Power data processing method and system based on obtained power index model

Abstract

The invention relates to the technical field of data processing, and particularly discloses an electric power data processing method and system based on an obtained electric power index model, and the method comprises the steps: carrying out the data analysis of to-be-processed electric power related data through employing a data analysis technology based on deep learning, and extracting key feature information in the to-be-processed electric power related data, meanwhile, a power related data set which is marked as normal is introduced, and semantic query search is performed on the power related data to be processed and the power related data which is marked as normal, so that anomaly diagnosis of the power related data to be processed is realized, and then the power related data to be processed is obtained based on an anomaly diagnosis result. And inputting the to-be-processed power related data without abnormality into the obtained power index model for calculation so as to obtain a comprehensive power condition index. In this way, abnormal data can be effectively identified and eliminated, it is ensured that only high-quality power-related data is used for calculation of the power index model, and therefore the efficiency and quality of power supply are evaluated more accurately.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and more specifically, to a power data processing method and system based on an obtained power index model. Background Art

[0002] With the development of social economy and technological progress, electricity, as an indispensable basic energy source in modern society, its service quality is directly related to the production and life of enterprises and residents. Therefore, in the operation and management of the power industry, it is of great significance to accurately and efficiently evaluate the quality of power supply and service. Since the World Bank's electricity access indicators include five secondary indicators: "procedures", "time", "cost", "power supply reliability and electricity bill transparency", and "electricity price", and China has added the "electricity access convenience" indicator on this basis, the original electricity access index algorithm cannot reflect the actual situation.

[0003] In response to this, the invention patent with the publication number CN115796427A proposes a method for constructing an electricity access index model. By obtaining the evaluation index formulas of electricity application procedures, electricity application time, electricity application cost, power supply reliability and electricity bill transparency, and electricity access convenience index, and weighted summing each evaluation index formula with a preset weight to obtain an electricity index formula, the electricity index model is constructed, thus realizing a comprehensive evaluation of the quality of power supply and service.

[0004] However, in practical applications, power-related data may be affected by various factors and become abnormal, such as sensor failures, human errors, or extreme weather events. If these abnormal data directly enter the electricity index model for calculation without being processed, it may lead to distortion of the index results, thereby affecting the judgment of the true level of power services and the formulation of relevant policies.

[0005] Therefore, there is an expectation for a power data processing method and system based on an obtained power index model. Summary of the Invention

[0006] To solve the above technical problems, the present application is proposed. Embodiments of the present application provide a power data processing method and system based on obtaining a power index model. By using data analysis technology based on deep learning to perform data analysis on the power-related data to be processed, key feature information in the power-related data to be processed is extracted. At the same time, a power-related data set labeled as normal is introduced, and semantic query search based on dynamic anchoring of the selection range is performed on the power-related data to be processed and the power-related data labeled as normal, so as to realize the anomaly diagnosis of the power-related data to be processed. Furthermore, based on the anomaly diagnosis result, the power-related data to be processed without anomalies is input into the obtained power index model for calculation to obtain a comprehensive power situation index. In this way, abnormal data can be effectively identified and excluded, ensuring that only high-quality power-related data is used for the calculation of the power index model, thereby more accurately evaluating the efficiency and quality of power supply.

[0007] According to one aspect of the present application, there is provided a power data processing method based on obtaining a power index model, which includes:

[0008] Obtain power-related data to be processed;

[0009] Based on the set of power-related data labeled as normal, perform anomaly diagnosis on the power-related data to be processed to obtain an anomaly diagnosis result;

[0010] In response to the anomaly diagnosis result indicating that there are no anomalies in the power-related data to be processed, input the power-related data to be processed into the obtained power index model to obtain a power situation index.

[0011] In the above power data processing method based on obtaining a power index model, based on the set of power-related data labeled as normal, performing anomaly diagnosis on the power-related data to be processed to obtain an anomaly diagnosis result includes: respectively performing vector quantization encoding on each power-related data labeled as normal in the set of power-related data labeled as normal and the power-related data to be processed to obtain a set of structured encoding vectors of normal power-related data and a structured encoding vector of the power-related data to be processed; performing dynamic semantic search based on the selection range ratio on the structured encoding vector of the power-related data to be processed and the set of structured encoding vectors of normal power-related data to obtain a semantic query response encoding vector of power-related data; determining the anomaly diagnosis result based on the semantic query response encoding vector of power-related data.

[0012] According to another aspect of the present application, there is provided a power data processing system based on obtaining a power index model, which includes:

[0013] A power-related data acquisition module for obtaining power-related data to be processed;

[0014] A data anomaly diagnosis module for performing anomaly diagnosis on the power-related data to be processed based on a set of power-related data marked as normal to obtain an anomaly diagnosis result;

[0015] A power situation index acquisition module for, in response to the anomaly diagnosis result indicating that there is no anomaly in the power-related data to be processed, inputting the power-related data to be processed into the power index model to obtain a power situation index.

[0016] Compared with the prior art, the power data processing method and system based on the power index model provided by the present application perform data analysis on the power-related data to be processed by using data analysis technology based on deep learning to extract key feature information in the power-related data to be processed. At the same time, a set of power-related data marked as normal is introduced, and semantic query search based on dynamic anchoring of the selection range is performed on the power-related data to be processed and the power-related data marked as normal to realize the anomaly diagnosis of the power-related data to be processed. Furthermore, based on the anomaly diagnosis result, the power-related data to be processed without anomalies is input into the power index model for calculation to obtain a comprehensive power situation index. In this way, abnormal data can be effectively identified and excluded, ensuring that only high-quality power-related data is used for the calculation of the power index model, thereby more accurately evaluating the efficiency and quality of power supply. Description of the Drawings

[0017] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 It is a flowchart of a power data processing method based on a power index model according to an embodiment of the present application.

[0019] Figure 2 It is a flowchart of sub-step S2 of a power data processing method based on a power index model according to an embodiment of the present application.

[0020] Figure 3 It is a schematic diagram of data flow of sub-step S2 of a power data processing method based on a power index model according to an embodiment of the present application.

[0021] Figure 4It is a flowchart of sub-step S22 of the power data processing method based on obtaining a power index model according to an embodiment of the present application.

[0022] Figure 5 It is a block diagram of a power data processing system based on obtaining a power index model according to an embodiment of the present application. Detailed implementation manners

[0023] As shown in the present application and the claims, unless the context clearly indicates otherwise, words such as "a", "an", "one", and / or "the" are not specifically singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0024] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server. The modules are merely illustrative, and different aspects of the system and method can use different modules.

[0025] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the operations before or below are not necessarily executed precisely in order. On the contrary, as needed, various steps can be executed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or several steps can be removed from these processes.

[0026] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0027] It is worth noting that in the present application, all actions of obtaining data are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where the data is located and obtaining the authorization given by the owner of the corresponding device.

[0028] As mentioned in the above background art, Patent CN115796427A proposes a construction method for obtaining a power index model. It obtains the evaluation index formulas of the power connection link, power connection time, power connection cost, power supply reliability, electricity fee transparency, and power supply convenience index, and constructs a power index model by weighted summing each evaluation index formula with a preset weight, thereby realizing a comprehensive evaluation of the power supply and service quality.

[0029] However, in practical applications, power-related data may be affected by various factors and become abnormal. For example, sensor failures, human errors, or extreme weather events. If such abnormal data directly enters the power index model for calculation without being processed, it may lead to distorted index results, thereby affecting the judgment of the true level of power services and the formulation of relevant policies. To address the above technical problems, this application proposes a power data processing method based on obtaining a power index model. It uses deep learning-based data analysis techniques to analyze the power-related data to be processed, extract key feature information from the power-related data to be processed, and at the same time introduce a dataset of power-related data labeled as normal. By performing semantic query search based on dynamically anchored selection ranges on the power-related data to be processed and the power-related data labeled as normal, the abnormal diagnosis of the power-related data to be processed is realized. Then, based on the abnormal diagnosis result, the power-related data to be processed without abnormalities is input into the obtained power index model for calculation to obtain a comprehensive power situation index. In this way, abnormal data can be effectively identified and excluded, ensuring that only high-quality power-related data is used for the calculation of the power index model, thus more accurately evaluating the efficiency and quality of power supply.

[0030] Figure 1 FIG. is a flowchart of a power data processing method based on obtaining a power index model according to an embodiment of the present application. As Figure 1 shown, the power data processing method based on obtaining a power index model includes the steps of: S1, obtaining power-related data to be processed; S2, performing abnormal diagnosis on the power-related data to be processed based on a set of power-related data labeled as normal to obtain an abnormal diagnosis result; S3, in response to the abnormal diagnosis result indicating that there are no abnormalities in the power-related data to be processed, inputting the power-related data to be processed into the obtained power index model to obtain a power situation index.

[0031] In the above-mentioned power data processing method based on the power acquisition index model, in step S1, power-related data to be processed is acquired. It should be understood that the power-related data to be processed is the basic data for power service evaluation. Specifically, in the evaluation of the electricity application process, the actual number of steps in the electricity application process (N) needs to be acquired, that is, the actual number of interactions between the enterprise and external parties during the electricity application process, as well as the maximum and minimum values of the evaluation criteria for the electricity application process. In the evaluation of the electricity application time, the actual operation time (t) of the power system needs to be acquired, that is, the calendar days from submitting the application to finally connecting to the power, as well as the maximum and minimum values of the evaluation criteria for the electricity application time. In the evaluation of the electricity application cost, the percentage of the electricity connection fee cost in the per capita national income of the economy needs to be acquired, which includes all relevant costs such as approval fees, application connection fees, inspection fees, etc., but does not include value-added tax. In the evaluation of power supply reliability and electricity fee transparency, the required power-related data includes the power cut frequency index (FCL), power cut duration index (Ucs), number of power cuts (Jt), number of power supply restorations (Ht), monitoring frequency (Gt), number of implemented measures (FCk), and electricity fee transparency (Md); among them, the power cut frequency index is jointly determined by the preset cycle time (L), the actual number of power cut users (k), the number of power cuts (NK) within the L time, and the number of affected users (Nns), and the power cut duration index is jointly determined by the power cut duration (Lik) of k actual power cut users at the i-th power cut and the total number of users (Nc). In the evaluation of the power supply convenience index, the required power-related data includes the user satisfaction level (Mn) and its corresponding weight (Wn), for example, the indicators are formulated according to four levels of "very satisfied", "satisfied", "average", and "dissatisfied", and the corresponding weights are 0.4, 0.3, 0.2, and 0.1 respectively. Based on this, in the technical solution of this application, by comprehensively and accurately collecting various power-related data, a comprehensive and accurate evaluation of the power service quality can be achieved.

[0032] First, in terms of the evaluation of the electricity application process, the process of power users from applying for electricity to finally connecting to the grid involves multiple steps, including submitting applications, on-site inspections, design reviews, construction and installation, etc. To evaluate the service quality of these links, power companies usually establish a set of standardized operation processes and record the key time nodes and service details of each stage. For example, for the access request of new users, the system will automatically generate a unique application number and track the entire process through an electronic workflow. During this process, the staff needs to complete the corresponding tasks within the specified time frame and update the status information to the database in a timely manner. In addition, a customer service platform should be established to allow users to query the processing progress online, and at the same time provide a suggestion box or hotline to collect feedback on the service experience. All these operation logs and service evaluations will be saved as raw data, providing rich materials for subsequent analysis.

[0033] Next is the evaluation of power supply time, which refers to the length of time required for users to formally connect to the power supply from the time they raise their electricity demand. Shortening this cycle means improving service efficiency and enhancing user satisfaction. In order to accurately capture the changing trend of power supply time, in addition to relying on the workflow management system mentioned above, a timestamp mechanism can also be introduced, that is, whenever an important event occurs (such as receiving an application, completing an investigation), a time stamp is automatically added. In this way, even in the case of multi-department collaboration, the actual time spent on each link can be accurately calculated. In addition, considering that different policy environments and technical conditions may exist in different regions, it is necessary to summarize the data from various regions in a layered manner, find out the common and characteristic factors, and optimize the process design in a targeted manner. At the same time, big data analysis tools are used to explore the potential laws in historical data, predict possible problems in the future, and formulate response strategies in advance.

[0034] The evaluation of electricity cost focuses on the various expenses borne by users in the entire electricity process, including but not limited to application fees, material fees, engineering fees, etc. The acquisition of this part of data requires a high degree of cooperation from the financial system to ensure that every expenditure is traceable. On the one hand, a strict budget management and approval system should be established to prevent unnecessary expenses; on the other hand, cost-benefit analysis should be emphasized to encourage the use of new technologies and new materials to reduce construction costs. For some complex projects, it may be necessary to introduce a third-party auditing agency for supervision to ensure the transparency and fairness of the use of funds. It is worth noting that with the intensification of market competition and technological progress, the cost of electricity is not fixed, but shows dynamic changes. Therefore, regular cost surveys and adjustment of charging standards have become necessary management measures. At the same time, an open and transparent price system helps to enhance users' sense of trust and reduce disputes.

[0035] The reliability of power supply is closely linked to the transparency of electricity charges, as both are important factors that affect user perception. Power supply reliability refers to the ability of power companies to provide continuous and stable power supply over a certain period of time, while transparency of electricity charges refers to whether the electricity price structure is clear and easy to understand, and whether the basis for charging is open and transparent. For the former, intelligent monitoring equipment can be installed to monitor the operation status of the power grid in real time, and key parameters such as the number of power outages, power outage time, and power restoration time can be recorded. For major failure events, a detailed accident investigation is required to find out the cause and take preventive measures. As for the latter, in addition to publishing the latest electricity price policy on the official website, mobile applications can also be developed to allow users to understand their own electricity consumption and bill details anytime and anywhere. In addition, the implementation of a tiered electricity price system can also help improve the transparency of electricity charges and encourage users to plan their electricity consumption behavior rationally. However, the effective data acquisition for these two evaluation indicators is inseparable from the support of high-level information technology and a sound management system. Only in this way can the authenticity and accuracy of the data be ensured.

[0036] Finally, the evaluation of power supply convenience index is a quantitative expression of the overall convenience of users in obtaining power services. It covers all touchpoints from pre - consultation to final acceptance, reflecting whether users feel convenient during the entire interaction process. To accurately measure this indicator, a multi - dimensional evaluation system can be constructed, covering multiple aspects such as physical distance, simplicity of procedures, response speed, etc. For example, promote online business halls and self - service terminals, simplify electricity application procedures, and shorten waiting times; optimize the customer service hotline to ensure 24 / 7 manned service and quickly answer questions; strengthen community publicity efforts, popularize electricity safety knowledge, and enhance public awareness. At the same time, establish a user satisfaction survey mechanism, regularly invite users to participate in questionnaire filling or telephone interviews, widely listen to opinions and suggestions, and continuously improve service quality. Through statistical analysis of a large amount of sample data, common problems and outstanding advantages can be discovered, providing strong support for the next strategic decision - making.

[0037] In the above - mentioned power data processing method based on the power access index model, in step S2, based on the set of power - related data marked as normal, the power - related data to be processed is subjected to anomaly diagnosis to obtain an anomaly diagnosis result. In particular, considering that in practical applications, power - related data may be affected by various factors such as sensor failures, human errors, or extreme weather events, resulting in anomalies. If directly input into the power index model for calculation, it may lead to distorted index results, thereby affecting the judgment of the true level of power services and the formulation of relevant policies. Therefore, to ensure the data input quality of the power index model and avoid the interference of abnormal data on the accuracy of the model output, this application further extracts the set of power - related data marked as normal from the database, and through the use of deep - learning algorithms, the data to be processed is compared and analyzed with the set of data marked as normal to achieve anomaly diagnosis of the power - related data to be processed. Among them, Figure 2 FIG. is a flowchart of sub - step S2 of the power data processing method based on the power access index model according to an embodiment of the present application. Figure 3 FIG. is a schematic diagram of data flow of sub - step S2 of the power data processing method based on the power access index model according to an embodiment of the present application. As Figure 2 and Figure 3As shown, step S2 includes steps: S21, respectively performing vector quantization encoding on each of the power-related data marked as normal in the set of the power-related data marked as normal and the to-be-processed power-related data to obtain a set of structured encoding vectors of the normal power-related data and a structured encoding vector of the to-be-processed power-related data; S22, performing dynamic semantic search based on the selection range ratio on the structured encoding vector of the to-be-processed power-related data and the set of structured encoding vectors of the normal power-related data to obtain a semantic query response encoding vector of the power-related data; S23, determining the abnormal diagnosis result based on the semantic query response encoding vector of the power-related data.

[0038] Specifically, in step S21, respectively performing vector quantization encoding on each of the power-related data marked as normal in the set of the power-related data marked as normal and the to-be-processed power-related data to obtain a set of structured encoding vectors of the normal power-related data and a structured encoding vector of the to-be-processed power-related data. It should be understood that in power data processing, since the original power-related data contains power information in multiple dimensions, with complexity and diversity, it may be relatively difficult to directly use it for comparative analysis in a deep learning model. Therefore, in order to convert the power-related data into a structured vector representation and eliminate the differences in the data dimensions and value ranges of different dimensions, the present application further performs vector quantization encoding on each of the power-related data marked as normal and the to-be-processed power-related data. Specifically, for numerical data in the power-related data, such as the actual number of electricity application links and the actual operation time of the power system, etc., it can be converted into dimensionless values through standardization processing, while for non-numerical data in the power-related data, such as the user satisfaction level, etc., it can be vectorized through methods such as one-hot encoding or word embedding. Finally, the vectorized representation results of the multi-dimensional data are concatenated to obtain a complete structured encoding vector of the power-related data. In this way, the feature information in the set of the power-related data marked as normal and the to-be-processed power-related data can be represented in a unified mathematical form, generating a set of structured encoding vectors of the normal power-related data and a structured encoding vector of the to-be-processed power-related data, thereby reducing the complexity of data processing and providing a more reliable input for subsequent anomaly detection algorithms.

[0039] Specifically, in step S22, a dynamic semantic search based on the selection range ratio is performed on the set of structured encoding vectors of the power-related data to be processed and the structured encoding vectors of the normal power-related data to obtain a semantic query response encoding vector for the power-related data. Specifically, the present application takes into account that the set of power-related data labeled as normal contains a large number of data features in normal power service scenarios. Directly comparing its features with the power-related data to be processed may lead to low computational efficiency due to the large scale of the data. Therefore, in order to improve the efficiency of anomaly detection, the present application proposes a dynamic semantic search method based on the selection range ratio. By performing a preliminary comparative analysis on the set of structured encoding vectors of the normal power-related data and the structured encoding vectors of the power-related data to be processed, the semantic query range and anchor points of the set of structured encoding vectors of the normal power-related data are dynamically adjusted, thereby significantly reducing the data range for comparative analysis and improving the efficiency of anomaly detection. Among them, Figure 4 is a flowchart of sub-step S22 of the power data processing method based on obtaining a power index model according to an embodiment of the present application. As Figure 4 shown, step S22 includes steps: S221, determining a dynamic search anchor vector for normal power-related data based on the internal association between the structured encoding vector of the power-related data to be processed and each structured encoding vector of the set of structured encoding vectors of the normal power-related data; S222, determining a selection range ratio for normal power-related data based on the characteristic distribution of the dynamic search anchor vector for normal power-related data; S223, performing semantic query response encoding on the set of structured encoding vectors of the power-related data to be processed and the set of structured encoding vectors of the normal power-related data based on the selection range ratio for normal power-related data to obtain the semantic query response encoding vector for the power-related data.

[0040] More specifically, step S221 includes: First, calculating the internal relationship score value of the structured encoding vector of the power-related data to be processed relative to each structured encoding vector of the set of structured encoding vectors of the normal power-related data to obtain a set of internal relationship score values of the power-related data to be processed - normal power-related data, which is expressed by the formula:

[0041] D = {d1, d2,..., d i ,..., d n}

[0042] s i = v T tanh(W r d i + W R vq )

[0043] Among them, D is the set of structured encoding vectors of the normal power-related data, and d1, d2, d i , d n are the 1st, 2nd, i-th, and n-th structured encoding vectors of the normal power-related data in the set of structured encoding vectors of the normal power-related data respectively, v q is the structured encoding vector of the power-related data to be processed, W r and W R are the weight parameter matrix of the normal power-related data and the weight parameter matrix of the power-related data to be processed respectively, tanh(·) is the hyperbolic tangent activation function, v T is the transposed vector of the reference vector for the internal relationship score of the power-related data to be processed - normal power-related data, s i is the score value of the internal relationship of the power-related data to be processed - normal power-related data corresponding to the d i ;

[0044] Then, use the structured encoding vector of the normal power-related data corresponding to the maximum value in the set of the internal relationship score values of the power-related data to be processed - normal power-related data as the dynamic search anchor vector of the normal power-related data, which is expressed by the formula as:

[0045]

[0046] where represents extracting the structured encoding vector of the normal power-related data corresponding to the maximum value in the set of the internal relationship score values of the power-related data to be processed - normal power-related data, and d * is the dynamic search anchor vector of the normal power-related data.

[0047] That is, by using the neural network model to calculate the internal relationship score values between the structured encoding vector of the power-related data to be processed and each structured encoding vector of the normal power-related data in the set of structured encoding vectors of the normal power-related data, the data feature correlation between the power-related data to be processed and the normal power-related data is measured. It should be understood that the higher the internal relationship score value, the more similar the data features between the two. Therefore, the present application further selects the structured encoding vector of the normal power-related data with the highest score value from the set of structured encoding vectors of the normal power-related data as the benchmark for comparative analysis, that is, the dynamic search anchor vector of the normal power-related data. In this way, by screening out the normal power-related data sample that is most similar to the power-related data to be processed as the anchor point for subsequent feature comparison analysis, it helps to improve the accuracy and efficiency of anomaly detection.

[0048] More specifically, in step S222, based on the characteristic distribution of the dynamic search anchor vector for the normal power-related data, the selection range ratio of the normal power-related data is determined, which is expressed by the formula:

[0049]

[0050] where max(d * ), μ, and σ 2 respectively represent the maximum eigenvalue, the mean value of the eigenvalues, and the eigenvalue variance of the dynamic search anchor vector for the normal power-related data. λ is an extremely small positive number used to prevent the denominator from being zero. represents the ceiling operation, and m is the window size of the selection range ratio of the normal power-related data.

[0051] That is, this application considers that relying solely on a single similar data sample for feature comparison may affect the accuracy of anomaly detection due to the limitations of data features. Therefore, to enhance the accuracy of anomaly detection, this application introduces the concept of a dynamic search anchor vector. By calculating the selection range ratio of the normal power-related data, multiple normal power-related data structured coding vectors are selected from the set of structured coding vectors of the normal power-related data as candidate samples for comparative analysis. In a specific example of this application, the vector at the starting position of the selection range ratio of the normal power-related data is the dynamic search anchor vector for the normal power-related data, and each normal power-related data structured coding vector in the selection range ratio of the normal power-related data is defined as a dynamically searched and optimized normal power-related data structured coding vector, thereby constructing a set of dynamically searched and optimized normal power-related data structured coding vectors. In this way, by restricting the data range for comparative analysis, not only can the computational amount be reduced and the efficiency of anomaly detection be improved, but also the focus can be on the normal power-related data samples that are most similar to the power-related data to be processed, avoiding the analysis error caused by a single data sample, thereby improving the accuracy of anomaly detection.

[0052] More specifically, step S223 includes: inputting the structured coding vector of the power-related data to be processed and each dynamically searched and optimized normal power-related data structured coding vector in the selection range ratio of the normal power-related data into a dynamic semantic search encoder to obtain the semantic query response coding vector of the power-related data.

[0053] In a specific example of the present application, inputting each dynamically searched and optimized normal power-related data structured encoding vector in the structured encoding vector of the power-related data to be processed and the ratio of the normal power-related data selection range into a dynamic semantic search encoder to obtain the power-related data semantic query response encoding vector further includes: calculating the position-wise difference between each dynamically searched and optimized normal power-related data structured encoding vector in the ratio of the normal power-related data selection range and the structured encoding vector of the power-related data to be processed to obtain a set of normal power - to-be-processed power-related data difference feature vectors; calculating the cosine similarity between the structured encoding vector of the power-related data to be processed and each dynamically searched and optimized normal power-related data structured encoding vector in the ratio of the normal power-related data selection range to obtain a set of to-be-processed power - normal power-related data cosine similarities; using the set of to-be-processed power - normal power-related data cosine similarities as a set of weights, performing weighted aggregation on the set of normal power - to-be-processed power-related data difference feature vectors, and then fusing the weighted aggregation result with the normal power-related data dynamic search anchor vector to obtain the power-related data semantic query response encoding vector, which is expressed by the formula as:

[0054] W = {d j ,..., d j+m-1}

[0055]

[0056] where d j is the vector at the starting position of the ratio of the normal power-related data selection range, that is, the d * , d j+m-1 is the vector at the ending position of the ratio of the normal power-related data selection range, W is the set composed of each dynamically searched and optimized normal power-related data structured encoding vector in the ratio of the normal power-related data selection range, ||·|| is the modulus of the vector, r j (d j , v q ) is the cosine similarity between d j and v q , and v a is the power-related data semantic query response encoding vector.

[0057] That is, in the present application, by performing feature difference calculations on each of the structured encoding vectors of the power-related data to be processed and the dynamic search optimization normal power-related data structured encoding vectors in the selection range ratio, the feature differences between the power-related data to be processed and the normal power-related data are extracted, and the feature differences are weighted and aggregated using the feature similarity between the two to comprehensively evaluate the overall difference information between the power-related data to be processed and the normal power-related data samples. Further, the overall difference information between the data to be processed and the normal data samples is fused with the dynamic search anchor vector of the normal power-related data to generate the final semantic query response encoding vector for power-related data. In this way, both the overall differences between the power-related data to be processed and multiple normal power-related data samples are considered, and the most similar normal power-related data sample is used as the basis, so that the final semantic query response encoding vector for power-related data can more accurately reflect the abnormal conditions of the power-related data to be processed, providing richer information support for subsequent anomaly diagnosis.

[0058] Specifically, in a specific example of the present application, step S23 includes: inputting the semantic query response encoding vector for power-related data into a data recognition module based on a classifier to obtain the anomaly diagnosis result, where the anomaly diagnosis result is used to indicate whether there is an anomaly in the power-related data to be processed. Specifically, in the training stage, a large number of labeled normal and abnormal power-related data samples are used to train the classifier so that it can learn the feature differences and classification boundaries between normal and abnormal data. In the actual application stage, after the power-related data to be processed is encoded into the semantic query response encoding vector for power-related data through the above processing, it is input into the trained classifier, and the classifier can perform feature pattern recognition and classification judgment on the semantic query response encoding vector for power-related data according to the feature differences and classification boundaries it has learned to determine whether it belongs to an anomaly, thereby realizing the automatic anomaly diagnosis of power-related data.

[0059] More specifically, inputting the semantic query response encoding vector for power-related data into a data recognition module based on a classifier to obtain the anomaly diagnosis result includes: using the fully connected layer of the data recognition module to perform a fully connected encoding on the semantic query response encoding vector for power-related data to obtain a fully connected encoding vector for the semantic query response for power-related data; inputting the fully connected encoding vector for the semantic query response for power-related data into the Softmax classification function of the data recognition module to obtain the probability values of the semantic query response encoding vector for power-related data belonging to each classification label, where the classification labels include that there is an anomaly in the power-related data and that there is no anomaly in the power-related data; and determining the classification label corresponding to the largest of the probability values as the anomaly diagnosis result.

[0060] In particular, when the set of the structured encoding vectors of the power-related data to be processed and the set of the structured encoding vectors of the normal power-related data respectively represent the set of the low-dimensional structured encoding features of the power-related data to be processed and the set of the low-dimensional structured encoding features of the power-related data labeled as normal, when performing feature dynamic semantic search encoding based on the selection range ratio anchoring, the complexity of the selection range ratio anchoring mechanism caused by the semantic differences of the power-related data sources will lead to insufficient long-distance dynamic semantic search encoding representation of the semantic query response encoding vector of the power-related data, thereby reducing the expression effect of the semantic query response encoding vector of the power-related data and affecting the accuracy of the abnormal diagnosis result obtained by its input into the data recognition module based on the classifier.

[0061] Therefore, in a preferred example of the present application, when the semantic query response encoding vector of the power-related data is input into the data recognition module based on the classifier, the semantic query response encoding vector of the power-related data is optimized, and the optimization includes the following steps:

[0062] First, arrange each eigenvalue of the semantic query response encoding vector of the power-related data in ascending order to obtain a semantic query response order encoding vector of the power-related data;

[0063] Second, in response to the absolute value of the difference between the i-th eigenvalue and the (i + 1)-th eigenvalue of the semantic query response order encoding vector of the power-related data being less than or equal to the distance difference hyperparameter ε, calculate the weighted sum of the i-th eigenvalue and the (i + 1)-th eigenvalue as the optimized (i + 1)-th eigenvalue, which is expressed by the formula:

[0064] v′ i+1 =ω1×v i +ω2×v i+1

[0065] where v i and v i+1 respectively represent the i-th eigenvalue and the (i + 1)-th eigenvalue of the semantic query response order encoding vector of the power-related data, ω1 and ω2 respectively represent the first weight parameter and the second weight parameter, and v' i+1 represents the (i + 1)-th eigenvalue of the optimized semantic query response order encoding vector of the power-related data;

[0066] Then, in response to the absolute value of the difference between the i-th eigenvalue and the (i + 1)-th eigenvalue of the semantic query response order encoding vector of the power-related data being greater than the distance difference hyperparameter ε, calculate the square root of the sum of the squares of all eigenvalues of the semantic query response encoding vector of the power-related data, which is expressed by the formula:

[0067]

[0068] Among them, root represents the square root of the sum of the squares of all eigenvalues of the power-related data semantic query response coding vector, and v1, v2 represent the respective eigenvalues of the power-related data semantic query response sequence coding vector;

[0069] Then, multiply the square root root by 2 and then divide by the square of the length of the power-related data semantic query response coding vector to obtain the power-related data semantic query response space primitive value, which is expressed by the formula:

[0070] base = 2×root / L 2

[0071] Among them, L represents the length of the power-related data semantic query response coding vector, and base represents the power-related data semantic query response space primitive value;

[0072] Then, after multiplying the power-related data semantic query response space primitive value by the i-th eigenvalue, calculate the weighted subtraction between the product and the (i + 1)-th eigenvalue to obtain the optimized (i + 1)-th eigenvalue, which is expressed by the formula:

[0073] v i+1 ' = ω3×v i ×base - ω4×v i+1

[0074] Among them, ω3 and ω4 respectively represent the third weight parameter and the fourth weight parameter, and v i+1 ' represents the (i + 1)-th eigenvalue of the optimized power-related data semantic query response sequence coding vector;

[0075] Finally, based on v'1 = v1, combine the optimized (i + 1)-th eigenvalue v i+1 ' to obtain the optimized power-related data semantic query response coding vector.

[0076] In this way, for the problem of insufficient global semantic search representation ability caused by long distances exceeding the predetermined local distribution interval threshold in the feature set of the power-related data semantic query response encoding vector under the predetermined eigenvalue order distribution, the high-dimensional feature space primitive representation based on self-inner product fusion of the power-related data semantic query response encoding vector is used to capture the complex structure of the eigenvalue global network interaction, so as to reconstruct the dynamic semantic search relationship between eigenvalues of the power-related data semantic query response encoding vector by simulating the scale-based high-dimensional feature space potential primitive, so as to realize the encoding reconstruction of the true sequence distribution behavior of the power-related data semantic query response encoding vector under long distances, improve the encoding expression effect of the power-related data semantic query response encoding vector, and improve the accuracy of the abnormal diagnosis result obtained by its input data recognition module based on the classifier.

[0077] In the above power data processing method based on obtaining the power index model, in step S3, in response to the abnormal diagnosis result indicating that there is no abnormality in the power-related data to be processed, the power-related data to be processed is input into the obtained power index model to obtain the power situation index. That is, if the power-related data to be processed is determined to be normal after abnormal diagnosis, it means that it belongs to normal power service data and can be directly input into the obtained power index model for the calculation of the power situation index. It is worth mentioning that the obtained power index model involved in this application can be constructed by the method disclosed in Chinese Patent CN115796427A. Of course, it can also be constructed by other equivalent methods according to actual needs, and this is not limited to this application.

[0078] The power index model is a comprehensive mathematical framework that obtains a comprehensive score based on a series of evaluation index formulas and through weighted summation. Each evaluation index corresponds to different aspects, such as the time efficiency, cost control, power supply reliability, electricity bill transparency and convenience of the electricity application process. Therefore, before inputting data into the model, it must be clearly known which data items should be associated with which specific evaluation index. For example, if the electricity application time is to be evaluated, then the timestamp records associated with the application acceptance date, on-site inspection completion date, etc. should be selected; if it is a consideration of power supply reliability, then the statistical information such as the number of power outages and the duration of power outages needs to be concerned. The establishment of this mapping relationship helps to ensure that the data can be correctly interpreted and each evaluation index can be fully supported.

[0079] To achieve a comprehensive assessment of power supply and service quality, when obtaining a power index model, corresponding weights are usually assigned to each evaluation index. These weights reflect the degree of importance of different factors in the entire service system. They may be fixed values set based on historical experience, or the result of dynamic adjustment through expert consultation or machine learning algorithms. When the data to be processed is ready, the next step is to perform weighted calculations on each evaluation index according to these preset weights. For example, if power supply reliability is considered one of the most critical factors, the corresponding weight may be relatively high; while other less directly related indicators will be given lower weights. Such a design ensures that the finally generated power access situation indicator can more objectively reflect the actual situation, rather than simply relying on the performance of certain single dimensions.

[0080] During the process of performing weighted summation, it is also necessary to consider how to handle the problems of missing values or outliers. Even after strict anomaly diagnosis, there may still be a small number of data points that do not meet the ideal accuracy standard. For such situations, interpolation methods can be used to fill in the missing positions, or reasonable thresholds can be set to eliminate data points that deviate extremely from the normal range. In addition, considering that there may be significant differences between different regions, sometimes it is also necessary to introduce regional correction coefficients so that the model can better fit local characteristics. In short, the entire calculation process should be as flexible as possible to adapt to various complex and changing actual scenarios.

[0081] After completing the weighted summation and obtaining the power situation indicator, this value becomes a comprehensive reference for measuring the power service level within a specific time period. It can be used to compare the gaps between different regions, help identify areas with excellent performance and areas that need improvement, and can also serve as an important basis for policymakers to adjust electricity price strategies and optimize the power grid layout.

[0082] In summary, the power data processing method based on the power access index model according to the embodiments of the present application is elucidated. It uses data analysis techniques based on deep learning to perform data analysis on the power-related data to be processed, so as to extract key feature information in the power-related data to be processed. At the same time, a power-related data set marked as normal is introduced, and semantic query search based on dynamic anchoring of the selection range is performed on the power-related data to be processed and the power-related data marked as normal, so as to realize the anomaly diagnosis of the power-related data to be processed. Then, based on the anomaly diagnosis result, the power-related data to be processed without anomalies is input into the power access index model for calculation to obtain a comprehensive power situation indicator. In this way, abnormal data can be effectively identified and excluded, ensuring that only high-quality power-related data is used for the calculation of the power index model, thereby more accurately evaluating the efficiency and quality of power supply.

[0083] Furthermore, a power data processing system based on an obtained power index model is also provided.

[0084] Figure 5 FIG. is a block diagram of a power data processing system based on an obtained power index model according to an embodiment of the present application. As Figure 5 shown, the power data processing system 100 based on an obtained power index model according to an embodiment of the present application includes: a power-related data acquisition module 110, configured to acquire power-related data to be processed; a data anomaly diagnosis module 120, configured to perform anomaly diagnosis on the power-related data to be processed based on a set of power-related data marked as normal to obtain an anomaly diagnosis result; and a power condition index acquisition module 130, configured to, in response to the anomaly diagnosis result indicating that the power-related data to be processed is normal, input the power-related data to be processed into the obtained power index model to obtain an obtained power condition index.

[0085] Here, those skilled in the art can understand that the specific operations of each module in the above power data processing system based on an obtained power index model have been described in detail in the above description of the power data processing method based on an obtained power index model, and therefore, the repeated description thereof will be omitted. Figures 1 to 4 The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, benefits, effects, etc. mentioned in the present invention are only examples and not limitations, and it cannot be considered that these advantages, benefits, effects, etc. are essential for each embodiment of the present invention. In addition, the specific details of the above embodiments are only for the purpose of illustration and facilitation of understanding, rather than limitations, and the above details do not limit the present invention to necessarily adopt the above specific details for implementation.

[0086] In the above embodiments, the descriptions of each embodiment have their own emphases. For parts not detailed or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only illustrative. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0087]

[0088] ​For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

[0089] In addition, it is obvious that the term "comprising" does not exclude other elements or steps, and the singular does not exclude the plural. The multiple elements stated in the system claims can also be implemented by one element through software or hardware.

[0090] Finally, it should be noted that the above description has been given for purposes of illustration and description. In addition, the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for processing power data based on obtaining a power index model, characterized in that: include: Acquire power-related data to be processed; Based on a set of power-related data marked as normal, an abnormal diagnosis is performed on the power-related data to be processed to obtain an abnormal diagnosis result, including: vector encoding each power-related data marked as normal in the set of power-related data marked as normal and the power-related data to be processed to obtain a set of normal power-related data structured coding vectors and a set of power-related data structured coding vectors to be processed; dynamic semantic search based on a selection range ratio is performed on the set of power-related data structured coding vectors to be processed and the normal power-related data structured coding vectors to obtain a power-related data semantic query response coding vector; based on the power-related data semantic query response coding vector, the abnormal diagnosis result is determined; In response to the abnormality diagnosis result that there is no abnormality in the power-related data to be processed, the power-related data to be processed is input into the obtained power index model to obtain an obtained power situation indicator.

2. The power data processing method based on obtaining the power index model according to claim 1 is characterized in that: Performing a dynamic semantic search based on a selection range ratio on the set of the structured coding vector of the power-related data to be processed and the structured coding vector of the normal power-related data to obtain a semantic query response coding vector of the power-related data, including: Determine a normal power-related data dynamic search anchor vector based on the internal association between the power-related data structured encoding vector to be processed and each normal power-related data structured encoding vector in the set of normal power-related data structured encoding vectors; Determining a normal power-related data selection range ratio based on the characteristic distribution characteristics of the normal power-related data dynamic search anchor vector; Based on the normal power-related data selection range ratio, semantic query response encoding is performed on the set of the to-be-processed power-related data structured encoding vector and the normal power-related data structured encoding vector to obtain the power-related data semantic query response encoding vector.

3. The power data processing method based on obtaining the power index model according to claim 2 is characterized in that: Determining a normal power-related data dynamic search anchor vector based on an internal association between the power-related data structured encoding vector to be processed and each normal power-related data structured encoding vector in the set of normal power-related data structured encoding vectors includes: Calculating the internal relationship score value of the structured coding vector of the power-related data to be processed relative to each normal power-related data structured coding vector in the set of normal power-related data structured coding vectors to obtain a set of internal relationship score values ​​of the power-related data to be processed and the normal power-related data; The normal power related data structured encoding vector corresponding to the maximum value in the set of internal relationship score values ​​of the to-be-processed power-normal power related data is used as the normal power related data dynamic search anchor vector.

4. The method for processing electric power data based on obtaining an electric power index model according to claim 3, characterized in that: The vector of the starting position of the normal power related data selection range ratio is the normal power related data dynamic search anchor vector, and each normal power related data structured coding vector in the normal power related data selection range ratio is defined as a dynamic search optimized normal power related data structured coding vector.

5. The method for processing electric power data based on obtaining an electric power index model according to claim 4, characterized in that: Based on the normal power-related data selection range ratio, semantic query response encoding is performed on the set of the to-be-processed power-related data structured encoding vector and the normal power-related data structured encoding vector to obtain the power-related data semantic query response encoding vector, including: The structured coding vector of the power-related data to be processed and each dynamically searched optimized normal power-related data structured coding vector in the selection range ratio of the normal power-related data are input into a dynamic semantic search encoder to obtain the power-related data semantic query response coding vector.

6. The method for processing electric power data based on obtaining an electric power index model according to claim 5, characterized in that: Inputting the structured encoding vector of the power-related data to be processed and each of the dynamic search optimized normal power-related data structured encoding vectors in the selection range ratio of the normal power-related data into a dynamic semantic search encoder to obtain the power-related data semantic query response encoding vector, including: Calculating the positional difference between each dynamically searched optimized normal power-related data structured encoding vector and the to-be-processed power-related data structured encoding vector in the normal power-related data selection range ratio to obtain a set of normal power-to-be-processed power-related data differential feature vectors; Calculating the cosine similarity between the structured encoding vector of the power-related data to be processed and each dynamically searched and optimized normal power-related data structured encoding vector in the selection range ratio of the normal power-related data to obtain a set of cosine similarities of the power-related data to be processed and the normal power-related data; Taking the set of cosine similarities of the data related to power to be processed and normal power as a set of weights, the set of differential feature vectors of the data related to power to be processed and normal power is weightedly aggregated, and then the weighted aggregation result is fused with the dynamic search anchor vector of the normal power related data to obtain the semantic query response encoding vector of the power related data.

7. The method for processing electric power data based on obtaining an electric power index model according to claim 6, characterized in that: Determining the abnormal diagnosis result based on the power-related data semantic query response encoding vector includes: The power-related data semantic query response encoding vector is input into a classifier-based data identification module to obtain the abnormality diagnosis result, and the abnormality diagnosis result is used to indicate whether there is an abnormality in the power-related data to be processed.

8. The method for processing electric power data based on obtaining an electric power index model according to claim 7, characterized in that: Inputting the power-related data semantic query response encoding vector into a classifier-based data recognition module to obtain the abnormal diagnosis result, including: Using the fully connected layer of the data identification module to perform fully connected encoding on the power-related data semantic query response encoding vector to obtain a power-related data semantic query response fully connected encoding vector; Inputting the fully connected encoding vector of the power-related data semantic query response into the Softmax classification function of the data identification module to obtain the probability value of the power-related data semantic query response encoding vector belonging to each classification label, wherein the classification label includes the power-related data having anomalies and the power-related data not having anomalies; The classification label corresponding to the largest probability value among the probability values ​​is determined as the abnormal diagnosis result.

9. A power data processing system based on obtaining a power index model, characterized in that: include: A power-related data acquisition module, used to obtain power-related data to be processed; A data anomaly diagnosis module, configured to perform an anomaly diagnosis on the power-related data to be processed based on a set of power-related data marked as normal to obtain an anomaly diagnosis result; The power situation index acquisition module is used for inputting the power related data to be processed into the obtained power index model to obtain the obtained power situation index in response to the abnormality diagnosis result that there is no abnormality in the power related data to be processed.

Images