth is obtained by mathematically processing the ratio of the actual discharge power to the undischarged power, and the mean and standard deviation of the discharge depth are calculated through statistical analysis tools (such as the pandas or NumPy libraries in Python). These reflect the discharge mode and its stability of the device. The discharge frequency is obtained by calculating the number of discharges of the device within a certain period of time, and it can usually be directly obtained on the monitoring platform. For the normalization of the discharge frequency, common standardization methods (such as min-max normalization or Z-Score standardization) are used to adjust it to a unified scale for subsequent calculations. By obtaining these key data and combining the preset weight coefficients, the discharge influence coefficient of each power plant group can be accurately calculated, which can provide strong support for the equipment scheduling, maintenance decision-making and abnormal detection of the power plant.

[0095] It should be noted that the greater the discharge influence coefficient of each power plant group, the higher the priority of the virtual power plant in the corresponding power plant group for anomaly detection through the monitoring platform. The reason is that when the discharge influence coefficient of each power plant group is greater, it means that the energy storage devices of the virtual power plants in this power plant group have experienced higher frequency and depth of discharge during the discharge process, and at the same time, the instability of their discharge is also higher. This indicates that the health status of these energy storage devices is relatively poor, and there may be a higher risk of failure. Because energy storage devices are prone to problems such as large battery losses and temperature rises during frequent and deep discharge, which increases the probability of equipment aging or failure. Therefore, the energy storage devices of the virtual power plants in the power plant groups with higher discharge influence coefficients should be given priority for anomaly detection. This priority ranking can help the monitoring platform detect anomalies in a timely manner before potential equipment failures occur, thereby reducing the risk of failures and improving the safety and stability of the power grid. Through this comprehensive assessment based on discharge depth, frequency, and instability, the monitoring platform can more effectively allocate resources and make maintenance decisions to ensure the reliability of the operation of the power system.

[0096] In one implementation method, the benefits of analyzing the discharge influence coefficients of each power plant group for the anomaly detection results of the energy storage devices of virtual power plants by the monitoring platform are as follows: Through the calculation of the discharge influence coefficients, the monitoring platform can give priority to detecting these devices before potential failures occur, so as to perform maintenance or adjustment before the equipment fails or its performance deteriorates; This method can not only reduce the risks of the system and avoid power supply interruptions caused by energy storage device failures, but also extend the service life of the equipment and reduce maintenance and replacement costs by identifying problem devices in advance; By closely linking the detection priority with the discharge situation, the monitoring platform can concentrate resources on the devices that require the most attention, improve the maintenance efficiency and reliability of the entire power system, and ultimately ensure the stability and safety of the power grid.

[0097] In one embodiment, the steps for calculating the priority detection coefficient of each power plant group based on the weighted anomaly total value and the discharge influence coefficient of each power plant group are as follows:

[0098] ; where is the priority detection coefficient of each power plant group, , are respectively the weighted anomaly total value and the discharge influence coefficient of each power plant group, are respectively , 's preset proportionality coefficients, and are both greater than 0;

[0099] It should be noted that is set by professionals according to the actual situation. Generally, 's sum is 1. For example They can be 0.3 and 0.7 respectively, or other numbers, without specific limitations.

[0100] In one embodiment, the monitoring platform performs anomaly detection on the energy storage devices of each power plant group in descending order of the priority detection coefficient. By calculating the priority detection coefficient based on the weighted total anomaly value and the discharge impact coefficient of each power plant group, the monitoring platform can accurately identify those energy storage devices that may have higher risks and perform timely anomaly detection. The weighted total anomaly value can reflect the potential problems and failure frequencies of the device during operation, while the discharge impact coefficient measures the usage pressure and possible health risks of the energy storage device. When these two indicators are combined, it can effectively identify those devices that have both historical records of anomaly detection and frequently experience high-load discharges. Due to long-term high-frequency discharges and unstable discharge depths, these devices are extremely prone to performance degradation or failures. Therefore, prioritizing the detection of these devices helps to avoid potential systemic risks and the spread of failures, thus ensuring the stable operation of the power grid. At the same time, by using the method of sorting the priority detection coefficients from large to small, the resource allocation can be optimized, enabling the monitoring platform to focus its efforts on the power plant groups that most need attention, improving the detection efficiency, avoiding ineffective detections, enhancing the accuracy and response speed of anomaly detection, and reducing the confusion or misjudgment of anomaly detection results, thereby ensuring the safety and stability of power supply to the greatest extent.

[0101] It should be noted that when the monitoring platform performs anomaly detection on the energy storage devices in the virtual power plants of each power plant group in descending order of the priority detection coefficient, it usually relies on a series of comprehensive data analysis and real-time monitoring technologies. First, the platform will perform real-time monitoring on the energy storage devices and collect various key indicators, such as the voltage, current, load, output power of the battery, etc. These data will be transmitted to the analysis system of the platform and compared with historical data to identify possible abnormal patterns of the device. For example, abnormal voltage fluctuations, increased load, large fluctuations in output power, etc. may all be signals that the device is about to fail. According to the sorting of the priority detection coefficient, the monitoring platform first conducts a detailed health check on the devices with higher weighted total anomaly values and higher discharge frequencies. These devices are usually first set as high-risk targets, and the platform will start more refined diagnostic mechanisms, such as predictive maintenance algorithms, machine learning models, etc., to further analyze the abnormal trends of the devices and predict potential failures. According to the detection results, the platform will automatically generate a detailed diagnostic report and automatically execute corresponding emergency response strategies according to the detection severity, such as shutdown inspection, adjustment of operating parameters, or arrangement of maintenance, etc., so as to effectively prevent the impact of device failures on the power grid.

[0102] Based on the same inventive concept, the embodiment of the present invention also provides a virtual power plant anomaly detection system. Refer to Figure 2 , Figure 2A framework diagram of an abnormal detection system for a virtual power plant provided by an embodiment of the present invention. The system includes:

[0103] Similarity module: For the energy storage devices of each virtual power plant, obtain the device types and working environments of the energy storage devices in every two virtual power plants, and calculate the similarity coefficients of the corresponding two virtual power plants according to the device types and working environments.

[0104] Grouping module: Group the energy storage devices in the virtual power plant according to the similarity coefficients of every two virtual power plants to obtain several power plant groups; and the difference in similarity coefficients between any two virtual power plant energy storage devices in each power plant group is less than a preset maximum difference.

[0105] Priority detection module: For each power plant group, obtain the historical abnormal times and historical discharge times of the energy storage devices in each virtual power plant in the power plant group, and calculate the priority detection coefficient of each power plant group.

[0106] Abnormal detection module: The monitoring platform performs abnormal detection on the energy storage devices of each power plant group in descending order of the priority detection coefficient.

[0107] Based on the abnormal detection system for a virtual power plant provided by an embodiment of the present invention, in the above manner, when the monitoring platform performs abnormal detection on the operation data of the energy storage devices in the virtual power plant, it can reasonably perform abnormal detection on the operation data of the energy storage devices of each virtual power plant in sequence, reduce the complexity of detection, reduce the influence of chaos or misjudgment on the abnormal detection results of the energy storage devices in the virtual power plant by the monitoring platform, and ensure the accuracy of the abnormal detection results of the monitoring platform.

[0108] The above has described an embodiment of the present invention in detail, but the content is only a preferred embodiment of the present invention and cannot be artificially used to limit the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of the present invention application should still fall within the scope covered by the patent of the present invention.

Claims

1. A method for abnormal detection of a virtual power plant, characterized in that It includes the following steps: For the energy storage devices of each virtual power plant, obtain the device types and working environments of the energy storage devices in every two virtual power plants, and calculate the similarity coefficients of the corresponding two virtual power plants according to the device types and working environments; Group the energy storage devices in the virtual power plants according to the similarity coefficients of every two virtual power plants to obtain several power plant groups; and the difference in similarity coefficients between any two virtual power plant energy storage devices in each power plant group is less than the preset maximum difference; For each power plant group, obtain the historical abnormal times and historical discharge times of the energy storage devices in each virtual power plant in the power plant group, and calculate the priority detection coefficient of each power plant group; The monitoring platform performs abnormal detection on the energy storage devices of each power plant group in the order from large to small of the priority detection coefficients; For each power plant group, the steps of obtaining the historical abnormal times and historical discharge times of the energy storage devices in each virtual power plant in the power plant group and calculating the priority detection coefficient of each power plant group are as follows: Obtain the total number of times of detecting abnormalities in the historical detection records of the energy storage devices of each virtual power plant in each power plant group, and divide the total number of times of abnormalities by the total working time of the energy storage devices to obtain the abnormal frequency of the energy storage devices; Obtain the capacity of the energy storage devices in each virtual power plant, and multiply the capacity of the energy storage devices by the corresponding abnormal frequency to obtain the weighted abnormal value of the energy storage devices in the corresponding virtual power plant; add up the weighted abnormal values of the energy storage devices in each virtual power plant in each power plant group to obtain the total weighted abnormal value of the corresponding power plant group; Calculate the priority detection coefficient of each power plant group according to the total weighted abnormal value of each power plant group and the historical discharge times of the energy storage devices in each virtual power plant. The steps are as follows: Obtain the historical discharge times of the energy storage devices in each virtual power plant in each power plant group, and obtain the actual discharge power and the power before discharge for each discharge. Divide the actual discharge power by the power before discharge to obtain the discharge depth; And calculate the average discharge depth of the energy storage devices in each virtual power plant as the average discharge depth; According to the discharge depth of the energy storage devices in each virtual power plant each time, calculate the standard deviation of the corresponding discharge depth, and use the standard deviation as the discharge depth instability value of the energy storage devices in each virtual power plant; Obtain the discharge frequency of the energy storage devices in each virtual power plant, normalize the discharge frequency to obtain the discharge frequency value, and calculate the discharge influence coefficient of each power plant group according to the discharge frequency value, the average discharge depth, and the discharge depth instability value. The calculation formula is as follows: , where is the discharge influence coefficient of the power plant group, represents the sequence number of the energy storage device in the virtual power plant within the power plant group, represents the total number of energy storage devices in the virtual power plant within the power plant group; , and respectively represent the average discharge depth, the discharge depth instability value, and the discharge frequency value; , , are respectively , and preset weight values, and , , are all greater than 0; Calculate the priority detection coefficient of each power plant group according to the total weighted abnormal value of each power plant group and the discharge influence coefficient.

2. The abnormal detection method of a virtual power plant according to claim 1, wherein The steps of obtaining the device types and working environments of the energy storage devices in every two virtual power plants and calculating the similarity value of the corresponding two virtual power plants according to the device types and working environments are as follows: Obtain the types of the device types of the energy storage devices in each virtual power plant to obtain the device type set of the corresponding energy storage devices; And calculate the Jaccard similarity between the two virtual power plants according to the device type sets of the energy storage devices in the two virtual power plants; In the device type sets of the energy storage devices in the two virtual power plants, record the device types with the same type as the common types, and record the minimum number of the devices in the energy storage devices of the two virtual power plants corresponding to each common type as the common number of the common type; Add up all types of common types, and divide the sum by the total number of the energy storage devices in the two virtual power plants to obtain the quantity similarity between the two virtual power plants; Obtain the equipment similarity value of two virtual power plants based on the Jaccard similarity and quantity similarity between the two virtual power plants: Calculate the similarity coefficient of the corresponding two virtual power plants according to the equipment similarity value and the working environment.

3. The abnormal detection method for a virtual power plant according to claim 2, characterized in that, The steps to calculate the similarity coefficient of the corresponding two virtual power plants according to the equipment similarity value and the working environment are as follows: For the energy storage devices in two virtual power plants, obtain the environmental data of the environments where the two energy storage devices are located within a preset time period, and preprocess the environmental data of the environments where the energy storage devices in the two virtual power plants are located; Construct a distance matrix, where each element is the distance between an environmental data point of the environment where an energy storage device in one virtual power plant is located and an environmental data point of the environment where an energy storage device in the other virtual power plant is located; Use the dynamic programming algorithm to calculate the shortest path in the distance matrix to find the best match between the environmental data sequences of the environments where the energy storage devices in the two virtual power plants are located; Align the environmental data sequences of the environments where the energy storage devices in the two virtual power plants are located according to the calculated shortest path; According to the aligned environmental data sequences of the environments where the energy storage devices in the two virtual power plants are located, calculate the similarity between them, and use the calculated similarity as the environmental similarity value of the energy storage devices in the corresponding two virtual power plants; Calculate the similarity coefficient of the corresponding two virtual power plants according to the equipment similarity value and the environmental similarity value.

4. A virtual power plant anomaly detection method according to claim 3, characterized in that, The steps to calculate the similarity coefficient of the corresponding two virtual power plants according to the equipment similarity value and the environmental similarity value are as follows: ; In the formula, is the similarity coefficient of two virtual power plants, , are the equipment similarity value and the environment similarity value respectively, are respectively , 's preset proportionality coefficients, and are both greater than 0.

5. A virtual power plant anomaly detection method according to claim 1, characterized in that, The steps to calculate the priority detection coefficient of each power plant group according to the weighted total abnormal value and discharge influence coefficient of each power plant group are as follows: ; wherein, is the priority detection coefficient for each power plant group, , are respectively the weighted total abnormal value and the discharge influence coefficient of each power plant group, are respectively , 's preset proportionality coefficients, and are all greater than 0.

6. A virtual power plant anomaly detection system for implementing the virtual power plant anomaly detection method according to any one of claims 1-5 above, characterized in that, The system includes: Similarity module: For the energy storage devices of each virtual power plant, obtain the equipment types and working environments of the energy storage devices in every two virtual power plants, and calculate the similarity coefficient of the corresponding two virtual power plants according to the equipment types and working environments; Grouping module: Group the energy storage devices in the virtual power plants according to the similarity coefficient of every two virtual power plants to obtain several power plant groups; and the difference in the similarity coefficient between any two virtual power plant energy storage devices in each power plant group is less than the preset maximum difference; Priority detection module: For each power plant group, obtain the historical abnormal times and historical discharge times of the energy storage devices in each virtual power plant within each power plant group, and calculate the priority detection coefficient of each power plant group; Abnormal detection module: The monitoring platform performs abnormal detection on the energy storage devices of each power plant group in descending order of the priority detection coefficient.

Images