Distributed energy storage intelligent dispatching method and system based on transformer area autonomy

Abstract

The application discloses a kind of distributed energy storage intelligent scheduling method and system based on district autonomy, comprising: first, the task instruction of current electric power into target district is obtained, and electric power load data is obtained according to task instruction. Then, the load data is evaluated using the pre-trained district load evaluation model, to determine whether the load state is normal. If abnormal, determine the main root cause of abnormality. Finally, based on load state evaluation and abnormal state analysis results, select the appropriate scheduling strategy from the pre-configured energy storage scheduling strategy, and issue to the distributed energy storage equipment of target district for execution. Thus designed, the stability and scheduling efficiency of power system are improved, suitable for intelligent management of distributed energy system.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a distributed energy storage intelligent scheduling method and system based on transformer area autonomy. Background Technology

[0002] With the rapid development of distributed energy and smart grid technologies, the autonomous operation of distribution transformers and the intelligent dispatching of distributed energy storage systems are becoming increasingly important. Traditional power dispatching methods typically rely on centralized control systems, which may not be flexible and efficient enough in distributed energy systems. Summary of the Invention

[0003] The purpose of this invention is to provide a distributed energy storage intelligent scheduling method and system based on transformer area autonomy.

[0004] In a first aspect, embodiments of the present invention provide a distributed energy storage intelligent scheduling method based on district autonomy, comprising:

[0005] Obtain the task instruction for the current power supply to the target transformer area;

[0006] According to the task instructions, obtain the current power load data;

[0007] The current power load data is input into a pre-trained transformer area load assessment model to obtain the corresponding load status assessment results.

[0008] When the load status assessment result is characterized as an abnormal load status, the target abnormal root cause with the highest correlation to the abnormal phenomenon of the abnormal load status is determined, and the abnormal status analysis result is obtained.

[0009] Based on the load status assessment results and the abnormal status analysis results, a target scheduling strategy is selected from the pre-configured energy storage scheduling.

[0010] The target scheduling strategy is distributed to the distributed energy storage devices included in the target distribution area for execution.

[0011] In a second aspect, embodiments of the present invention provide a server system, including a server, the server being used to execute the method described in the first aspect.

[0012] Compared to existing technologies, the beneficial effects of this invention include: This invention discloses a distributed energy storage intelligent scheduling method and system based on transformer area autonomy, comprising: firstly, acquiring the task instruction for current power entering the target transformer area, and acquiring power load data according to the task instruction; then, evaluating the load data using a pre-trained transformer area load assessment model to determine whether the load status is normal; if abnormal, identifying the main root cause of the abnormality; finally, based on the load status assessment and abnormal status analysis results, selecting a suitable scheduling strategy from pre-configured energy storage scheduling strategies and issuing it to the distributed energy storage devices in the target transformer area for execution. This design improves the stability and scheduling efficiency of the power system and is suitable for the intelligent management of distributed energy systems. Attached Figure Description

[0013] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart illustrating the steps of a distributed energy storage intelligent scheduling method based on transformer area autonomy provided in an embodiment of the present invention;

[0015] Figure 2 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0017] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0018] In order to solve the technical problems mentioned in the background art Figure 1 This is a flowchart illustrating the distributed energy storage intelligent scheduling method based on district autonomy provided in this embodiment of the disclosure. The following is a detailed description of the distributed energy storage intelligent scheduling method based on district autonomy.

[0019] Step S201: Obtain the task instruction for the current power supply to the target transformer area;

[0020] Step S202: According to the task instruction, obtain the current power load data;

[0021] Step S203: Input the current power load data into the pre-trained transformer area load assessment model to obtain the corresponding load status assessment result;

[0022] Step S204: When the load state assessment result is characterized as an abnormal load state, determine the target abnormal root cause with the highest correlation to the abnormal phenomenon of the abnormal load state, and obtain the abnormal state analysis result.

[0023] Step S205: Based on the load status assessment results and the abnormal status analysis results, select a target scheduling strategy from the pre-configured energy storage scheduling;

[0024] Step S206: The target scheduling strategy is sent to the distributed energy storage devices included in the target distribution area for execution.

[0025] In this embodiment of the invention, for example, the server receives a task instruction from the power management system, requesting the monitoring and scheduling of the power load of "Transformer Area No. 10, XX District, XX City". This task instruction includes the specific location of the transformer area, the monitoring time period, and the main power parameters to be monitored. Based on the received task instruction, the server acquires real-time power load data for "Transformer Area No. 10, XX District, XX City" over the past hour through a remote data acquisition system. This data includes key parameters such as voltage, current, active power, and reactive power. The server inputs the collected power load data into a pre-trained transformer area load assessment model. This model is constructed by analyzing and learning from historical power load data using machine learning algorithms. After processing by the model, the server obtains a load status assessment result, showing that the transformer area is currently under high load and exhibits abnormal fluctuations. Since the load status assessment result shows an anomaly, the server further analyzes the data, comparing the differences between historical and current data, and using anomaly detection algorithms to determine that the main root cause of the anomaly is "a line short circuit or equipment failure at some point in the transformer area". Based on the high load and abnormal fluctuations in the load status, and the identified root cause of the anomaly, the server selected an "emergency load reduction and priority power supply to energy storage devices" strategy from a variety of pre-configured energy storage scheduling strategies. This strategy aims to ensure a stable power supply to the distribution area by reducing some non-critical loads while activating distributed energy storage devices. The server then distributes the selected scheduling strategy to all distributed energy storage devices within "Distribution Area No. 10, XX District, XX City". Upon receiving the instruction, these devices adjust according to the strategy requirements, such as activating energy storage battery discharge and disconnecting some non-critical circuits, to ensure the smooth operation of the power supply in the distribution area.

[0026] In this embodiment of the invention, the transformer area load assessment model is obtained in the following manner.

[0027] The power load data recorded during the transmission of the target power to the target distribution area is obtained, and the power load data is associated with transformer equipment to obtain sub-power load data corresponding to multiple transformer equipment respectively.

[0028] Based on each of the sub-power load data, the load status corresponding to the transformer equipment is determined, and the load status includes safe load status and abnormal load status;

[0029] Using each of the sub-power load data as a sample instance and the corresponding load state as the instance target value, a sample instance array is constructed for training the distribution area load assessment model, and a training process is performed on the distribution area load assessment model based on the sample instance array.

[0030] The transformer area load assessment model is used to assess the load status of the transformer area where the target power is located based on the power load data of the target power.

[0031] In this embodiment of the invention, for example, the server first collects load data generated when "XX City Power Grid" transmits power to "various transformer substations in XX District of XX City". This data covers voltage, current, and power changes throughout the entire transmission process from the power plant to the user. Next, the server uses data analysis technology to correlate this raw power load data with the transformer equipment in each transformer substation. For example, for the three transformer equipment (equipment A, equipment B, and equipment C) in "Transformer Substation No. 10 in XX District of XX City", the server calculates their respective sub-power load data, including the average load and peak load of each equipment at different time periods. The server analyzes the sub-power load data of each transformer equipment and determines the load status of each transformer equipment based on preset safety thresholds and abnormal thresholds. For example, for equipment A, its average load over the past 24 hours is lower than the safety threshold, so it is determined to be in a "safe load state"; while for equipment B, its load peak at certain time points exceeds the abnormal threshold, so it is determined to be in an "abnormal load state". The server takes the sub-load data of each transformer as input and the corresponding load status (safe or abnormal) as output, constructing a large array of sample instances. This array is used to train the transformer load assessment model. Through machine learning algorithms (such as support vector machines and neural networks), the server continuously adjusts the model's parameters to more accurately predict the load status of the transformer based on the input load data. After the above training process, the server obtains a highly efficient transformer load assessment model. This model can receive the load data of "Transformer Area No. 10, XX District, XX City" in real time and quickly determine whether the overall load status of the current transformer area is safe or abnormal. For example, when the model receives a new set of load data, it automatically analyzes the data and outputs a load status assessment result, such as "The current transformer area load is high, but still within the safe range" or "Abnormal load fluctuations have been detected; further inspection of lines and equipment is recommended."

[0032] In this embodiment of the invention, the process of training the transformer area load assessment model based on the sample instance array can be implemented through the following example.

[0033] For each of the sub-electrical load data in the sample instance array, the following processing is performed:

[0034] The sub-electric load data is mapped and processed from multiple data processing domains to obtain the conversion domain data corresponding to each data processing domain.

[0035] Perform feature extraction operation on each of the transformation domain data to obtain the transformation domain features corresponding to each of the transformation domain data, and perform feature integration processing on each of the transformation domain features to obtain integrated features;

[0036] Using the aforementioned transformer area load assessment model and based on the integrated features, the target transformer area is assessed for load to obtain the inferred load status of the target transformer area.

[0037] The training process is performed on the transformer area load assessment model by combining the inferred load status and the corresponding load status.

[0038] In this embodiment of the invention, for example, after the server obtains the power load data of "Transformer Area No. 10, XX District, XX City", it begins to process this data. First, the server performs mapping processing on the sub-power load data from multiple data processing domains, including the time domain, frequency domain, and wavelet domain. For example, in the time domain, the server arranges the data in chronological order to form time series data; in the frequency domain, it uses Fourier transform to convert the data from the time domain to the frequency domain to analyze the spectral characteristics of the power; in the wavelet domain, it uses wavelet transform to extract local features of the data. These mapping processes help the server comprehensively understand the characteristics of the power load data from multiple perspectives. Next, the server extracts features from the data in each transformation domain. In the time domain data, statistical features such as peak value, valley value, and average value of the load are extracted; in the frequency domain data, features such as dominant frequency and spectral energy are extracted; in the wavelet domain data, features such as wavelet coefficients and energy distribution are extracted. Then, the server integrates these features from different transformation domains to form a comprehensive and rich integrated feature set. This integrated feature set can more comprehensively reflect the inherent laws and abnormal patterns of the power load data. The server integrates features into the transformer substation load assessment model, which then infers the load status of "Transformer Substation No. 10 in XX District, XX City" based on these features. Internally, the model employs complex calculations and learning processes to output an inferred load status, such as "The current transformer substation is under abnormal load conditions, with an overload risk." The server compares the inferred load status with the actual observed load status. If a discrepancy is found, the server adjusts the model's parameters and structure to better reflect reality. This process utilizes optimization techniques within machine learning algorithms, aiming to improve the model's accuracy in future inferences. Through repeated training and adjustment, the performance of the transformer substation load assessment model gradually improves, ultimately enabling accurate assessment of the substation's load status.

[0039] In this embodiment of the invention, the data processing domain includes a load time series analysis domain, an energy spectrum analysis domain, and a power quality fluctuation analysis domain. The sub-power load data is mapped from multiple data processing domains to obtain the conversion domain data corresponding to each data processing domain. This can be implemented through the following example.

[0040] From the load time series analysis domain, the sub-power load data is subjected to load time series mapping processing to obtain the load time series transformation domain data corresponding to the load time series analysis domain;

[0041] From the energy spectrum analysis domain, the sub-electric load data is subjected to energy spectrum mapping processing to obtain the energy spectrum conversion domain data corresponding to the energy spectrum analysis domain;

[0042] From the power quality fluctuation analysis domain, the sub-power load data is subjected to power quality fluctuation mapping processing to obtain the power quality fluctuation conversion domain data corresponding to the power quality fluctuation analysis domain.

[0043] In this embodiment of the invention, for example, the server first processes the data from the perspective of the load time series analysis domain. For instance, for the power load data of "Transformer Area No. 10, XX District, XX City" within a certain time period (e.g., 24 hours a day), the server arranges these data according to the time series to form a continuous load time series. This series reflects the change of the power load of the transformer area over time. Through mapping processing, the server obtains the load time series transformed domain data corresponding to the load time series analysis domain. This data can intuitively show the fluctuation trend and peak periods of the power load. Next, the server processes the same set of sub-power load data from the energy spectrum analysis domain. Through mathematical methods such as Fourier transform, the server converts these time series data into frequency domain data, that is, analyzes the spectral characteristics of the power. This conversion process reveals the energy distribution of different frequency components in the power load. For example, the server finds that the power load of "Transformer Area No. 10, XX District, XX City" has obvious energy concentration in a certain frequency band, which can indicate a certain periodic or regular electricity consumption behavior. Through energy spectrum mapping processing, the server obtains the energy spectrum transformed domain data corresponding to the energy spectrum analysis domain. Finally, the server processes the sub-power load data from the power quality fluctuation analysis domain. This step primarily analyzes abnormal fluctuations and disturbances in the power load data to assess the stability of power quality. For example, the server might detect a brief but significant fluctuation in the power load at "Transformer Area No. 10, XX District, XX City" within a certain timeframe. This could be due to equipment failure, line problems, or external interference. Through power quality fluctuation mapping processing, these abnormal fluctuations are clearly identified, forming power quality fluctuation transformation domain data corresponding to the power quality fluctuation analysis domain. This data is crucial for subsequent identification and resolution of power quality issues.

[0044] In this embodiment of the invention, the sub-power load data includes instantaneous load data corresponding to each sampling time node during the transmission of the target power to the corresponding transformer equipment. The step of performing load time series mapping processing on the sub-power load data from the load time series analysis domain to obtain the load time series transformation domain data corresponding to the load time series analysis domain can be implemented through the following example.

[0045] Obtain the load mean, root mean square load value, load fluctuation, load dispersion, and peak data in each instantaneous load data.

[0046] Divide the load dispersion by the load mean to obtain the rate of change deviation of the sub-power load data; divide the peak data by the root mean square load value to obtain the peak-to-average ratio of the sub-power load data; and divide the peak data by the load mean to obtain the impact ratio of the sub-power load data.

[0047] The average load, the load fluctuation, the rate of change deviation, the peak-to-average ratio, and the impact ratio are integrated and processed to obtain the load time series transformation domain data corresponding to the load time series analysis domain.

[0048] In this embodiment of the invention, for example, the server first acquires the instantaneous load data of a transformer in "Transformer Area No. 10, XX District, XX City" for each sampling time point over the past 24 hours. This data reflects the real-time changes in power load over a short period. Next, the server calculates statistical indicators for these instantaneous load data, including the load mean (i.e., the average of all instantaneous load data), the root mean square load value (a statistical measure reflecting load intensity), load fluctuation (indicating the severity of load changes), load dispersion (indicating the degree of dispersion of load data), and the peak data (i.e., the maximum load value) within the instantaneous load data. After acquiring the above statistical indicators, the server further processes the data. First, the load dispersion is divided by the load mean to calculate the rate of change deviation of the sub-power load data; this value reflects the relative fluctuation of the load data. Next, the peak data is divided by the root mean square load value to obtain the peak-to-average ratio of the sub-power load data; this value represents the ratio between the maximum load and the average load intensity. Finally, the peak data is divided by the load mean to calculate the impact ratio of the sub-power load data; this value reflects the impact of the maximum load relative to the average load. The server integrates and processes the calculated load average, load fluctuation, rate of change deviation, peak-to-average ratio, and surge ratio. These data describe the characteristics of the power load from different perspectives. Through integration, a more comprehensive and richer data representation can be formed—the load time series transformed domain data corresponding to the load time series analysis domain. This data not only reflects the basic statistical characteristics of the power load but also reveals important information such as load volatility, stability, and peak characteristics, providing strong data support for subsequent distribution area load assessment.

[0049] In this embodiment of the invention, the step of performing energy spectrum mapping processing on the sub-electricity load data from the energy spectrum analysis domain to obtain the energy spectrum conversion domain data corresponding to the energy spectrum analysis domain can be implemented through the following example.

[0050] Obtain the mean values ​​of the spectral distribution intensity and power spectral density corresponding to the sub-power load data, and determine the spectral center frequency of the sub-power load data in the energy spectrum based on the spectral distribution intensity;

[0051] By combining the center frequency of the spectrum and the intensity of the spectrum distribution, the spectral fluctuation of the sub-power load data on the energy spectrum is determined, and the upper limit frequency of the spectrum in the intensity of the spectrum distribution is obtained;

[0052] The spectral distribution intensity, the mean power spectral density, the spectral center frequency, the spectral fluctuation, and the upper limit frequency of the spectral spectrum are integrated and processed to obtain the energy spectral conversion domain data corresponding to the energy spectral analysis domain.

[0053] In this embodiment of the invention, taking "Transformer Area No. 10, XX District, XX City" as an example, the server first acquires the sub-power load data of this transformer area within a specific time period and converts this time-series data into frequency domain data using signal processing techniques such as Fourier transform. During the conversion process, the server calculates the amplitude of each frequency component to obtain the spectral distribution intensity, i.e., the energy distribution at each frequency. Simultaneously, the server also calculates the mean power spectral density, a statistic representing the distribution of average signal power with frequency. Next, the server determines the spectral center frequency of the sub-power load data in the energy spectrum based on the spectral distribution intensity. This center frequency can be understood as the frequency point where energy is most concentrated, reflecting the main periodic components in the power load data. For example, in the power load data of "Transformer Area No. 10, XX District, XX City," the server finds the spectral center frequency to be 50Hz, indicating that the power load of this transformer area mainly exhibits a periodic variation of 50Hz. The server also combines the spectral center frequency and the spectral distribution intensity to determine the spectral volatility of the sub-power load data in the energy spectrum. This volatility reflects the degree of dispersion of spectral energy, i.e., the distribution of energy at different frequencies. Simultaneously, the server extracts the upper frequency limit from the spectral distribution intensity. This is the highest frequency component in the energy distribution and is typically associated with the highest operating frequency of equipment in the power system. Finally, the server integrates data such as spectral distribution intensity, mean power spectral density, spectral center frequency, spectral fluctuation, and upper frequency limit. These data characterize the frequency domain properties of the power load from different perspectives. Through integrated processing, the server obtains a comprehensive data representation reflecting the frequency domain characteristics of the power load—the energy spectrum conversion domain data corresponding to the energy spectrum analysis domain. This data is crucial for analyzing the operating status of the power system and identifying potential faults and anomalies.

[0054] In this embodiment of the invention, the step of performing feature integration processing on each of the transformation domain features to obtain integrated features can be implemented through the following example.

[0055] Obtain the feature domain attributes corresponding to each of the transformation domain features, and perform a weighted average of each of the feature domain attributes to obtain the reference domain attributes;

[0056] Each of the feature domain attributes is compared with the reference domain attribute to obtain the attribute comparison result corresponding to each feature domain attribute;

[0057] When the attribute comparison result indicates that the feature domain attribute is consistent with the reference domain attribute, the corresponding transformation domain feature is determined as the target transformation domain feature corresponding to the transformation domain feature;

[0058] When the attribute comparison result indicates that the feature domain attribute is inconsistent with the reference domain attribute, the feature domain attribute of the corresponding transformation domain feature is updated to the reference domain attribute to obtain the target transformation domain feature corresponding to the transformation domain feature.

[0059] Each of the target transformation domain features is subjected to feature integration processing to obtain the integrated features.

[0060] In this embodiment of the invention, for example, the server first acquires conversion domain features from the load time series analysis domain, energy spectrum analysis domain, and power quality fluctuation analysis domain. Each conversion domain feature has its corresponding feature domain attributes. For example, the feature domain attributes of load time series conversion domain data may include the stability and periodicity of the time series; the feature domain attributes of energy spectrum conversion domain data may include the concentration and volatility of the spectrum; and the feature domain attributes of power quality fluctuation conversion domain data may include the amplitude and frequency of fluctuations. Next, the server performs a weighted average of these feature domain attributes to obtain a comprehensive reference domain attribute. This reference domain attribute reflects an average or typical characteristic of all conversion domain features. The server then compares each feature domain attribute with the reference domain attribute. For example, if a feature domain attribute of a load time series conversion domain data shows high stability, while the reference domain attribute shows slightly lower overall stability, the comparison result between the two will show inconsistency. When the attribute comparison result shows that the feature domain attribute is consistent with the reference domain attribute, the server retains that conversion domain feature as the target conversion domain feature. If inconsistencies are found, the server will update the feature domain attributes of the transformed domain feature to the reference domain attributes to ensure consistency across all transformed domain features during integration. After completing the attribute comparison and update, the server will perform feature integration processing on all target transformed domain features. This process may include operations such as feature weighting, fusion, or dimensionality reduction to obtain a comprehensive integrated feature that fully reflects the characteristics of the power load data. For example, in the power load data analysis of "Transformer Area No. 10, XX District, XX City," the server may integrate transformed domain features from different analysis domains to obtain an integrated feature that considers both time series stability and incorporates spectral characteristics and power quality fluctuations. This integrated feature provides a more accurate and comprehensive data foundation for subsequent tasks such as transformer area power load assessment, prediction, and fault detection.

[0061] In this embodiment of the invention, the acquisition of power load data recorded during the transmission of target power to the target distribution area can be implemented through the following examples.

[0062] The voltage fluctuations and current changes at each sampling time point during the transmission of the target power to the target distribution area are obtained.

[0063] For each sampling time node, the standardized values ​​of the corresponding voltage fluctuation and the standardized values ​​of the current change are integrated to obtain the instantaneous load data corresponding to the sampling time node;

[0064] The instantaneous load data is obtained by arranging each instantaneous load data according to the time sequence of the sampling time nodes.

[0065] In this embodiment of the invention, taking "Transformer Area No. 10, XX District, XX City" as an example, the server monitors and records the voltage and current conditions of the transformer area when receiving target power in real time. At each sampling time node (e.g., per second or per minute), the server captures the voltage fluctuation and current change data at that moment. These data reflect the actual operating state of the power system at that moment. After obtaining the raw voltage and current data, the server performs standardization processing to eliminate the influence of different dimensions and orders of magnitude on data analysis. For example, the server converts the voltage and current data into standardized values ​​relative to their mean and standard deviation. Next, for each sampling time node, the server integrates the standardized values ​​of the corresponding voltage fluctuation and current change. This integration may include weighted averaging, summation, or other mathematical operations to obtain instantaneous load data that comprehensively reflects the voltage and current state at that moment. After completing the data integration, the server arranges each instantaneous load data according to the time sequence of the sampling time nodes. In this way, the server obtains a series of instantaneous load data arranged in chronological order, which together constitute the power load data. This power load data not only reflects the operating status of the power system at various times, but also provides an important data foundation for subsequent load analysis, forecasting, and fault detection. Through the above steps, the server can accurately acquire and process the power load data recorded during the transmission of target power to the target distribution area, providing strong data support for the stable operation and optimization of the power system.

[0066] In this embodiment of the invention, the sub-power load data includes instantaneous load data corresponding to each sampling time node during the transmission of the target power to the corresponding transformer equipment. The determination of the load status of the corresponding transformer equipment based on each sub-power load data can be implemented through the following example.

[0067] For each of the aforementioned sub-power load data, the following processing is performed:

[0068] For each sampling time node corresponding to the sub-power load data, the target load safety level corresponding to the sampling time node is determined based on the instantaneous load data corresponding to the sampling time node;

[0069] When the target load safety level indicates that the load of the target distribution area through which the target power passes at the sampling time node is within the safe load range, the sampling time node corresponding to the target load safety level is determined as the target sampling time node;

[0070] Based on the number of nodes at the target sampling time node, the load state corresponding to the transformer is determined.

[0071] In this embodiment of the invention, taking the transformer equipment of "Transformer Area No. 10, XX District, XX City" as an example, the server acquires the sub-power load data of the transformer equipment when receiving target power. This data includes instantaneous load data recorded at each sampling time node (e.g., every minute). For each sampling time node, the server determines the target load safety level of that node based on the instantaneous load data (such as voltage, current, power factor, etc.) and a preset safety threshold. Safety levels can be divided into different grades such as "safe," "warning," and "dangerous." After determining the target load safety level for each sampling time node, the server filters out those sampling time nodes characterized as "safe," meaning that at these time points, the load of the transformer area through which the target power passes is within the safe load range. These filtered sampling time nodes are called target sampling time nodes. Based on the number of filtered target sampling time nodes, the server further determines the load status of the corresponding transformer equipment. For example, if the proportion of the target sampling time points to the total number exceeds a certain preset value (e.g., 90%), the server can determine that the transformer is in a "normal load state"; if the proportion is low (e.g., below 60%), it can be determined as a "high load state" or an "overload risk state". Specifically, suppose that in the past hour, the server sampled the transformer in "Transformer Area No. 10, XX District, XX City" 60 times (once per minute). After analysis, the target load safety level of 54 of these sampling time points was "safe". Based on this data, the server can determine that in the past hour, the transformer was within a safe load range for most of the time, therefore its load state can be determined as a "normal load state". Through the above steps, the server can monitor and accurately determine the load state of the transformer in real time, thus providing an important guarantee for the stable operation of the power system.

[0072] In this embodiment of the invention, the determination of the target load safety level corresponding to the sampling time node based on the instantaneous load data corresponding to the sampling time node can be implemented through the following example.

[0073] At least one adjacent sampling time node is obtained from the sampling time node, and the instantaneous load data corresponding to the sampling time node is calculated by difference with the instantaneous load data corresponding to each adjacent sampling time node to obtain the load fluctuation change corresponding to each adjacent sampling time node.

[0074] When at least one of the load fluctuation changes exceeds a preset change threshold, the target load safety level is determined as the first load safety level.

[0075] When no load fluctuation exceeds the preset change threshold, the target load safety level is determined as the second load safety level.

[0076] Wherein, the first load safety level is used to characterize the load of the transformer area through which the target power passes at the sampling time node as an abnormal load range, and the second load safety level is used to characterize the load of the transformer area through which the target power passes at the sampling time node as a safe load range.

[0077] In this embodiment of the invention, taking "City XX District No. 10 transformer station" as an example, the server acquires the instantaneous load data of a specific sampling time node (e.g., T1). To analyze load fluctuation, the server also acquires the instantaneous load data of at least one sampling time node adjacent to T1 (e.g., one minute before T1, T0, and one minute after T1, T2). The server then performs a difference calculation between the instantaneous load data at time T1 and the instantaneous load data at times T0 and T2, respectively. For example, if the load at time T1 is 100 kW, at time T0 it is 90 kW, and at time T2 it is 110 kW, then the load fluctuation change with T0 is 10 kW, and the load fluctuation change with T2 is also 10 kW. The server presets a threshold for load fluctuation change (e.g., 5 kW) to determine whether abnormal load fluctuations occur. In the above example, the load fluctuation change at both adjacent time nodes is 10 kW, exceeding the preset threshold of 5 kW. Because at least one load fluctuation exceeds a preset threshold, the server will determine the target load safety level at time T1 as the first load safety level. This indicates that at time T1, the power load of "Transformer Area No. 10, XX District, XX City" experienced abnormal fluctuations, posing a safety hazard. If the load fluctuations at all adjacent sampling time points do not exceed the preset threshold, the server will determine the target load safety level as the second load safety level, indicating that the power load of that transformer area is fluctuating within a safe range. Through these steps, the server can monitor the fluctuation of power load in real time, promptly detect and warn of potential power grid safety issues, thereby ensuring the stable operation of the power system.

[0078] In this embodiment of the invention, determining the load state of the corresponding transformer based on the number of nodes at the target sampling time node can be implemented through the following example.

[0079] The number of nodes at the target sampling time node is compared with the node number threshold to obtain the node number comparison result.

[0080] When the node number comparison result indicates that the number of nodes at the target sampling time node exceeds the node number threshold, the load state is determined as the safe load state.

[0081] When the node number comparison result indicates that the number of nodes at the target sampling time node does not exceed the node number threshold, the load state is determined as the abnormal load state.

[0082] In this embodiment of the invention, for example, the server first sets a node number threshold, which is determined based on historical data and the safe operation requirements of the power grid. For instance, assuming the server sets a node number threshold of 50, it means that in 100 consecutively sampled time points, at least 50 nodes must have power loads within a safe range for the overall load status to be considered safe. Subsequently, the server counts the number of target sampling time points that are determined to be within a safe load range over a period of time (e.g., 60 sampling points within one hour). Assuming that 55 sampling time points have power loads within a safe range during this period, the server then compares these 55 safe node numbers with the set node number threshold of 50. Since 55 is greater than 50, the node number comparison result indicates that the number of nodes at the target sampling time points exceeds the node number threshold. Based on the above comparison result, the server determines the load status of the transformer equipment as a safe load state. This means that during this period, the power load borne by the transformer equipment is mostly safe, and there is no sustained overload. Conversely, if the number of safe nodes is less than or equal to the node number threshold, for example, only 45 nodes are within the safe load range, the server will classify the load status as an abnormal load state. This means that the transformer equipment may have experienced significant overload during this period, posing a safety risk and requiring further inspection or adjustment. Through this analysis process, the server can monitor and accurately assess the load status of the transformer equipment in real time, providing crucial assurance for the safe and stable operation of the power grid.

[0083] In this embodiment of the invention, determining the load state of the corresponding transformer based on the number of nodes at the target sampling time node can be implemented through the following example.

[0084] The load stability rate of the target distribution area is obtained by dividing the number of nodes at the target sampling time node by the number of nodes at the sampling time node.

[0085] When the load stability rate exceeds the load stability rate threshold, the load state is determined as the safe load state; when the load stability rate does not exceed the load stability rate threshold, the load state is determined as the abnormal load state.

[0086] In this embodiment of the invention, taking the transformer equipment of "Transformer Area No. 10, XX District, XX City" as an example, the server sampled the transformer area 240 times in one day (10 times per hour, for a total of 24 hours), recording the power load data at each sampling time point. After analysis, the power load at 216 sampling time points was within the safe range and was marked as the target sampling time points. To evaluate the load stability of the transformer area, the server divides the number of these 216 target sampling time points by the total number of sampling time points, 240, to obtain the load stability rate. In this example, the load stability rate is 216 / 240 = 0.9, or 90%. The server presets a load stability rate threshold to determine whether the load status is safe. This threshold is usually determined based on historical data and the safety requirements of the power grid. Assuming that in this scenario, the load stability rate threshold is set to 85%, the calculated load stability rate of 90% is compared with the set load stability rate threshold of 85%. Since 90% is greater than 85%, the load stability rate exceeds the load stability rate threshold. Based on the above comparison results, the server will determine the load status of the transformer equipment in "Transformer Area No. 10, XX District, XX City" as a safe load status. This means that the power load of this transformer area remained within a safe range for most of the past day, exhibiting high stability. Conversely, if the calculated load stability rate is below 85%, for example, only 80%, the server will determine the load status as an abnormal load status. This means that the power load of the transformer area fluctuates significantly, posing a safety hazard and requiring further inspection or adjustment. Through this method, the server can objectively assess the load stability of the transformer equipment, promptly identify potential safety issues, and take corresponding measures to ensure the safe and stable operation of the power grid.

[0087] In this embodiment of the invention, after the training process is performed on the transformer area load assessment model based on the sample instance array, the following implementation methods are also provided.

[0088] Obtain the power dispatch plan for the target power, and search the load status of the base distribution area corresponding to the power dispatch plan from the distribution area load database to obtain the search results;

[0089] When the search result indicates that the load status of the benchmark transformer area is not stored in the transformer area load database, the target load data of the benchmark transformer area in a similar time period is obtained, and a feature extraction operation is performed on the target load data to obtain the target load features.

[0090] Using the trained transformer area load assessment model, based on the target load characteristics, the load of the benchmark transformer area is assessed to obtain the inferred load status of the benchmark transformer area.

[0091] The inferred load status of the reference transformer area is updated in the transformer area load database to obtain the updated transformer area load database.

[0092] In this embodiment of the invention, for example, the server first obtains the power dispatch plan for the target power, which details the power allocation and dispatch arrangements for a future period. Based on this plan, the server needs to search the load status of the corresponding base distribution area from the distribution area load database. For example, the power dispatch plan may specify "Distribution Area No. 5, District XX, City XX" as the base distribution area, and the load status during a specific time period (e.g., 9:00 AM to 11:00 AM) is the information the server needs to query. After performing the search, the server finds that the distribution area load database does not store the load status information for "Distribution Area No. 5, District XX, City XX" during the specified time period. This means that additional data processing is needed to fill this information gap. Since the database lacks directly relevant data, the server then searches for the target load data for "Distribution Area No. 5, District XX, City XX" within a similar time period (e.g., 8:00 AM to 10:00 AM). This data includes key indicators such as voltage, current, and power factor, which can indirectly reflect the load status of the distribution area. The server performs feature extraction operations on the obtained target load data. For example, by calculating characteristic values ​​such as average voltage, peak current, and power fluctuation, the server summarizes the load characteristics of the distribution area within a similar time period. These feature values ​​will help the load assessment model make accurate inferences. Next, the server uses the pre-trained distribution area load assessment model to infer the load status of "Distribution Area No. 5, XX District, XX City" within a specified time period based on the extracted target load features. The model will output a numerical value or classification label representing the load status, such as "Safe," "Warning," or "Danger." Finally, the server updates the inferred load status information to the distribution area load database. In this way, the server can directly use this updated data in future power dispatching and load assessment processes without having to make inferences again. Through this process, the server can not only fill the information gaps in the distribution area load database in a timely manner but also ensure the accuracy of power dispatching and load assessment, thereby maintaining the stable operation of the power system.

[0093] In this embodiment of the invention, the following implementation methods are also provided.

[0094] When the search result indicates that the load status of the reference area is stored in the area load database, and the load status of the reference area indicates that the area load of the reference area is within the abnormal load range, a first notification message is generated.

[0095] When the search result indicates that the load status of the reference distribution area is stored in the distribution area load database, and the load status of the reference distribution area indicates that the load of the reference distribution area is within the safe load range, a second notification message is generated.

[0096] The first notification message is used to notify the adjustment of the power dispatch plan in the load database of the distribution area, and the second notification message is used to notify the base distribution area corresponding to the power dispatch plan that the power load is normal.

[0097] In this embodiment of the invention, for example, the server first searches the distribution area load database for the load status of a specific benchmark distribution area. For example, for "Distribution Area No. 3, XX District, XX City", the server finds the corresponding load status record in the database. If the search result shows that the load status of "Distribution Area No. 3, XX District, XX City" is within an abnormal load range, such as excessively high load or abnormal fluctuations, the server immediately generates a first notification message. This message may include the specific location of the distribution area, detailed data on the abnormal load, possible cause analysis, and suggested adjustment measures. The server sends the first notification message to the power dispatch center or relevant management personnel. For example, the message may state: "Distribution Area No. 3, XX District, XX City is currently experiencing an abnormal load. Please adjust the power dispatch plan as soon as possible to reduce the load or conduct equipment checks." Upon receiving the notification, relevant personnel can react quickly to avoid potential power failures. On the other hand, if the search result shows that the load status of "Distribution Area No. 3, XX District, XX City" is within a safe load range, the server generates a second notification message. This message confirms the current power dispatch plan, indicating that the distribution area load is normal and no urgent adjustment is needed. The server sends the second notification message to the power dispatch center or management personnel for reference and recording. For example, a message could state: "The current load on transformer substation No. 3 in XX District, XX City is normal; the power dispatch plan can continue." Such notifications help maintain system transparency and management continuity. Through this process, the server can provide timely feedback based on actual conditions, ensuring the flexibility of the power dispatch plan and the safety of the transformer substation load. This real-time monitoring and notification mechanism is crucial for maintaining the stable operation of the power system.

[0098] In this embodiment of the invention, the aforementioned step S204 can be implemented through the following examples.

[0099] Identify the abnormal phenomena and multiple undetermined root causes of abnormalities under abnormal load conditions;

[0100] Based on the feature parameters corresponding to each of the undetermined root causes of the anomaly and the anomaly phenomenon, a sample instance array is constructed.

[0101] Iterative expert model training is performed based on the sample instance array to determine the target anomaly root cause from each of the undetermined anomaly root causes through multiple rounds of training. The sample instances in each subsequent round of expert model training include the feature parameters of the anomaly phenomenon and the feature parameters of the key anomaly root cause determined in the previous round of expert model training. The feature parameters of the key anomaly root cause determined in the previous round of expert model training differ from the feature parameter distribution of the anomaly root cause in the sample instances of the previous round of expert model training. The correlation coefficient between the key anomaly root cause and the anomaly phenomenon conforms to a preset correlation coefficient threshold.

[0102] The correlation and matching relationship between the abnormal phenomenon and the target abnormal root cause is used as the abnormal state analysis result of the abnormal load state.

[0103] In this embodiment of the invention, for example, the server first detects an abnormal load state in a certain transformer substation (e.g., "Substation No. 7, XX District, XX City"), manifested as a sudden increase in power load and voltage fluctuations. Next, the server lists several potential root causes that could lead to this abnormality, such as equipment failure, abnormal external power grid supply, or changes in user electricity consumption behavior. For each potential root cause and abnormal phenomenon, the server collects relevant feature parameters. For example, for the potential root cause of equipment failure, feature parameters may include equipment temperature, operating time, and historical fault records; for abnormal external power grid supply, feature parameters may include grid voltage fluctuations and frequency changes. These feature parameters and abnormal phenomena (such as increased load and voltage fluctuations) together constitute a sample instance array. The server uses these sample instance arrays for iterative expert model training. In each round of training, the model learns the relationship between abnormal phenomena and potential root causes. Importantly, each round of training utilizes the feature parameters of key root causes identified in the previous round of training, and the correlation coefficients between these key root causes and abnormal phenomena meet a preset threshold. In this way, the model can gradually focus on the root causes of anomalies that are more strongly correlated with the abnormal phenomena. For example, in the first round of training, the model finds that both equipment failure and abnormal power supply from the external grid are related to abnormal load conditions, but the characteristic parameters of equipment failure have a higher correlation with the abnormal phenomena. In the second round of training, the model will consider the characteristic parameters of equipment failure more, further refining the analysis. After multiple rounds of training, the model will identify a target root cause of the anomaly with the highest correlation to the abnormal phenomena. In this example, the model can ultimately determine that equipment failure is the main cause of the abnormal load condition of "Transformer Area No. 7, XX District, XX City". Finally, the server will use the correlation matching relationship between the abnormal phenomena (sudden increase in load, voltage fluctuation) and the target root cause of the anomaly (equipment failure) as the anomaly state analysis result of the abnormal load condition. This result can help power workers quickly locate problems and take corresponding maintenance measures.

[0104] In this embodiment of the invention, the abnormal phenomena of abnormal load states and multiple undetermined root causes of abnormalities can be determined by the following example.

[0105] Obtain an anomaly analysis indication for an abnormal load state, and extract the abnormal phenomena of the abnormal load state from the anomaly analysis indication;

[0106] From the preset hypergraph where the abnormal phenomenon is located, determine multiple undetermined root causes of the abnormal load state.

[0107] In an embodiment of the invention, for example, the server receives an alarm signal regarding abnormal load in "Transformer N, YY District, XX City," which is an anomaly analysis indication. This indication may originate from smart meters, sensors, or other monitoring systems within the transformer area. The server analyzes and extracts specific anomalies from the anomaly analysis indication. For example, the indication may include information such as "Transformer N, YY District, XX City has experienced a continuous increase in load and voltage fluctuations exceeding the normal range over the past 30 minutes." The server identifies these specific manifestations as anomalies, namely, abnormal load increase and abnormal voltage fluctuations. The server then consults a pre-defined hypergraph, which is constructed based on historical data and expert knowledge, containing complex relationships between various possible anomalies and potential root causes. In this hypergraph, nodes may represent different devices, environmental factors, or user behaviors, while hyperedges represent the associations or causal relationships between these nodes. Based on the anomalies extracted from the anomaly analysis indication (abnormal load increase and abnormal voltage fluctuations), the server searches for nodes associated with these phenomena in the pre-defined hypergraph. For example, factors that may lead to increased load and voltage fluctuations include: internal equipment failure within the distribution area, abnormal external power supply, and sudden changes in user electricity consumption behavior. The server lists these factors as potential root causes of anomalies because they could all be the cause of the current abnormal phenomenon. Through the above steps, the server can quickly identify abnormal phenomena in abnormal load states and determine multiple possible potential root causes of anomalies based on anomaly analysis indicators and preset hypergraphs. This provides important clues and directions for subsequent troubleshooting and problem solving.

[0108] In this embodiment of the invention, determining multiple undetermined root causes of the abnormal load state from a preset hypergraph where the abnormal phenomenon is located can be implemented through the following example.

[0109] Determine the distribution system to which the abnormal load state belongs, and obtain a preset hypermap of the distribution system; each entity in the preset hypermap contains the abnormal phenomenon entity where the abnormal phenomenon is located, and potential abnormal root cause entities other than the abnormal phenomenon entity.

[0110] Based on the entity relationship corresponding to each entity in the preset hypergraph, multiple key anomaly root cause entities are determined from each potential anomaly root cause entity;

[0111] The abnormal factors characterized by the key abnormal root cause entity are identified as the undetermined abnormal root causes of the abnormal load state.

[0112] In this embodiment of the invention, for example, firstly, the server needs to determine the specific transformer substation system where the abnormal load state occurred. For example, suppose the server detects an abnormal load state in "Transformer Substation No. M in District ZZ of XX City". Subsequently, the server retrieves a preset hypergraph corresponding to this transformer substation system from its database. This preset hypergraph is a complex relational network containing various possible abnormal phenomenon entities (such as voltage anomalies, overload, etc.) and associated potential root cause entities (such as equipment failure, line aging, changes in user electricity consumption behavior, etc.). The server begins to analyze the relationships between the entities in the preset hypergraph. These relationships can be causal, correlational, or conditional, etc. For example, in the hypergraph, the voltage anomaly entity may be connected to multiple potential root cause entities such as transformer failure entity and poor line contact entity. By analyzing these relationships, the server can initially understand which potential root causes may be related to the current abnormal phenomenon. Next, the server, based on the entity relationships in the preset hypergraph, historical data, expert knowledge, etc., filters out several key root cause entities from the numerous potential root cause entities. These key root cause entities are the causes with a high degree of correlation and high probability to the current abnormal phenomenon. For example, when analyzing voltage anomalies, the server identified transformer faults and poor line contact as two key root cause entities because they had caused similar voltage anomalies multiple times in the past. Finally, the server determined the anomaly factors represented by the selected key root cause entities as potential root causes. These potential root causes will serve as targets for further investigation and verification. In this example, "transformer fault" and "poor line contact" were identified as potential root causes leading to the voltage anomaly at "Transformer No. M, ZZ District, XX City". Through these steps, the server can quickly and accurately identify multiple potential root causes related to abnormal load conditions based on a pre-defined hypergraph and entity relationship analysis, providing strong support for subsequent problem-solving.

[0113] In this embodiment of the invention, a sample instance array is constructed based on the feature parameters corresponding to each of the undetermined root causes of the anomaly and the anomaly phenomenon. This can be implemented through the following example.

[0114] Standardize the characteristic parameters corresponding to each of the undetermined root causes of the anomaly and the anomaly phenomenon to obtain standardized data;

[0115] Perform feature extraction on the standardized data to obtain abnormal factor features that meet the training requirements of the expert model;

[0116] Construct an array of sample instances containing the features of the aforementioned anomalous factors.

[0117] In this embodiment of the invention, for example, the server first collects feature parameters related to each potential root cause and anomaly. For instance, for the potential root cause "equipment aging," relevant feature parameters may include equipment runtime, number of maintenance records, equipment temperature, etc.; while for the anomaly "voltage anomaly," feature parameters may include voltage fluctuation range, anomaly duration, etc. Next, the server standardizes these raw feature parameters to eliminate differences in dimensions and numerical ranges between different feature parameters. Standardization methods may include min-max standardization, Z-score standardization, etc. Standardization ensures that all feature parameters are within the same numerical range, facilitating subsequent model training. After standardization, the server performs feature extraction on the standardized data. The purpose of feature extraction is to extract information useful for model training while reducing the dimensionality and complexity of the data. For example, for the two feature parameters, equipment runtime and number of maintenance records, the server extracts more representative features by calculating their statistics (such as mean, standard deviation, etc.) or applying methods such as principal component analysis (PCA). During this process, the server adjusts its feature extraction strategy based on expert knowledge and historical data to ensure that the extracted anomaly features meet the requirements for expert model training. Finally, the server uses the extracted anomaly features to construct an array of sample instances. Each sample instance contains feature parameter values ​​related to a specific undetermined root cause and anomaly phenomenon. For example, a sample instance might contain feature parameter values ​​related to equipment aging (such as standardized runtime, maintenance record statistics, etc.) and feature parameter values ​​related to voltage anomalies (such as standardized voltage fluctuation range, etc.). These arrays of sample instances will be used for subsequent expert model training to help the model learn how to identify root causes and predict anomalies based on the feature parameters. Through these steps, the server can construct an array of sample instances that meets the requirements for expert model training based on the feature parameters corresponding to each undetermined root cause and anomaly phenomenon. This provides an important data foundation for subsequent model training and root cause identification.

[0118] In this embodiment of the invention, iterative expert model training is performed based on the sample instance array to determine the target abnormal root cause among each of the undetermined abnormal root causes after multiple rounds of cycles. This can be implemented through the following example.

[0119] The target expert model is obtained by training an expert model on the array of sample instances.

[0120] Based on the target expert model, the abnormal phenomenon is analyzed for its root causes, and the correlation coefficient between each of the undetermined root causes and the abnormal phenomenon is determined.

[0121] From each of the undetermined root causes of anomalies, identify the key root causes whose correlation coefficients meet the preset correlation coefficient threshold;

[0122] Based on the key anomaly root cause, update the feature parameter distribution of the sample instances, determine the adjusted sample instance array, and repeat the step of training the expert model on the sample instance array until a preset termination state is met to obtain the target anomaly root cause; the adjusted sample instance array includes the key anomaly root cause and the feature parameters corresponding to the anomaly phenomenon.

[0123] In this embodiment of the invention, for example, the server first trains an expert model using an initial array of sample instances. This array contains feature parameters of various undetermined root causes and abnormal phenomena. For example, the sample instance array may contain feature parameters such as device temperature, grid voltage fluctuations, and changes in user electricity consumption behavior. Using this data, the server trains an initial expert model, which can be called a "target expert model." Once the target expert model is trained, the server uses this model to analyze the root causes of abnormal phenomena. The model analyzes the correlation coefficient between each undetermined root cause and the abnormal phenomenon. For example, the model finds a high correlation coefficient between excessively high device temperature and abnormal voltage, while a low correlation coefficient between changes in user electricity consumption behavior and abnormal voltage. The server then identifies key root causes from each undetermined root cause whose correlation coefficient meets a preset correlation coefficient threshold. In this example, if the correlation coefficient between excessively high device temperature and abnormal voltage exceeds the preset threshold, then "excessively high device temperature" is identified as a key root cause. After identifying the key root causes, the server updates the feature parameter distribution of the sample instances based on these key root causes. This means that in the next round of expert model training, the server will pay more attention to feature parameters related to the key anomaly root cause. For example, if "device temperature too high" is identified as the key anomaly root cause, then in the next round of sample instance array, feature parameters related to device temperature will be given higher weights. Subsequently, the server will repeat the expert model training steps using the updated sample instance array. This process will iterate until a preset termination state is met, such as reaching a preset number of iterations or the model performance improvement no longer being significant. After multiple iterations, the server will finally determine the target anomaly root cause with the highest correlation to the anomaly. In this example, if "device temperature too high" is consistently identified as the key anomaly root cause by the model after multiple iterations, and its correlation coefficient with voltage anomaly remains at a high level, then the server will ultimately determine "device temperature too high" as the target anomaly root cause. Through this iterative expert model training method, the server can gradually focus on the key factors that truly cause the anomaly, thereby improving the accuracy and efficiency of anomaly root cause identification.

[0124] In this embodiment of the invention, the target expert model is obtained by training the sample instance array. This can be implemented through the following example.

[0125] Determine the feature parameter attribute corresponding to each feature parameter in the sample instance array;

[0126] Using a core training architecture associated with each of the aforementioned feature parameter attributes, an expert model is trained on the array of sample instances to obtain the target expert model.

[0127] In this embodiment of the invention, for example, the server first analyzes each feature parameter in the sample instance array to determine their attributes. For example, for the feature parameter "equipment temperature," its attributes may include "continuous numerical value," "real-time change," etc.; while for the feature parameter "equipment model," its attributes may include "discrete classification," "static information," etc. Based on the attributes of each feature parameter, the server selects the associated core training architecture. For example, for continuous numerical features, the server selects a neural network architecture suitable for regression problems, such as a deep neural network (DNN) or a long short-term memory network (LSTM), to capture the complex nonlinear relationships between feature parameters. For discrete classification features, the server selects an architecture suitable for classification problems, such as a support vector machine (SVM) or a decision tree. Using the selected core training architecture, the server begins expert model training on the sample instance array. Taking equipment temperature and equipment model as examples, the server uses the training architecture for continuous numerical features to train a model that can predict the impact of equipment temperature on the power grid load, and simultaneously uses the training architecture for discrete classification features to train a model that can predict the failure probability of the equipment model. After multiple rounds of training and optimization, the server will eventually obtain one or more target expert models. These models can perform efficient prediction and analysis based on specific feature parameter attributes. For example, for continuously monitored equipment temperature data, the target expert model can predict its impact on the grid load in real time, thereby adjusting the energy storage system's scheduling strategy in a timely manner; for static information such as equipment models, the target expert model can assess the failure risk of different equipment models, providing decision support for equipment maintenance and replacement. Through the above steps, the server can select appropriate training architectures for training expert models for different types of feature parameters, thereby obtaining more accurate and efficient target expert models. These models play an important role in the intelligent scheduling of distributed energy storage in autonomous distribution areas, helping to improve system stability and operational efficiency.

[0128] In this embodiment of the invention, the target expert model is obtained by training the sample instance array. This can be implemented through the following example.

[0129] The sample instance array is divided into multiple sub-sample instance arrays; each sub-sample instance array includes the feature parameters of the abnormal phenomenon and at least a portion of the feature parameters of the undetermined root cause of the abnormality.

[0130] Each of the sub-sample instance arrays is trained with an expert model to obtain the target expert model associated with each sub-sample instance array.

[0131] In this embodiment of the invention, for example, the server first divides the entire sample instance array into multiple sub-sample instance arrays. This is done to distribute training pressure, improve training efficiency, and also to enable more refined model training for different types of anomaly root causes. For instance, the server can divide feature parameters related to equipment failure into one sub-sample instance array and feature parameters related to changes in the external environment (such as weather and temperature) into another sub-sample instance array. Each sub-sample instance array contains feature parameters of the abnormal phenomenon and at least some feature parameters of the undetermined anomaly root causes. Taking equipment failure as an example, the relevant sub-sample instance array may contain feature parameters such as equipment operating status, fault history, and equipment usage time, as well as feature parameters related to anomalies that may be caused by equipment failure (such as voltage fluctuations and load anomalies). The server then trains expert models for each sub-sample instance array separately. This means that the server trains different target expert models for different types of anomaly root causes. For example, for root causes of anomalies like equipment failure, the server trains an expert model capable of identifying precursors and predicting the impact of failures. For root causes like changes in the external environment, the server trains another expert model capable of analyzing the relationship between external environmental factors and the power grid's operating status. After training, the server obtains the target expert model associated with each sub-sample instance array. These models are optimized and trained for different types of root causes, thus achieving higher accuracy and efficiency in handling corresponding anomalies. For instance, when the server detects voltage fluctuations in the power grid, it can use the target expert model related to equipment failure to quickly determine if the anomaly is caused by equipment failure and take appropriate countermeasures. Through this divide-and-conquer strategy, the server can more effectively utilize limited computing resources for model training and provide more accurate prediction and analysis capabilities for different types of root causes.

[0132] In this embodiment of the invention, the abnormal phenomenon is analyzed based on the target expert model to determine the correlation coefficient between each of the undetermined abnormal root causes and the abnormal phenomenon. This can be implemented through the following example.

[0133] Based on the target expert model, the abnormal phenomenon is analyzed for its root causes, and the initial correlation coefficient between each of the undetermined root causes and the abnormal phenomenon is determined.

[0134] For each of the undetermined root causes of anomalies, feature masking is performed on the feature parameters of the undetermined root causes of anomalies to obtain masked data;

[0135] Based on the characteristic parameters of the anomaly and the deviation between them and the output response of the target expert model to the masked data, the model stability evaluation value of the undetermined anomaly root cause for the target expert model is determined.

[0136] By combining the initial correlation coefficient and the model stability evaluation value, the correlation coefficient between the undetermined root cause of the anomaly and the anomaly phenomenon is determined.

[0137] In this embodiment of the invention, for example, the server first uses a pre-trained target expert model to perform root cause analysis on the occurring anomaly. For instance, when a voltage drop occurs in the power grid, the server uses model analysis to preliminarily determine the initial correlation coefficients between each potential root cause of the anomaly (such as equipment failure, line aging, external interference, etc.) and this anomaly. These initial correlation coefficients reflect the probability that each potential root cause of the anomaly will lead to the anomaly under the model training data. Next, the server performs masking processing on the feature parameters of each potential root cause of the anomaly. This is equivalent to "shielding" the influence of the root cause of the anomaly in the model to observe the model's performance without this factor. For example, if "equipment failure" is a potential root cause of the anomaly, the server will mask its related feature parameters (such as equipment operating status, historical fault records, etc.) to generate masked data. The server then inputs the masked data into the target expert model and observes the model's output response. By comparing the deviation between the characteristic parameters of the anomaly (such as the magnitude and duration of the voltage drop) and the model's output response to the masked data, the impact of the undetermined root cause on model stability can be assessed. A smaller deviation indicates a smaller impact of the root cause on the model output, and vice versa. This deviation is defined as the model stability assessment value. Finally, the server combines the initial correlation coefficient and the model stability assessment value to comprehensively determine the final correlation coefficient between each undetermined root cause and the anomaly. For example, if the initial correlation coefficient for the root cause "equipment failure" is high, but the deviation between the model output and the anomaly (i.e., the model stability assessment value) is small after masking its characteristic parameters, this means that although equipment failure and voltage drop are related, it is not the main cause of the anomaly. Therefore, under comprehensive consideration, the server will adjust the correlation coefficient between "equipment failure" and the voltage drop anomaly. Through these steps, the server can more accurately identify the key factors causing the anomaly, providing strong support for subsequent troubleshooting and repair.

[0138] In this embodiment of the invention, the abnormal phenomenon is analyzed based on the target expert model to determine the initial correlation coefficient between each of the undetermined abnormal root causes and the abnormal phenomenon. This can be implemented through the following example.

[0139] The feature parameters corresponding to each of the undetermined root causes of anomalies are used as the input of the target expert model, and the feature parameters of the anomaly are used as the output of the target expert model. The expected influence degree of the input of the target expert model is calculated through backpropagation. The expected influence degree of the input is characterized by the expected superposition influence degree corresponding to each of the undetermined root causes of anomalies.

[0140] Based on the expected superimposed influence degree of each of the aforementioned factors, an initial correlation coefficient between each of the undetermined root causes of anomalies and the aforementioned anomalies is determined; the initial correlation coefficient is matched with the expected superimposed influence degree of the undetermined root causes of anomalies.

[0141] In this embodiment of the invention, for example, the server uses the feature parameters corresponding to each undetermined root cause of anomaly as input to the target expert model, such as equipment temperature, grid voltage fluctuations, and line aging. Simultaneously, the feature parameters of the abnormal phenomenon are used as the model's output, such as voltage anomalies or load anomalies in the power grid. To determine the degree of influence of the input (undetermined root cause of anomaly) on the output (abnormal phenomenon), the server employs a backpropagation algorithm. This method is typically used in the training process of neural networks, but here it is used to analyze the expected influence of the input on the output in a trained model. Specifically, the server starts from the model's output layer and calculates the contribution of each input feature parameter to the output in reverse; this is called the expected influence of the input. This influence is characterized by the expected superimposed influence corresponding to each undetermined root cause of anomaly. During backpropagation, the server calculates the expected superimposed influence of each undetermined root cause of anomaly on the output. For example, for the root cause of "overheating equipment temperature," the server analyzes its impact on the voltage anomaly and calculates a specific expected superimposed influence value. This value reflects the potential impact of "overheating equipment temperature" on the voltage anomaly in the model's prediction. Based on the expected cumulative impact of each potential root cause of an anomaly, the server determines its initial correlation coefficient with the anomaly phenomenon. This correlation coefficient matches the expected cumulative impact; that is, the greater the impact of the root cause, the higher its initial correlation coefficient. For example, if the expected cumulative impact of "equipment overheating" is large, its initial correlation coefficient with voltage anomaly will also be high, indicating a strong correlation between the two. Through these steps, the server can quantify the initial correlation coefficient between each potential root cause of an anomaly and the anomaly phenomenon, providing important reference for subsequent root cause analysis and fault handling.

[0142] In this embodiment of the invention, the key abnormal root cause whose correlation coefficient meets the preset correlation coefficient threshold is determined from each of the undetermined abnormal root causes, which can be implemented through the following example.

[0143] Determine the preset correlation coefficient threshold associated with the current training cycle;

[0144] The correlation coefficient corresponding to each of the undetermined root causes of anomalies is compared with the preset correlation coefficient threshold, and the key root causes of anomalies whose correlation coefficients meet the preset correlation coefficient threshold are determined from each of the undetermined root causes of anomalies.

[0145] In this embodiment of the invention, for example, at the beginning of each training cycle, the server sets a preset correlation coefficient threshold associated with the current training cycle based on historical data and experience. This threshold is used to determine which potential root causes of anomalies have a sufficiently strong correlation with the abnormal phenomenon, and are thus considered key root causes of anomalies. For example, the server can set the preset correlation coefficient threshold to 0.8, meaning that only when the correlation coefficient between a potential root cause of anomalies and an abnormal phenomenon reaches or exceeds 0.8 will it be identified as a key root cause of anomalies. The server calculates the correlation coefficient between each potential root cause of anomalies and the abnormal phenomenon, and compares these correlation coefficients with the preset correlation coefficient threshold. For example, the server calculates that the correlation coefficient between the potential root cause of "equipment aging" and the phenomenon of "abnormal power grid load" is 0.9, while the correlation coefficient between "external environmental temperature change" and the same abnormal phenomenon is 0.7. Based on the comparison results, the server determines those key root causes of anomalies whose correlation coefficients meet (i.e., are greater than or equal to) the preset correlation coefficient threshold from the potential root causes of anomalies. In the example above, the correlation coefficient of "equipment aging" (0.9) exceeds the preset correlation coefficient threshold of 0.8, therefore it is identified as a key root cause of "abnormal power grid load." The correlation coefficient of "external ambient temperature change" (0.7) is below the threshold and is therefore not considered a key root cause. Through these steps, the server can accurately identify the key root causes most strongly correlated with the abnormal phenomenon from multiple potential root causes, which is crucial for quickly locating and resolving practical problems in power grid operation.

[0146] This invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned distributed energy storage intelligent scheduling method based on regional autonomy. Figure 2 As shown, Figure 2 This is a structural block diagram of a computer device 100 provided in an embodiment of the present invention. The computer device 100 includes a memory 111, a processor 112, and a communication unit 113. To enable data transmission or interaction, the memory 111, processor 112, and communication unit 113 are electrically connected to each other directly or indirectly. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.

[0147] For illustrative purposes, the foregoing description has been made with reference to specific embodiments. However, the foregoing illustrative discussions are not intended to be exhaustive or to limit the present disclosure to the precise forms disclosed. Numerous modifications and variations are possible in accordance with the foregoing teachings. These embodiments were chosen and described in order to best illustrate the principles of the present disclosure and its practical application, thereby enabling those skilled in the art to best utilize the disclosure and to employ various embodiments with different modifications to suit a particular intended application.

Claims

1. A distributed energy storage intelligent scheduling method based on transformer area autonomy, characterized in that, include: Obtain the task instruction for the current power supply to the target transformer area; According to the task instructions, obtain the current power load data; The current power load data is input into a pre-trained transformer area load assessment model to obtain the corresponding load status assessment results. When the load status assessment result is characterized as an abnormal load status, the target abnormal root cause with the highest correlation to the abnormal phenomenon of the abnormal load status is determined, and the abnormal status analysis result is obtained. Based on the load status assessment results and the abnormal status analysis results, a target scheduling strategy is selected from the pre-configured energy storage scheduling. The target scheduling strategy is distributed to the distributed energy storage devices included in the target distribution area for execution; The transformer area load assessment model is obtained through the following methods: The power load data recorded during the transmission of the target power to the target transformer area is acquired, and the power load data is associated with transformer equipment to obtain sub-power load data corresponding to multiple transformer equipment; the sub-power load data includes the instantaneous load data corresponding to each sampling time node during the transmission of the target power to the corresponding transformer equipment. Based on each of the sub-power load data, the load status corresponding to the transformer equipment is determined, and the load status includes safe load status and abnormal load status; Using each of the sub-power load data as a sample instance and the corresponding load state as the instance target value, a sample instance array is constructed for training the transformer area load assessment model, and the following processing is performed on each of the sub-power load data in the sample instance array: Obtain the load mean, root mean square load value, load fluctuation, load dispersion, and peak data in each instantaneous load data. Divide the load dispersion by the load mean to obtain the rate of change deviation of the sub-power load data; divide the peak data by the root mean square load value to obtain the peak-to-average ratio of the sub-power load data; and divide the peak data by the load mean to obtain the impact ratio of the sub-power load data. The average load, the load fluctuation, the rate of change deviation, the peak-to-average ratio, and the impact ratio are integrated and processed to obtain the load time series transformation domain data corresponding to the load time series analysis domain. Obtain the mean values ​​of the spectral distribution intensity and power spectral density corresponding to the sub-power load data, and determine the spectral center frequency of the sub-power load data in the energy spectrum based on the spectral distribution intensity; By combining the center frequency of the spectrum and the intensity of the spectrum distribution, the spectral fluctuation of the sub-power load data on the energy spectrum is determined, and the upper limit frequency of the spectrum in the intensity of the spectrum distribution is obtained; The spectral distribution intensity, the mean power spectral density, the spectral center frequency, the spectral fluctuation, and the upper limit frequency of the spectral spectrum are integrated and processed to obtain the energy spectral conversion domain data corresponding to the energy spectral analysis domain; From the power quality fluctuation analysis domain, the sub-power load data is subjected to power quality fluctuation mapping processing to obtain the power quality fluctuation conversion domain data corresponding to the power quality fluctuation analysis domain. Perform feature extraction operation on each of the transformed domain data to obtain the transformed domain features corresponding to each of the transformed domain data; Obtain the feature domain attributes corresponding to each of the transformation domain features, and perform a weighted average of each of the feature domain attributes to obtain the reference domain attributes; Each of the feature domain attributes is compared with the reference domain attribute to obtain the attribute comparison result corresponding to each feature domain attribute; When the attribute comparison result indicates that the feature domain attribute is consistent with the reference domain attribute, the corresponding transformation domain feature is determined as the target transformation domain feature corresponding to the transformation domain feature; When the attribute comparison result indicates that the feature domain attribute is inconsistent with the reference domain attribute, the feature domain attribute of the corresponding transformation domain feature is updated to the reference domain attribute to obtain the target transformation domain feature corresponding to the transformation domain feature. Each of the target transformation domain features is subjected to feature integration processing to obtain integrated features; Using the aforementioned transformer area load assessment model and based on the integrated features, the target transformer area is assessed for load to obtain the inferred load status of the target transformer area. The training process is performed on the transformer area load assessment model by combining the inferred load status and the corresponding load status; The transformer area load assessment model is used to assess the load status of the transformer area where the target power is located based on the power load data of the target power.

2. The method according to claim 1, characterized in that, The acquisition of power load data recorded during the transmission of target power to the target distribution area includes: The voltage fluctuations and current changes at each sampling time point during the transmission of the target power to the target distribution area are obtained. For each sampling time node, the standardized values ​​of the corresponding voltage fluctuation and the standardized values ​​of the current change are integrated to obtain the instantaneous load data corresponding to the sampling time node; The power load data is obtained by arranging each instantaneous load data according to the time sequence of the sampling time nodes.

3. The method according to claim 1, characterized in that, The sub-power load data includes instantaneous load data corresponding to each sampling time point during the transmission of the target power to the corresponding transformer equipment. Determining the load state of the corresponding transformer equipment based on each sub-power load data includes: For each of the aforementioned sub-power load data, the following processing is performed: For each sampling time node corresponding to the sub-power load data, at least one adjacent sampling time node is obtained, and the instantaneous load data corresponding to the sampling time node is respectively compared with the instantaneous load data corresponding to each adjacent sampling time node to obtain the load fluctuation change corresponding to each adjacent sampling time node. When at least one of the load fluctuation changes exceeds a preset change threshold, the target load safety level is determined as the first load safety level. When no load fluctuation exceeds the preset change threshold, the target load safety level is determined as the second load safety level. Wherein, the first load safety level is used to characterize the load of the transformer area through which the target power passes at the sampling time node as an abnormal load range, and the second load safety level is used to characterize the load of the transformer area through which the target power passes at the sampling time node as a safe load range; When the target load safety level indicates that the load of the target distribution area through which the target power passes at the sampling time node is within the safe load range, the sampling time node corresponding to the target load safety level is determined as the target sampling time node; Based on the number of nodes at the target sampling time node, the load state corresponding to the transformer is determined.

4. The method according to claim 3, characterized in that, Determining the load state corresponding to the transformer based on the number of nodes at the target sampling time node includes: The number of nodes at the target sampling time node is compared with the node number threshold to obtain the node number comparison result. When the node number comparison result indicates that the number of nodes at the target sampling time node exceeds the node number threshold, the load state is determined as the safe load state. When the node number comparison result indicates that the number of nodes at the target sampling time node does not exceed the node number threshold, the load state is determined as the abnormal load state. The step of determining the load state corresponding to the transformer based on the number of nodes at the target sampling time node further includes: The load stability rate of the target distribution area is obtained by dividing the number of nodes at the target sampling time node by the number of nodes at the sampling time node. When the load stability rate exceeds the load stability rate threshold, the load state is determined as the safe load state; when the load stability rate does not exceed the load stability rate threshold, the load state is determined as the abnormal load state.

5. The method according to claim 1, characterized in that, After combining the inferred load state and the corresponding load state to perform a training process on the transformer area load assessment model, the method further includes: Obtain the power dispatch plan for the target power, and search the load status of the base distribution area corresponding to the power dispatch plan from the distribution area load database to obtain the search results; When the search result indicates that the load status of the benchmark transformer area is not stored in the transformer area load database, the target load data of the benchmark transformer area in a similar time period is obtained, and a feature extraction operation is performed on the target load data to obtain the target load features. Using the trained transformer area load assessment model, based on the target load characteristics, the load of the benchmark transformer area is assessed to obtain the inferred load status of the benchmark transformer area. The inferred load status of the reference transformer area is updated in the transformer area load database to obtain the updated transformer area load database.

6. The method according to claim 5, characterized in that, The method further includes: When the search result indicates that the load status of the reference area is stored in the area load database, and the load status of the reference area indicates that the area load of the reference area is within the abnormal load range, a first notification message is generated. When the search result indicates that the load status of the reference distribution area is stored in the distribution area load database, and the load status of the reference distribution area indicates that the load of the reference distribution area is within the safe load range, a second notification message is generated. The first notification message is used to notify the adjustment of the power dispatch plan in the load database of the distribution area, and the second notification message is used to notify the base distribution area corresponding to the power dispatch plan that the power load is normal.

7. The method according to claim 1, characterized in that, The process of identifying the target anomaly root cause with the highest correlation to the abnormal phenomenon of the abnormal load state, and obtaining the anomaly state analysis results, includes: Obtain an anomaly analysis indication for an abnormal load state, and extract the abnormal phenomena of the abnormal load state from the anomaly analysis indication; Determine the distribution system to which the abnormal load state belongs, and obtain a preset hypermap of the distribution system; each entity in the preset hypermap contains the abnormal phenomenon entity where the abnormal phenomenon is located, and potential abnormal root cause entities other than the abnormal phenomenon entity. Based on the entity relationship corresponding to each entity in the preset hypergraph, multiple key anomaly root cause entities are determined from each of the potential anomaly root cause entities. The abnormal factors characterized by the key abnormal root cause entity are identified as the undetermined abnormal root causes of the abnormal load state. Standardize the characteristic parameters corresponding to each of the undetermined root causes of the anomaly and the anomaly phenomenon to obtain standardized data; Perform feature extraction on the standardized data to obtain abnormal factor features that meet the training requirements of the expert model; Construct an array of sample instances containing the features of the aforementioned anomaly factors; The target expert model is obtained by training an expert model on the array of sample instances. The feature parameters corresponding to each of the undetermined root causes of anomalies are used as the input of the target expert model, and the feature parameters of the anomaly are used as the output of the target expert model. The expected influence degree of the input of the target expert model is calculated through backpropagation. The expected influence degree of the input is characterized by the expected superposition influence degree corresponding to each of the undetermined root causes of anomalies. Based on the expected superimposed influence degree of each of the aforementioned factors, an initial correlation coefficient between each of the undetermined root causes of anomalies and the aforementioned anomalies is determined; the initial correlation coefficient is matched with the expected superimposed influence degree of the undetermined root causes of anomalies. For each of the undetermined root causes of anomalies, feature masking is performed on the feature parameters of the undetermined root causes of anomalies to obtain masked data; Based on the characteristic parameters of the anomaly and the deviation between them and the output response of the target expert model to the masked data, the model stability evaluation value of the undetermined anomaly root cause for the target expert model is determined. By combining the initial correlation coefficient and the model stability evaluation value, the correlation coefficient between the undetermined root cause of the anomaly and the anomaly phenomenon is determined; Determine the preset correlation coefficient threshold associated with the current training cycle; The correlation coefficient corresponding to each of the undetermined root causes of anomalies is compared with the preset correlation coefficient threshold, and the key root causes of anomalies whose correlation coefficients meet the preset correlation coefficient threshold are determined from each of the undetermined root causes of anomalies. Based on the key anomaly root cause, the feature parameter distribution of the updated sample instances is determined, and the adjusted sample instance array is determined. The step of training the expert model on the sample instance array is repeated until a preset termination state is met, thus obtaining the target anomaly root cause. The adjusted sample instance array includes the key anomaly root cause and the feature parameters corresponding to the anomaly phenomenon. The sample instances in the next round of expert model training include the feature parameters of the anomaly phenomenon and the feature parameters of the key anomaly root cause determined in the previous round of expert model training. The feature parameters of the key anomaly root cause determined in the previous round of expert model training are different from the feature parameter distribution of the anomaly root cause in the sample instances of the previous round of expert model training. The correlation coefficient between the key anomaly root cause and the anomaly phenomenon meets a preset correlation coefficient threshold. The correlation and matching relationship between the abnormal phenomenon and the target abnormal root cause is used as the abnormal state analysis result of the abnormal load state.

8. The method according to claim 7, characterized in that, The step of training an expert model on the sample instance array to obtain a target expert model includes: Determine the feature parameter attribute corresponding to each feature parameter in the sample instance array; Using a core training architecture associated with each of the aforementioned feature parameter attributes, an expert model is trained on the array of sample instances to obtain the target expert model; The step of training an expert model on the sample instance array to obtain a target expert model further includes: The sample instance array is divided into multiple sub-sample instance arrays; each sub-sample instance array includes the feature parameters of the abnormal phenomenon and at least a portion of the feature parameters of the undetermined root cause of the abnormality. Each of the sub-sample instance arrays is trained with an expert model to obtain the target expert model associated with each sub-sample instance array.

9. A server system, characterized in that, Includes a server, the server being used to perform the method according to any one of claims 1-8.

Images