A method and system for identifying the heating load of a transformer area

Abstract

The application discloses a kind of table area electric heating load identification method and system, belong to power grid data analysis and intelligent identification technical field.The method includes: obtaining the historical load of to-be-identified table area, time and multi-dimensional time series data such as meteorology by data acquisition module;Utilize a deep learning model comprising time series feature extraction module and key feature weighting module to process data to generate preliminary classification results;Based on the business rules of the growth of heating season electricity consumption and the stability of non-heating season electricity consumption, the historical load is compared and verified across years;The preliminary classification result of model is combined with the verification result of rule, and whether table area is electric heating load is comprehensively judged.This application combines the pattern recognition ability of deep learning with the deterministic logic of business rules, realizes the high-precision, high-robustness automatic identification of electric heating load, and provides an effective technical means for power grid intelligent management.

Description

Technical Field

[0001] This invention belongs to the field of power grid data analysis technology, and in particular relates to a method and system for identifying electric heating loads in transformer substations. Background Technology

[0002] The application of new energy sources, with electricity substitution as the core, is becoming increasingly widespread. Among them, the large-scale promotion of new energy heating methods such as "coal-to-electricity" in northern regions has led to significant changes in the characteristics of power grid load. This type of heating load typically exhibits a sharp increase in electricity consumption during winter, especially at night, showing seasonal and temporal characteristics that are quite different from traditional residential and industrial loads.

[0003] Accurately and efficiently identifying new energy heating areas is of vital importance for power grid planning and construction (such as increasing distribution transformer capacity), operation and dispatch (such as peak shaving and valley filling), and marketing services (such as formulating differentiated electricity pricing strategies).

[0004] Existing methods for identifying heating loads mainly suffer from the following problems:

[0005] Manual statistics and investigation: This method relies on maintenance personnel to manually screen data based on the transformer area files and user installation information. This method is inefficient, has a huge workload, and is prone to errors and omissions due to untimely information updates.

[0006] Rule-based judgment based on simple thresholds: This method uses a fixed threshold, such as determining whether winter electricity consumption exceeds a certain absolute value. However, this method is too rigid, cannot adapt to differences in different regions and transformer substations of varying sizes, has poor generalization ability, and suffers from high false positive and false negative rates.

[0007] Traditional time series analysis models: Although some methods attempt to use traditional statistical models for analysis, they have limited ability to handle complex nonlinear relationships under the coupled influence of multiple variables such as load and weather, and it is difficult to accurately capture the unique time series patterns of electric heating loads.

[0008] Therefore, there is an urgent need for a technical solution that can automatically, accurately and robustly identify electric heating loads in order to address the shortcomings of existing technologies. Summary of the Invention

[0009] The present invention provides a method and system for identifying electric heating loads in a transformer substation, so as to achieve high-precision and high-efficiency automatic identification of electric heating loads.

[0010] To achieve the above objectives, the present invention adopts the following specific solution:

[0011] In a first aspect, the present invention provides a method for identifying the electric heating load of a transformer substation, comprising the following steps:

[0012] S1: Obtain multi-dimensional time-series data of the transformer area to be identified, including historical load data, time characteristic data, and meteorological characteristic data;

[0013] S2: Based on a preset deep learning model, the multidimensional time-series data is processed to generate a preliminary classification result of whether the substation to be identified is an electric heating load; the deep learning model includes: a time-series feature extraction module, used to extract time-series features from the multidimensional time-series data; and a key feature weighting module, used to weight the time-series features to strengthen key features related to the heating load;

[0014] S3: Based on preset business rules, perform cross-year comparison and verification of the historical load data;

[0015] S4: Combining the preliminary classification results and the cross-year comparison verification results, it is finally determined whether the area to be identified is an electric heating load.

[0016] To further improve the above technical solution, the acquisition of the multidimensional time series data includes: collecting data at a preset time resolution, and constructing the multidimensional time series data into input samples containing multiple time steps in units of a preset time window, for processing by the deep learning model.

[0017] Furthermore, the deep learning model is pre-trained through the following steps: obtaining a training dataset containing known substation category labels, the category labels indicating whether the substation is an electric heating load; constructing multiple training samples covering a preset time window from the multidimensional time-series data of each substation in the training dataset; and training the deep learning model using the training samples through supervised learning to minimize the cross-entropy loss between the classification prediction and the category label.

[0018] Furthermore, the temporal feature extraction module is a Long Short-Term Memory (LSTM) network model; the key feature weighting module is a Key Feature Weighted Network (KAN) model based on an attention mechanism.

[0019] Furthermore, during the processing, the key feature weighting module adaptively increases the contribution of time-series features corresponding to the electric heating load characteristic pattern by using weights formed after training. The electric heating load characteristic pattern includes a sustained high load value during low-temperature periods or an atypical increase in the load value at night.

[0020] Furthermore, the preset business rules include: a heating season electricity consumption growth rule, used to determine whether the current year's heating season electricity consumption is higher than the previous year's heating season electricity consumption; and a non-heating season electricity consumption stability rule, used to determine whether the deviation between the current year's non-heating season electricity consumption and the previous year's non-heating season electricity consumption is within a preset threshold range.

[0021] Furthermore, the start and end times of the heating season or the preset threshold in the non-heating season electricity consumption stability rules are dynamically adjusted based on at least one of the following factors: the geographical location of the transformer substation to be identified, historical meteorological data, or the type of user in the substation.

[0022] The specific rule for stabilizing electricity consumption during the non-heating season is as follows: determine whether the absolute difference between the total electricity consumption during the current year's non-heating season and the total electricity consumption during the previous year's non-heating season is less than or equal to the allowable deviation value determined by the product of the total electricity consumption during the previous year's non-heating season and a preset proportional coefficient.

[0023] Furthermore, S4 also includes: when the preliminary classification result is inconsistent with the cross-year comparison verification result, the area to be identified is marked as awaiting manual review.

[0024] Secondly, the present invention provides an electric heating load identification system for implementing the above-mentioned method for identifying electric heating loads in a transformer substation, comprising:

[0025] The data acquisition module is used to acquire multidimensional time-series data of the area to be identified; wherein, the multidimensional time-series data includes historical load data, time characteristic data, and meteorological characteristic data;

[0026] A deep learning processing module is used to process the multi-dimensional time-series data based on a preset deep learning model to generate a preliminary classification result of whether the area to be identified is an electric heating load. The deep learning model includes: a time-series feature extraction module for extracting time-series features from the multi-dimensional time-series data; a key feature weighting module for weighting the time-series features to obtain enhanced key features related to the heating load; and a classification module for determining the preliminary classification result based on the key features.

[0027] The business rule verification module is used to perform cross-year comparison verification on the historical load data based on preset business rules, and obtain the comparison verification results.

[0028] The comprehensive judgment module is used to determine whether the area to be identified is an electric heating load based on the preliminary classification results and the comparison verification results.

[0029] Furthermore, the temporal feature extraction module is configured to use a Long Short-Term Memory (LSTM) network model or a Gated Recurrent Unit (GRU) model; the key feature weighting module is configured to use a Key Feature Weighting Network (KAN) model based on an attention mechanism.

[0030] Compared with the prior art, the advantages of the present invention are as follows:

[0031] By accurately capturing the temporal patterns of the load through deep learning models and combining them with business rules for dual verification, the system effectively avoids misjudgment and missed judgment by a single model, significantly improving the accuracy and recall rate of identification.

[0032] The collaborative mechanism between AI models and business rules, especially the dynamically adjusted business rules, enables this method to adapt to changes in load characteristics under different regions and climatic conditions, exhibiting good robustness and generalization ability.

[0033] The entire identification process can be executed automatically, greatly reducing manual intervention, improving work efficiency, and providing strong support for the intelligent management of large-scale power grid distribution areas.

[0034] It innovatively combines the complex pattern recognition capabilities of deep learning with the deterministic logic of power business, and strengthens the specific patterns of heating load through a key feature weighting module, thus solving the complex identification problem that traditional methods cannot handle. Attached Figure Description

[0035] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0036] Figure 1 This is the overall process of a method for identifying the electric heating load in a transformer substation, as provided in an embodiment of the present invention.

[0037] Figure 2 This is a flowchart of the training process for a deep learning model.

[0038] Figure 3 This is a block diagram of a substation electric heating load identification system provided in an embodiment of the present invention.

[0039] Figure 4 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0040] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings, but the scope of protection of the present invention is not limited to the embodiments described.

[0041] Example 1

[0042] Reference Figure 1 This figure illustrates the overall process of a method for identifying the electric heating load in a transformer substation provided in this embodiment. The method includes the following steps:

[0043] Step S1: Obtain multi-dimensional time-series data of the station area to be identified.

[0044] like Figure 1 As shown in the "Data Loading and Processing" section on the left, the data acquisition module collects multi-dimensional time-series data of the distribution area to be identified from the power grid's load monitoring system, meteorological data service interface, and system clock and holiday database. This data undergoes preprocessing such as cleaning, missing value imputation, and standardization before being input into the model.

[0045] Specific data types include:

[0046] Time-series electricity load data, such as the active power of the transformer area, is the core basis for identification;

[0047] Meteorological data, including temperature and humidity, are key external factors affecting heating load;

[0048] Time-related data, including hours, days of the week, and whether it is a holiday, are used to capture the periodicity of the load.

[0049] A preferred implementation is to collect data with a preset time resolution of 15 minutes and a preset time window of 24 hours, and construct the data into input samples containing 96 time steps for subsequent processing by the deep learning model.

[0050] Step S2: Perform preliminary classification based on a deep learning model.

[0051] like Figure 1 As shown in the "Preliminary Classification Based on LSTMKAN" process in the middle section, the deep learning processing module receives the preprocessed input samples and performs preliminary classification.

[0052] It should be noted that the deep learning model is pre-trained, and its core innovation lies in its two-layer structure:

[0053] Temporal feature extraction module (corresponding) Figure 1 (In the “LSTM time series modeling” section): Preferably, the time series feature extraction module employs a Long Short-Term Memory (LSTM) network model. LSTM can effectively learn and remember the dependencies of load data over long time spans, accurately capturing the dynamic patterns of the load curve.

[0054] Key feature weighting module (corresponding) Figure 1(KAN Enhanced Features): After high-dimensional temporal features are extracted by LSTM, these features are fed into the Key Feature Weighted Network (KAN) model. The core function of the KAN model is to adaptively amplify features that are highly correlated with electric heating load patterns through its internal weights formed after training. For example, for patterns such as "continuously high load during low-temperature periods" or "atypical increase in nighttime load," KAN will assign them higher weights, thus playing a dominant role in classification decisions.

[0055] Classified output (corresponding) Figure 1 The "classification layer" in the model: After weighting, the features are finally passed through the classification module (such as the classification output layer of Softmax) to output the probability that the area is an electric heating load, thus generating the "preliminary classification result".

[0056] In one alternative embodiment, refer to Figure 2 The diagram details how the deep learning model described in step S2 above is pre-trained. The training process begins with "data input," which includes a large amount of historical data from substations with known category labels (i.e., "heated substation" or "non-heated substation"). This data flows through feature engineering modules such as "data preprocessing and cleaning" and "constructing the training set," and is constructed into training samples that meet the model input requirements. Subsequently, these training samples are fed into the "LSTM_KAN model" in the "model training" process. Through supervised learning, the model performs forward propagation to obtain prediction results and calculates cross-entropy loss with the true labels. Then, under the action of the "optimizer," backpropagation is performed to update the model weights. This process (i.e., "model iterative training") is repeated until the model's performance on the validation set reaches the preset standard. Finally, the optimal model after training, along with its parameters (such as normalized parameters), is fixed by the "model training and saving" module for use during online identification.

[0057] Step S3: Perform cross-year comparison verification based on business rules.

[0058] like Figure 1 As shown in the "Business Rule Verification" module on the right, in order to compensate for possible misjudgments by the AI ​​model, the Business Rule Verification module introduces deterministic verification logic based on power business knowledge.

[0059] In one optional embodiment, cross-year comparison verification based on business rules includes:

[0060] (1) Rules for the growth of electricity consumption during the heating season, including: comparing the total electricity consumption of the current heating season in the area to be identified with the total electricity consumption of the same period of the previous year. If the current year's electricity consumption is significantly higher than that of the previous year, it meets the characteristics of the heating load being put into operation. (2) Rules for the stability of electricity consumption during the non-heating season, including: comparing the total electricity consumption of the current non-heating season in the area to be identified with that of the same period of the previous year. If the two are within a preset reasonable threshold range, the annual growth of load caused by other factors such as industrial capacity expansion can be ruled out, and it can be further confirmed that the growth in winter is due to the seasonal heating load.

[0061] In another optional embodiment, cross-year comparison verification based on business rules includes:

[0062] 1. Data preparation and preprocessing

[0063] Obtain historical daily electricity consumption data for at least two full years for the area to be identified. Based on the geographical location and climate characteristics of the area, clearly define the start and end dates of the heating season and the non-heating season. For example, define the heating season as November 15th to March 15th of the following year, and the remaining dates as the non-heating season.

[0064] Calculate the total electricity consumption for the current heating season (E_heat_current), the total electricity consumption for the previous heating season (E_heat_previous), the total electricity consumption for the current non-heating season (E_non_heat_current), and the total electricity consumption for the previous non-heating season (E_non_heat_previous) based on historical daily electricity consumption data.

[0065] 2. Calculation of the confidence level P_growth_confidence for electricity consumption growth during the heating season. The main purpose is to assess the extent to which the observed increase in electricity consumption exceeds the reasonable range of natural growth in the power distribution area, thereby confirming that the increase is caused by the new electric heating load.

[0066] Specifically, it includes the following sub-steps:

[0067] a. Calculate the observed growth rate r_observed:

[0068] r_observed=(E_heat_currentE_heat_previous) / E_heat_previous

[0069] The observed growth rate r_observed is used to reflect the magnitude of the increase in electricity consumption.

[0070] b. Establish a prior model of natural growth:

[0071] The natural growth prior model is trained on historical data. Specifically, a large amount of historical annual electricity consumption data for non-heating substations is collected, and the year-on-year electricity consumption growth rate of these substations is calculated, forming a dataset R_natural={r1,r2,r3,...}. Assuming that R_natural follows a normal distribution, the mean μ and standard deviation σ of this distribution are fitted using the maximum likelihood estimation method, i.e., r_natural~N(μ,σ). 2 For example, μ could be 0.02 (representing 2% annual growth) and σ could be 0.01.

[0072] c. Calculate the confidence level of the growth anomaly, including:

[0073] Calculate the right-tail probability of the observed growth rate r_observed in the natural growth distribution. The formula is:

[0074] P_growth=P(r_natural>=r_observed)

[0075] The right-tailed probability value P_growth represents the likelihood that a typical unheated terracing area will naturally grow at or above the observed astonishing growth rate.

[0076] Define the confidence level for abnormal growth as: P_growth_confidence = 1 - P_growth;

[0077] The closer the value of P_growth_confidence is to 1, the more confident that the current growth is abnormal and unnatural, and is most likely due to the commissioning of electric heating equipment.

[0078] 3. Calculation of the confidence level P_stable for electricity consumption stability during the non-heating season.

[0079] The confidence level P_stable for electricity consumption stability during the non-heating season is mainly used to assess whether the base load of a distribution area is stable. If electricity consumption also changes significantly during the non-heating season, it indicates that the annual electricity consumption pattern has changed, which reduces confidence in attributing the winter increase solely to seasonal heating.

[0080] Specifically, it includes the following sub-steps:

[0081] a. Use the Z-test to calculate the average daily electricity consumption for the two non-heating seasons:

[0082] μ_current=E_non_heat_current / N_days,μ_previous=E_non_heat_previous / N_days.

[0083] Assuming the variances of the two series are unknown but equal, calculate the pooled standard deviation, and then calculate the Z-statistic. Calculate the p-value based on the Z-statistic.

[0084] Define the stability confidence level as P_stable = p_value.

[0085] 4. Combine the two independent confidence scores into a single overall business rule confidence score.

[0086] A weighted linear combination method can be used:

[0087] P_rule=α*P_growth_confidence+β*P_stable

[0088] Where α + β = 1. The weights α and β can be adjusted according to actual needs. For example, if abnormal growth is considered stronger evidence, α = 0.7 and β = 0.3 can be set. P_rule is a continuous value between [0, 1].

[0089] Step S4: Combine the results to finally determine the identification conclusion.

[0090] In one alternative embodiment, such as Figure 1 As shown in the "Comprehensive Judgment Result" and "Result Output" flowchart, the comprehensive judgment module performs a comprehensive judgment on the preliminary classification result of step S2 and the verification result of step S3. If the model judgment is "yes" and the business rule verification passes ( Figure 1 If the output shows "Consistent Processing: Output Results", it indicates that the area is an electric heating load. An important robust design principle is that when the two results are inconsistent ( Figure 1 In cases of inconsistency (manual or weight adjustment), the system will not draw hasty conclusions but will mark the area as "awaiting manual review" and submit it to professionals for final confirmation, thus ensuring the high reliability of the system.

[0091] In one alternative embodiment, if P_rule is used for subsequent judgments, the P_rule value is used in a collaborative decision-making process with the preliminary classification probability P_model given by the deep learning model.

[0092] For example: P_final=λ*P_model+(1-λ)*P_rule.

[0093] Here, λ is a trade-off parameter that can be obtained through optimization using historical validation sets.

[0094] After calculating P_final, a judgment is made based on the dual thresholds, for example:

[0095] If P_final >= 0.75, it is finally determined as an electric heating load.

[0096] If P_final <= 0.25, it is finally determined as a non-electric heating load.

[0097] If 0.25 < P_final < 0.75, it is determined as uncertain, and manual review is recommended.

[0098] Embodiment 2

[0099] As Figure 3 shown, based on the same inventive concept as the above embodiment, the present invention also provides a system for implementing the method for identifying the electric heating load in the area of Embodiment 1, including:

[0100] A data acquisition module for acquiring multi-dimensional time-series data of the area to be identified; wherein, the multi-dimensional time-series data includes historical load data, time feature data, and meteorological feature data;

[0101] A deep learning processing module for processing the multi-dimensional time-series data based on a preset deep learning model to generate a preliminary classification result of whether the area to be identified is an electric heating load; wherein, the deep learning model includes: a time-series feature extraction module for extracting time-series features from the multi-dimensional time-series data; a key feature weighting module for weighting the time-series features to obtain enhanced key features related to the heating load; a classification module for determining the preliminary classification result according to the key features;

[0102] A business rule verification module for performing cross-year comparison verification on the historical load data based on preset business rules to obtain a comparison verification result;

[0103] A comprehensive determination module for determining whether the area to be identified is an electric heating load according to the preliminary classification result and the comparison verification result.

[0104] Furthermore, the time-series feature extraction module is configured to use a long short-term memory network model or a gated recurrent unit model; the key feature weighting module is configured to use a key feature weighting network model based on an attention mechanism.

[0105] The functions of the modules in the system correspond to the method steps of Embodiment 1 above. Its internal units, such as the time-series feature extraction module and the key feature weighting module inside the deep learning processing module, use the LSTM and KAN models. This system can be deployed on a server or a cloud platform, and by executing the stored program instructions, it can achieve Figure 1 and Figure 2 all the processes shown.

[0106] Embodiment 3

[0107] like Figure 4 As shown, the present invention also provides an electronic device 100 for implementing a method for identifying the electric heating load in a transformer substation;

[0108] The electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0109] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the method for identifying the electric heating load of a transformer substation in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0110] The memory 101 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0111] At least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 102 may be a microprocessor or any conventional processor. Processor 102 is the control center of electronic device 100, connecting various parts of electronic device 100 via various interfaces and lines.

[0112] The memory 101 in the electronic device 100 stores multiple instructions to implement a method for identifying the electric heating load in a transformer substation, and the processor 102 can execute multiple instructions to achieve the following:

[0113] Acquire multidimensional time-series data of the area to be identified; wherein, the multidimensional time-series data includes historical load data, time characteristic data, and meteorological characteristic data;

[0114] Based on a preset deep learning model, the multidimensional time-series data is processed to generate a preliminary classification result of whether the area to be identified is an electric heating load. The deep learning model includes: a time-series feature extraction module for extracting time-series features from the multidimensional time-series data; a key feature weighting module for weighting the time-series features to obtain enhanced key features related to the heating load; and a classification module for determining the preliminary classification result based on the key features.

[0115] Based on preset business rules, the historical load data is compared and verified across years to obtain the comparison and verification results.

[0116] Based on the preliminary classification results and the comparative verification results, it is determined whether the area to be identified is an electric heating load.

[0117] Example 4

[0118] If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0119] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0120] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0121] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0122] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0123] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying the electric heating load of a transformer substation, characterized in that, Includes the following steps: Acquire multidimensional time-series data of the transformer area to be identified; The multidimensional time-series data includes historical load data, time characteristic data, and meteorological characteristic data; Based on a pre-defined deep learning model, the multi-dimensional time-series data is processed to generate a preliminary classification probability that the substation to be identified is an electric heating load. The deep learning model includes a temporal feature extraction module, a key feature weighting module, and a classification module. The temporal feature extraction module is used to extract temporal features from the multidimensional time-series data. The key feature weighting module is used to weight the temporal features to obtain enhanced key features related to heating load. The classification module is used to output the preliminary classification probability based on the key features. ; Based on preset business rules, the historical load data is compared and verified across years to obtain the confidence level of the business rules. ; The cross-year comparison verification includes: Obtain historical daily electricity consumption data for at least two full years for the area to be identified, and determine the start and end dates of the heating season and non-heating season based on the geographical location and climate characteristics of the area to be identified. Calculate the total electricity consumption for the current heating season based on the historical daily electricity consumption data. Total electricity consumption during the previous heating season ; based on and The relative differences between the observed growth rates are used to calculate the natural growth prior model. This natural growth prior model is established by collecting historical annual electricity consumption data from known non-heating areas, calculating the annual electricity consumption growth rate of these known non-heating areas, and forming a dataset. The result was obtained through post-fitting. Calculate the right tail probability value And define the confidence level of electricity consumption growth during the heating season. ; A Z-test is performed on the average daily electricity consumption during the current year's non-heating season and the previous year's non-heating season, corresponding to two non-heating seasons, and a stability confidence level for non-heating season electricity consumption is defined. ; according to Obtain the confidence level of the business rule α and β are weights. ; Based on the preliminary classification probability confidence level with respect to the business rules Determine the overall confidence level , λ is the trade-off parameter obtained through optimization using the historical validation set; and according to The comparison result with the preset dual threshold determines whether the area to be identified is an electric heating load, or marks the area to be identified as awaiting manual review.

2. The method for identifying the electric heating load of a transformer substation according to claim 1, characterized in that, Obtain multi-dimensional time-series data of the transformer area to be identified, including: Data is collected at a preset time resolution, and the multidimensional time series data is constructed into an input sample containing multiple time steps using a preset time window as the unit.

3. The method for identifying the electric heating load in a transformer substation according to claim 1, characterized in that, The deep learning model is pre-trained through the following steps: Obtain a training dataset containing known transformer substation category labels, which indicate whether the substation is an electric heating load; The multidimensional time-series data of each station area in the training dataset are constructed into multiple training samples covering a preset time window; The deep learning model is trained using the training samples through supervised learning to minimize the cross-entropy loss between the classification prediction and the category label.

4. The method for identifying the electric heating load of a transformer substation according to any one of claims 1 to 3, characterized in that, The temporal feature extraction module is a long short-term memory network model; the key feature weighting module is a key feature weighting network model based on an attention mechanism.

5. The method for identifying the electric heating load in a transformer substation according to claim 4, characterized in that, The key feature weighting module is specifically used for: The contribution of time-series features corresponding to electric heating load characteristic patterns is adaptively increased by weights formed after training; wherein, the electric heating load characteristic patterns include sustained high load values ​​during low-temperature periods or atypical increases in load values ​​at night.

6. The method for identifying the electric heating load in a transformer substation according to claim 1, characterized in that, The start and end dates of the heating season and non-heating season are determined based on the geographical location and climate characteristics of the area to be identified, including: defining November 15th of each year to March 15th of the following year as the heating season, and the remaining dates as the non-heating season.

7. The method for identifying the electric heating load in a transformer substation according to claim 1, characterized in that, Determining whether the area to be identified is an electric heating load based on the comparison result between P_final and the preset dual thresholds, or marking the area to be identified as awaiting manual review, includes: If P_final ≥ 0.75, the area to be identified is ultimately determined to be an electric heating load; if P_final ≤ 0.25, the area to be identified is ultimately determined to be a non-electric heating load; if 0.25 < P_final < 0.75, it is determined to be uncertain, and the area to be identified is marked as awaiting manual review.

8. A system for identifying electric heating loads in implementing the method for identifying electric heating loads in a transformer substation as described in claim 1, characterized in that, include: The data acquisition module is used to acquire multi-dimensional time-series data of the station area to be identified; The multidimensional time-series data includes historical load data, time characteristic data, and meteorological characteristic data; The deep learning processing module is used to process the multidimensional time-series data based on a preset deep learning model to generate a preliminary classification probability that the substation to be identified is an electric heating load. The deep learning model includes a temporal feature extraction module, a key feature weighting module, and a classification module. The temporal feature extraction module is used to extract temporal features from the multidimensional time-series data. The key feature weighting module is used to weight the temporal features to obtain enhanced key features related to heating load. The classification module is used to output the preliminary classification probability based on the key features. ; The business rule verification module is used to perform cross-year comparison verification of the historical load data based on preset business rules to obtain the business rule confidence level. The business rule verification module is specifically used to obtain historical daily electricity consumption data for at least two complete years for the area to be identified, and to calculate the total electricity consumption for the current heating season. Total electricity consumption during the previous heating season ,based on and The relative differences between them and the confidence level of electricity consumption growth during the heating season calculated using the natural growth prior model. The stability confidence level of electricity consumption during the non-heating season was calculated based on the Z-test results of the average daily electricity consumption during the two non-heating seasons. and in accordance with Obtain the confidence level of the business rule ; The comprehensive judgment module is used to determine based on Determine the overall confidence level and according to The comparison result with the preset dual threshold determines whether the area to be identified is an electric heating load, or marks the area to be identified as awaiting manual review.

9. The system according to claim 8, characterized in that, The temporal feature extraction module is configured to use a long short-term memory network model or a gated recurrent unit model; the key feature weighting module is configured to use a key feature weighting network model based on an attention mechanism.

Images