Electric power engineering intelligent management and control system and method based on big data

Abstract

The invention discloses an electric power engineering intelligent management and control system and method based on big data, and relates to the technical field of engineering risk management and control, and the method comprises the steps: collecting multi-source heterogeneous data in a management and control region, and dividing the multi-source heterogeneous data into suspicious data and non-suspicious data; when the multi-source heterogeneous data are non-suspicious data, carrying out integrating degree analysis and coverage degree analysis on each piece of multi-source heterogeneous data; evaluating the accuracy of the multi-source heterogeneous data; according to an evaluation result, integrating accurate data into a deep learning model; and on the basis of inaccurate data in the multi-source heterogeneous data, predicting an overall management and control risk score of the electric power engineering through a deep learning model. The method has the advantages that the coverage and the consistency of the data are ensured through a multi-level data verification and risk assessment mechanism, the adverse effect of data distortion on electric power engineering management and control is fundamentally solved, and the reliability and the practicability of an electric power engineering management and control system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering risk management and control, and in particular to an intelligent management and control system and method for electric power engineering based on big data. Background Art

[0002] With the rapid development of information technology, big data and artificial intelligence technologies are increasingly being used in engineering control; especially in power engineering control, the use of real-time multi-source heterogeneous data can greatly improve the efficiency and accuracy of engineering risk control. However, in practical applications, the quality of data has become a key issue; multi-source heterogeneous data often come from different data sources, including meteorological characteristic data and population density characteristic data, etc. These data may be inconsistent, incomplete or distorted. If there is multi-source heterogeneous data distortion in the process of training big data analysis models, the multi-source heterogeneous data will not match the actual situation, which may lead to engineering controllers being misled when making or executing decisions; moreover, in the process of emergency response, if the distortion of multi-source heterogeneous data cannot be predicted in advance in a timely manner, it may lead to decision makers' understanding deviation when executing decisions, making it impossible to predict risk events in a timely and accurate manner, increasing engineering management risks. Summary of the invention

[0003] In order to solve the above technical problems, a big data-based intelligent power engineering management and control system and method are provided. This technical solution solves the above-mentioned problem that the distortion of multi-source heterogeneous data cannot be predicted in advance in a timely manner, which may lead to decision makers' understanding deviation when executing decisions, making it impossible to predict risk events in a timely and accurate manner, thereby increasing the risk of engineering management.

[0004] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0005] A method for intelligent control of electric power engineering based on big data, comprising:

[0006] Collect multi-source heterogeneous data within the control area, compare each multi-source heterogeneous data with historical multi-source heterogeneous data, and classify the multi-source heterogeneous data into suspicious data and non-suspicious data based on the comparison results;

[0007] When the multi-source heterogeneous data is non-suspicious data, a fit analysis is performed on each multi-source heterogeneous data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model;

[0008] When the multi-source heterogeneous data is non-suspicious data, the coverage degree analysis is performed on each multi-source heterogeneous data to evaluate the overall coverage of the multi-source heterogeneous data;

[0009] Comprehensively analyze the overall fit between multi-source heterogeneous data and deep learning models and the overall coverage of multi-source heterogeneous data to evaluate the accuracy of multi-source heterogeneous data;

[0010] Based on the evaluation results, the multi-source heterogeneous data are divided into accurate data and inaccurate data, and the accurate data are integrated into the deep learning model;

[0011] Based on the inaccurate data in multi-source heterogeneous data, the overall management and control risk score of the power project is predicted through a deep learning model.

[0012] Preferably, comparing each multi-source heterogeneous data with the historical multi-source heterogeneous data specifically includes:

[0013] The data feature vectors between different independent sources and the feature vectors of corresponding historical data are regarded as a set of data for comparison;

[0014] Represent the new data collected at each time point as a feature vector x n , where n is the number of features and x n The feature vector representing the currently collected multi-source heterogeneous data;

[0015] Among them, the corresponding historical data is represented as the feature vector y n ,y n A feature vector representing multi-source heterogeneous data collected historically;

[0016] Use the Euclidean distance calculation formula to calculate the difference between feature vectors;

[0017] The Euclidean distance calculation formula is:

[0018]

[0019] Among them, d(x, y) is the difference between historical data and current data.

[0020] Preferably, the method of classifying the multi-source heterogeneous data into suspicious data and non-suspicious data according to the comparison result is specifically as follows:

[0021] It is determined whether the difference between the data feature vectors of the independent sources and the feature vectors of the corresponding historical data is greater than or equal to the difference threshold. If so, it is recorded as suspicious data; if not, it is recorded as non-suspicious data.

[0022] Preferably, when the multi-source heterogeneous data is non-suspicious data, performing a fit analysis on each multi-source heterogeneous data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model specifically includes:

[0023] Select features that can reflect the relationship between data and deep learning models from multi-source heterogeneous data as nodes of the Bayesian network model;

[0024] For each node Ys, calculate its conditional probability using the conditional probability calculation formula;

[0025] Wherein, the conditional probability calculation formula is:

[0026] Where P(Yp, Xp) is the joint probability distribution of node Yp and all its parent nodes Xp, P(Xp) is the marginal probability distribution of parent node Xp, and P(Ys|Xs) is the conditional probability of each node Ys;

[0027] For the target variable Xs under the overall goal of the deep learning model, calculate the probability distribution;

[0028] The probability distribution calculation formula is:

[0029] Where P(Xs|Ys) is the probability distribution of the target variable Xs under the overall goal of the deep learning model;

[0030] By calculating the conditional probability P(Ys|Xs) and the probability distribution P(Xs|Ys) of the target variable, the deviation value of the overall fit between multi-source heterogeneous data and the deep learning model is calculated;

[0031] The calculation formula of the fit deviation value is:

[0032]

[0033] Where q is the number of data samples, D z It is the deviation value of the fit between multi-source heterogeneous data and the deep learning model.

[0034] Preferably, when the multi-source heterogeneous data is non-suspicious data, performing coverage analysis on each multi-source heterogeneous data, and evaluating the overall coverage of the multi-source heterogeneous data specifically includes:

[0035] Divide the collected time series data into a series of fixed-length time windows, each of which represents one hour of data;

[0036] The coverage rate of the data source is calculated in each time window. The coverage rate is defined as the area or time proportion actually covered by the data source in the time window.

[0037] The calculation formula for the coverage of the data source is:

[0038]

[0039] Where R(t) represents the coverage rate within the time window t, A(t) is the area or time actually covered within the time window t, and A total is the total area or time;

[0040] Based on the coverage rate in each time window, the coverage fluctuation index is calculated. The coverage fluctuation index reflects the changes in coverage rate between different time windows.

[0041] The calculation formula of the coverage volatility index is:

[0042]

[0043] Where V is the coverage volatility index, N is the number of time windows, and R N is the coverage rate of the Nth time window, R y is the average coverage of all time windows.

[0044] Preferably, the comprehensive analysis of the overall fit between the multi-source heterogeneous data and the deep learning model and the overall coverage of the multi-source heterogeneous data to evaluate the accuracy of the multi-source heterogeneous data specifically includes:

[0045] The overall fit deviation value D of multi-source heterogeneous data and deep learning model z Normalized with the coverage volatility index V, the overall fit deviation value D of the normalized multi-source heterogeneous data and the deep learning model is obtained z Calculate the accuracy coefficient of multi-source heterogeneous data with the coverage volatility index V;

[0046] The accuracy coefficient calculation formula for multi-source heterogeneous data is as follows:

[0047]

[0048] Wherein, SA is the accuracy coefficient of multi-source heterogeneous data, a1 and a2 are preset proportional coefficients, and both a1 and a2 are greater than 0.

[0049] Preferably, the dividing the multi-source heterogeneous data into accurate data and inaccurate data according to the evaluation result specifically includes:

[0050] It is determined whether the accuracy coefficient SA of the multi-source heterogeneous data is greater than or equal to a preset accuracy coefficient threshold of the multi-source heterogeneous data. If so, it is recorded as accurate data; if not, it is recorded as inaccurate data.

[0051] Preferably, the method of predicting the overall control risk score of the power project through a deep learning model based on inaccurate data in multi-source heterogeneous data specifically includes:

[0052] For each inaccurate data L j, calculate the deviation between it and the model prediction result, and use the loss function S to quantify this deviation, where j represents the number of inaccurate data and j is a positive integer greater than 0;

[0053] Wherein, the loss function expression is:

[0054] e j =S(L j ,Ly j );

[0055] Where, L j Indicates the actual observed value, Ly j represents the predicted value of the deep learning model, e j is the prediction bias;

[0056] The prediction deviations of all inaccurate data are summed up to calculate the overall prediction risk score;

[0057] The overall risk calculation formula is:

[0058]

[0059] In the formula, w j is the weight factor for each inaccurate data, e j is the prediction bias value of the jth inaccurate data, and R is the overall prediction risk score.

[0060] Furthermore, a power engineering intelligent management and control system based on big data is proposed, which is used to implement the above-mentioned power engineering intelligent management and control method based on big data, specifically including:

[0061] A multi-source heterogeneous data acquisition module, which is used to collect multi-source heterogeneous data in the control area;

[0062] A multi-source heterogeneous data comparison module, which is used to compare each multi-source heterogeneous data with the historical multi-source heterogeneous data, and classify the multi-source heterogeneous data into suspicious data and non-suspicious data according to the comparison result;

[0063] A fit analysis module, which performs fit analysis on each multi-source heterogeneous data based on non-suspicious data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model;

[0064] A coverage analysis module, which performs coverage analysis on each multi-source heterogeneous data based on non-suspicious data to evaluate the overall coverage of the multi-source heterogeneous data;

[0065] An accuracy division module, which divides the multi-source heterogeneous data into accurate data and inaccurate data according to the evaluation results, and integrates the accurate data into the deep learning model;

[0066] A risk prediction module predicts the overall management and control risk score of the power project through a deep learning model based on inaccurate data in multi-source heterogeneous data.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] (1) By strictly comparing real-time multi-source heterogeneous data with historical data, using Euclidean distance to calculate the difference between new data and historical data, and comparing with the preset difference threshold, the data is divided into suspicious data and non-suspicious data; this process effectively screens out possible distorted data, avoids these data from directly entering the model training link, and reduces the model prediction deviation caused by data distortion; in addition, for non-suspicious data, the overall fit with the deep learning model is evaluated by the fit analysis module, and the conditional probability is calculated by the Bayesian network model to further verify the authenticity and reliability of the data; the coverage analysis module evaluates the coverage of data in different time windows by calculating the coverage fluctuation index to ensure the continuity and integrity of the data in time and space. These steps work together to significantly improve the quality of the data, thereby ensuring the prediction accuracy and stability of the deep learning model, allowing engineering controllers to rely on more reliable prediction results when making decisions, and reducing decision-making errors caused by data distortion. Through the technical solution of the present invention, not only can distorted data be discovered and eliminated in a timely manner, but also the coverage and consistency of the data can be ensured through comprehensive analysis, fundamentally solving the adverse effects of data distortion on power engineering management and control;

[0069] (2) The present invention further ensures that only high-accuracy data will be integrated into the deep learning model by calculating the accuracy coefficient of multi-source heterogeneous data and comparing it with a preset accuracy coefficient threshold. This mechanism not only improves the prediction performance of the model, but also enhances the robustness of the system; for data that is judged to be inaccurate, the present invention calculates the deviation between it and the model prediction result and summarizes it to obtain the overall prediction risk score, thereby evaluating the impact of these inaccurate data on the model prediction; when the overall prediction risk score exceeds the preset threshold, the system will prompt the decision maker to pay attention to the potential risks and recommend taking corresponding remedial measures; this process enables the system to discover and respond to data distortion problems in the first place, avoiding decision-making deviations caused by data quality problems; by comprehensively evaluating the fit and coverage of the data, the technical solution of the present invention can ensure the accuracy and reliability of the data in the entire engineering management and control system, and improve the speed and effect of emergency response; in addition, by dynamically monitoring the accuracy and stability of the data, the present invention can also adjust the model parameters in time and optimize the data processing process, thereby continuously improving the overall performance of the system; in summary, the present invention not only solves the problem of data distortion, but also significantly improves the reliability and practicality of the power engineering management and control system through a multi-level data verification and risk assessment mechanism. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 This is a flow chart of the intelligent management and control method of electric power engineering based on big data proposed by the present invention;

[0071] Figure 2 is a flow chart of a method for calculating differences between feature vectors in the present invention;

[0072] Figure 3 A flow chart of a method for evaluating the overall fit between multi-source heterogeneous data and a deep learning model in the present invention;

[0073] Figure 4 Flowchart of the method for predicting the risks of engineering management and control using the computational deep learning model in the present invention. DETAILED DESCRIPTION

[0074] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0075] Example 1

[0076] See also Figure 1 As shown, the present invention is a method for intelligent management and control of electric power engineering based on big data, comprising the following steps:

[0077] Collect multi-source heterogeneous data in the monitoring area, compare each multi-source heterogeneous data with historical multi-source heterogeneous data, and classify the multi-source heterogeneous data into suspicious data and non-suspicious data according to the comparison results;

[0078] When the multi-source heterogeneous data is non-suspicious data, a fit analysis is performed on each multi-source heterogeneous data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model;

[0079] When the multi-source heterogeneous data is non-suspicious data, the coverage degree analysis is performed on each multi-source heterogeneous data to evaluate the overall coverage of the multi-source heterogeneous data;

[0080] Comprehensively analyze the overall fit between multi-source heterogeneous data and deep learning models and the overall coverage of multi-source heterogeneous data to evaluate the accuracy of multi-source heterogeneous data;

[0081] Based on the evaluation results, the multi-source heterogeneous data are divided into accurate data and inaccurate data, and the accurate data are integrated into the deep learning model;

[0082] Based on the inaccurate data in multi-source heterogeneous data, the overall management and control risk score of the power project is predicted through a deep learning model.

[0083] Example 2

[0084] Based on Example 1, various types of multi-source heterogeneous data are collected from the monitoring area;

[0085] The features of data from different independent sources are represented as vectors, each feature rule is standardized, and a vector score for each rule is set, wherein each feature rule is set with a corresponding score;

[0086] The data feature vectors between different independent sources and the feature vectors of corresponding historical data are regarded as a set of data for comparison;

[0087] For details, please refer to Figure 2 As shown:

[0088] Represent the new data collected at each time point as a feature vector x n , where n is the number of features and x n The feature vector representing the currently collected multi-source heterogeneous data;

[0089] Among them, x n The corresponding historical data is represented as the feature vector y n ,y n A feature vector representing multi-source heterogeneous data collected historically;

[0090] Use the Euclidean distance calculation formula to calculate the difference between feature vectors;

[0091] Among them, the Euclidean distance calculation formula is:

[0092]

[0093] Among them, d(x, y) is the difference between historical data and current data;

[0094] Compare the difference between the data feature vectors from different independent sources and the feature vectors of the corresponding historical data with a difference threshold;

[0095] If the difference between the data feature vectors of the independent sources and the feature vectors of the corresponding historical data is greater than or equal to the difference threshold, the multi-source heterogeneous data is recorded as suspicious data;

[0096] If the difference between the feature vectors of the data from the independent sources and the feature vectors of the corresponding historical data is less than the difference threshold, the multi-source heterogeneous data is recorded as non-suspicious data;

[0097] It should be noted that: the greater the difference between the data feature vectors between different independent sources and the feature vectors of corresponding historical data, the greater the difference between the data between different independent sources and the historical data, and the higher the degree of data suspicion.

[0098] Example 3

[0099] See also Figure 3 As shown, on the basis of Example 2, when the multi-source heterogeneous data is non-suspicious data, a fit analysis is performed on each multi-source heterogeneous data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model, specifically including:

[0100] Select features that can reflect the relationship between data and deep learning models from multi-source heterogeneous data as nodes of the Bayesian network model;

[0101] For each node Ys, calculate its conditional probability using the conditional probability calculation formula;

[0102] The conditional probability calculation formula is:

[0103] Where P(Yp, Xp) is the joint probability distribution of node Yp and all its parent nodes Xp, P(Xp) is the marginal probability distribution of parent node Xp, and P(Ys|Xs) is the conditional probability of each node Ys;

[0104] For the target variable Xs under the overall goal of the deep learning model, calculate the probability distribution;

[0105] The probability distribution calculation formula is:

[0106] Where P(Xs|Ys) is the probability distribution of the target variable Xs under the overall goal of the deep learning model;

[0107] By calculating the conditional probability P(Ys|Xs) and the probability distribution P(Xs|Ys) of the target variable, the deviation value of the overall fit between multi-source heterogeneous data and the deep learning model is calculated;

[0108] The specific calculation formula is:

[0109]

[0110] Where q is the number of data samples, D z It is the deviation value of the fit between multi-source heterogeneous data and the deep learning model;

[0111] It should be noted that: if the deviation value of the fit between multi-source heterogeneous data and the deep learning model is smaller, it means that the fit between the multi-source heterogeneous data and the deep learning model is higher; if the deviation value of the fit between the multi-source heterogeneous data and the deep learning model is larger, it means that the fit between the multi-source heterogeneous data and the deep learning model is lower.

[0112] When the multi-source heterogeneous data is non-suspicious data, the coverage degree of each multi-source heterogeneous data is analyzed to obtain the coverage fluctuation index during the update process of the non-suspicious data, and evaluate the overall coverage of the multi-source heterogeneous data. The process of obtaining the coverage fluctuation index during the update process of the non-suspicious data is as follows:

[0113] Divide the collected time series data into a series of fixed-length time windows, and divide a day's data into 24-hour time windows, with each window representing one hour's data;

[0114] The coverage rate of the data source is calculated in each time window. The coverage rate is defined as the area or time proportion actually covered by the data source in the time window.

[0115] The calculation formula for the coverage of the data source is:

[0116]

[0117] Where R(t) represents the coverage rate within the time window t, A(t) is the area or time actually covered within the time window t, and A total is the total area or time;

[0118] Based on the coverage rate in each time window, the coverage fluctuation index is calculated. The coverage fluctuation index reflects the changes in coverage rate between different time windows.

[0119] The calculation formula of the coverage volatility index is:

[0120]

[0121] Where V is the coverage volatility index, N is the number of time windows, N = 1, 2, ..., m, R N is the coverage rate of the Nth time window, R y is the average coverage of all time windows;

[0122] It should be noted that the lower the coverage fluctuation index, the relatively stable the data coverage; the higher the coverage fluctuation index, the greater the volatility of the data coverage. This fluctuation index can help us evaluate the overall coverage and stability of the data, thereby better supporting the operation of the power engineering management and control system.

[0123] Comprehensively analyze the overall fit between multi-source heterogeneous data and deep learning models and the overall coverage of multi-source heterogeneous data, calculate the data accuracy coefficient, and evaluate the accuracy of multi-source heterogeneous data;

[0124] Among them, the process of obtaining the data accuracy coefficient is:

[0125] The overall fit deviation value D of multi-source heterogeneous data and deep learning model z Normalized with the coverage volatility index V, the overall fit deviation value D of the normalized multi-source heterogeneous data and the deep learning model is obtained z Calculate the accuracy coefficient of multi-source heterogeneous data with the coverage volatility index V;

[0126] The calculation formula for the accuracy coefficient of multi-source heterogeneous data is:

[0127]

[0128] Wherein, SA is the accuracy coefficient of multi-source heterogeneous data, a1 and a2 are preset proportional coefficients, and a1 and a2 are both greater than 0;

[0129] It should be noted that the overall fit deviation value D of multi-source heterogeneous data and deep learning model z The coverage volatility index V is positively correlated with the accuracy coefficient SA of multi-source heterogeneous data, and when the overall fit deviation value D of multi-source heterogeneous data and the deep learning model z The larger the coverage volatility index V is, the larger the accuracy coefficient SA of multi-source heterogeneous data is, and the higher the accuracy of the data is.

[0130] Example 5

[0131] On the basis of Embodiment 4, the accuracy coefficient SA of the multi-source heterogeneous data is compared with a preset accuracy coefficient threshold of the multi-source heterogeneous data;

[0132] If the accuracy coefficient SA of the multi-source heterogeneous data is greater than or equal to the preset accuracy coefficient threshold of the multi-source heterogeneous data, it means that the multi-source heterogeneous data is accurate and is recorded as accurate data;

[0133] If the accuracy coefficient SA of the multi-source heterogeneous data is less than the preset accuracy coefficient threshold of the multi-source heterogeneous data, it means that the multi-source heterogeneous data is inaccurate and is recorded as inaccurate data;

[0134] Integrate accurate data from multi-source heterogeneous data into deep learning models to facilitate subsequent predictions for engineering control.

[0135] See also Figure 4 As shown in the figure, based on the inaccurate data of multi-source heterogeneous data, the deep learning model is calculated to predict the risk of engineering control. The specific calculation process is:

[0136] For each inaccurate data L j , calculate the deviation between it and the model prediction result. Use the loss function S to quantify this deviation, where j represents the number of inaccurate data and j is a positive integer greater than 0;

[0137] Among them, the loss function expression is:

[0138] e j =S(L j ,Ly j );

[0139] Where, L j Indicates the actual observed value, Ly j represents the predicted value of the deep learning model, e j is the prediction bias;

[0140] The prediction deviations of all inaccurate data are summed up to calculate the overall prediction risk score;

[0141] The overall risk calculation formula is:

[0142]

[0143] In the formula, w j is the weight factor for each inaccurate data, e j is the prediction bias value of the jth inaccurate data, and R is the overall prediction risk score;

[0144] The overall predicted risk score R is compared with a preset threshold;

[0145] If the overall prediction risk score R is greater than or equal to the preset threshold, it means that the inaccurate data has a high risk for the overall prediction of deep learning, resulting in instability of the overall prediction of deep learning;

[0146] If the overall prediction risk score R is less than the preset threshold, it means that the inaccurate data has a low risk for the overall prediction of deep learning and will not cause instability in the overall prediction of deep learning.

[0147] It should be noted that the deep learning model can be further optimized or corresponding remedial measures can be taken by calculating the risks brought by inaccurate data.

[0148] Example 6

[0149] An intelligent control system for electric power engineering based on big data, comprising:

[0150] Multi-source heterogeneous data acquisition module, which is used to collect multi-source heterogeneous data in the monitoring area;

[0151] A multi-source heterogeneous data comparison module is used to compare each multi-source heterogeneous data with the historical multi-source heterogeneous data, and classify the multi-source heterogeneous data into suspicious data and non-suspicious data according to the comparison result;

[0152] The fit analysis module performs fit analysis on each multi-source heterogeneous data based on non-suspicious data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model.

[0153] Coverage analysis module: The coverage analysis module performs coverage analysis on each multi-source heterogeneous data based on non-suspicious data to evaluate the overall coverage of the multi-source heterogeneous data.

[0154] The accuracy division module divides the multi-source heterogeneous data into accurate data and inaccurate data according to the evaluation results, and integrates the accurate data into the deep learning model;

[0155] Risk prediction module,The risk prediction module calculates the deep learning model based on the inaccurate data in multi-source heterogeneous data to predict the risks of engineering control.

[0156] The working principle of the present invention is to improve the accuracy and reliability of engineering control response by collecting, comparing, analyzing and evaluating multi-source heterogeneous data; specifically, the technical solution includes a multi-source heterogeneous data collection module, which is used to collect multi-source heterogeneous data from the monitoring area; a multi-source heterogeneous data comparison module, which judges the suspicious degree of data by calculating the Euclidean distance between new data and historical data, and divides the data into suspicious data and non-suspicious data; a fit analysis module, which performs fit analysis on non-suspicious data, evaluates its overall fit with the deep learning model, and calculates the conditional probability through the Bayesian network model; a coverage analysis module, which evaluates the coverage of non-suspicious data in different time windows, and calculates the coverage fluctuation index; an accuracy classification module, which calculates the accuracy coefficient of the data by combining the fit deviation value and the coverage fluctuation index, and divides the data into accurate data and inaccurate data according to a preset threshold; a risk prediction module, which calculates the risk of engineering control predicted by the deep learning model based on inaccurate data, and evaluates the prediction stability of the model by summarizing the prediction deviation and calculating the overall risk score. This technical solution effectively improves the data quality of the power engineering management and control system through a systematic multi-step data processing and evaluation process, enhances the accuracy and stability of predictions, and thus better supports power engineering management and decision-making.

[0157] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0158] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0159] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0160] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0161] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A method for intelligent management and control of electric power engineering based on big data, characterized in that: include: Collect multi-source heterogeneous data within the control area, compare each multi-source heterogeneous data with historical multi-source heterogeneous data, and classify the multi-source heterogeneous data into suspicious data and non-suspicious data based on the comparison results; When the multi-source heterogeneous data is non-suspicious data, a fit analysis is performed on each multi-source heterogeneous data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model; When the multi-source heterogeneous data is non-suspicious data, the coverage degree analysis is performed on each multi-source heterogeneous data to evaluate the overall coverage of the multi-source heterogeneous data; Comprehensively analyze the overall fit between multi-source heterogeneous data and deep learning models and the overall coverage of multi-source heterogeneous data to evaluate the accuracy of multi-source heterogeneous data; Based on the evaluation results, the multi-source heterogeneous data are divided into accurate data and inaccurate data, and the accurate data are integrated into the deep learning model; Based on the inaccurate data in multi-source heterogeneous data, the overall management and control risk score of the power project is predicted through a deep learning model.

2. According to the big data-based intelligent management and control method for electric power engineering according to claim 1, it is characterized in that: The comparing each multi-source heterogeneous data with the historical multi-source heterogeneous data specifically includes: The data feature vectors between different independent sources and the feature vectors of corresponding historical data are regarded as a set of data for comparison; Represent the new data collected at each time point as a feature vector x n , where n is the number of features and x n The feature vector representing the currently collected multi-source heterogeneous data; Among them, x n The corresponding historical data is represented as the feature vector y n ,y n A feature vector representing multi-source heterogeneous data collected historically; Use the Euclidean distance calculation formula to calculate the difference between feature vectors; The Euclidean distance calculation formula is: Among them, d(x, y) is the difference between historical data and current data.

3. The method for intelligent management and control of electric power engineering based on big data according to claim 2 is characterized in that: The method of classifying multi-source heterogeneous data into suspicious data and non-suspicious data according to the comparison results is specifically as follows: It is determined whether the difference between the data feature vectors of the independent sources and the feature vectors of the corresponding historical data is greater than or equal to the difference threshold. If so, it is recorded as suspicious data; if not, it is recorded as non-suspicious data.

4. The method for intelligent management and control of electric power engineering based on big data according to claim 3 is characterized in that: When the multi-source heterogeneous data is non-suspicious data, performing a fit analysis on each multi-source heterogeneous data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model specifically includes: Select features that can reflect the relationship between data and deep learning models from multi-source heterogeneous data as nodes of the Bayesian network model; For each node Ys, calculate its conditional probability using the conditional probability calculation formula; Wherein, the conditional probability calculation formula is: Where P(Yp, Xp) is the joint probability distribution of node Yp and all its parent nodes Xp, P(Xp) is the marginal probability distribution of parent node Xp, and P(Ys|Xs) is the conditional probability of each node Ys; For the target variable Xs under the overall goal of the deep learning model, calculate the probability distribution; The probability distribution calculation formula is: Where P(Xs|Ys) is the probability distribution of the target variable Xs under the overall goal of the deep learning model; By calculating the conditional probability P(Ys|Xs) and the probability distribution P(Xs|Ys) of the target variable, the deviation value of the overall fit between multi-source heterogeneous data and the deep learning model is calculated; The calculation formula of the fit deviation value is: Where q is the number of data samples, D z It is the deviation value of the fit between multi-source heterogeneous data and the deep learning model.

5. The method for intelligent management and control of electric power engineering based on big data according to claim 4 is characterized in that: When the multi-source heterogeneous data is non-suspicious data, performing coverage analysis on each multi-source heterogeneous data and evaluating the overall coverage of the multi-source heterogeneous data specifically includes: Divide the collected time series data into a series of fixed-length time windows, each of which represents one hour of data; The coverage rate of the data source is calculated in each time window. The coverage rate is defined as the area or time proportion actually covered by the data source in the time window. The calculation formula for the coverage of the data source is: Where R(t) represents the coverage rate within the time window t, A(t) is the area or time actually covered within the time window t, and A total is the total area or time; Based on the coverage rate in each time window, the coverage fluctuation index is calculated. The coverage fluctuation index reflects the changes in coverage rate between different time windows. The calculation formula of the coverage volatility index is: Where V is the coverage volatility index, N is the number of time windows, and R N is the coverage rate of the Nth time window, R y is the average coverage of all time windows.

6. The method for intelligent management and control of electric power engineering based on big data according to claim 5 is characterized in that: The comprehensive analysis of the overall fit between the multi-source heterogeneous data and the deep learning model and the overall coverage of the multi-source heterogeneous data to evaluate the accuracy of the multi-source heterogeneous data specifically includes: The overall fit deviation value D of multi-source heterogeneous data and deep learning model z Normalized with the coverage volatility index V, the overall fit deviation value D of the normalized multi-source heterogeneous data and the deep learning model is obtained z Calculate the accuracy coefficient of multi-source heterogeneous data with the coverage volatility index V; The accuracy coefficient calculation formula for multi-source heterogeneous data is as follows: Wherein, SA is the accuracy coefficient of multi-source heterogeneous data, a1 and a2 are preset proportional coefficients, and both a1 and a2 are greater than 0.

7. The method for intelligent management and control of electric power engineering based on big data according to claim 6 is characterized in that: The method of classifying the multi-source heterogeneous data into accurate data and inaccurate data according to the evaluation results specifically includes: It is determined whether the accuracy coefficient SA of the multi-source heterogeneous data is greater than or equal to a preset accuracy coefficient threshold of the multi-source heterogeneous data. If so, it is recorded as accurate data; if not, it is recorded as inaccurate data.

8. The method for intelligent management and control of electric power engineering based on big data according to claim 7 is characterized in that: The method of predicting the overall control risk score of the power project through a deep learning model based on inaccurate data in multi-source heterogeneous data specifically includes: For each inaccurate data L j , calculate the deviation between it and the model prediction result, and use the loss function S to quantify this deviation, where j represents the number of inaccurate data and j is a positive integer greater than 0; Wherein, the loss function expression is: and j =S(L j ,Ly j ); Where, L j Indicates the actual observed value, Ly j represents the predicted value of the deep learning model, e j is the prediction bias; The prediction deviations of all inaccurate data are summed up to calculate the overall prediction risk score; The overall risk calculation formula is: In the formula, w j is the weight factor for each inaccurate data, e j is the prediction bias value of the jth inaccurate data, and R is the overall prediction risk score.

9. An intelligent management and control system for power engineering based on big data, characterized in that: The method for intelligent control of electric power engineering based on big data as claimed in any one of claims 1 to 8 specifically comprises: A multi-source heterogeneous data acquisition module, which is used to collect multi-source heterogeneous data in the control area; A multi-source heterogeneous data comparison module, which is used to compare each multi-source heterogeneous data with the historical multi-source heterogeneous data, and classify the multi-source heterogeneous data into suspicious data and non-suspicious data according to the comparison result; A fit analysis module, which performs fit analysis on each multi-source heterogeneous data based on non-suspicious data to evaluate the overall fit between the multi-source heterogeneous data and the deep learning model; A coverage analysis module, which performs coverage analysis on each multi-source heterogeneous data based on non-suspicious data to evaluate the overall coverage of the multi-source heterogeneous data; An accuracy division module, which divides the multi-source heterogeneous data into accurate data and inaccurate data according to the evaluation results, and integrates the accurate data into the deep learning model; A risk prediction module predicts the overall management and control risk score of the power project through a deep learning model based on inaccurate data in multi-source heterogeneous data.

Images