Power load data situation awareness-oriented data mining visualization method and system

Abstract

The invention discloses a power load data situation awareness-oriented data mining visualization method and system, and the method comprises the steps: collecting power load data and carrying out the preprocessing of the power load data, and the preprocessing comprises data cleaning, normalization and abnormal value processing; multi-dimensional features of the preprocessed power load data are extracted, and the multi-dimensional features are extracted through statistical analysis and a machine learning algorithm; on the basis of the extracted multi-dimensional features, a comprehensive situation index is constructed, and the comprehensive situation index is used for reflecting the load situation; and the load situation information is visually displayed through a graphical interface. The system is used for executing the method. According to the power load data situation awareness-oriented data mining visualization method and system provided by the invention, the internal law of load change can be deeply revealed through a data mining technology, and a complex data situation can be presented in a simple and understandable manner through a visualization means.

Description

Technical Field

[0001] The present invention relates to the technical field of power system management, and in particular to a data mining visualization method and system for power load data situation awareness. Background Art

[0002] With the rapid development and widespread application of smart grid technology, the power system continuously generates massive load data during operation. These data not only include traditional power consumption information, but also cover multi-dimensional characteristics such as power consumption time, geographical location, and user type. Effective analysis and utilization of these data have important practical significance and academic value for improving power grid operation efficiency, ensuring power supply safety, optimizing resource allocation, and realizing intelligent management.

[0003] However, in the face of such large-scale, high-dimensional and rapidly updated load data, traditional data analysis methods are increasingly showing their limitations. First, conventional database systems and analysis tools are often unable to handle TB or even PB-level load data, resulting in low data processing efficiency and failure to meet the needs of real-time analysis. Second, traditional statistical analysis methods are difficult to extract deep-level features and patterns from complex load data, and are prone to ignore important information contained in the data, thus affecting the accuracy of subsequent decisions. In addition, in terms of load forecasting, the prediction accuracy of classic time series models is often unsatisfactory when dealing with nonlinear load data affected by multiple factors, making it difficult to provide a reliable basis for power grid dispatching and planning. In terms of data visualization and situational awareness, existing technologies also have many shortcomings. The current data display methods are mostly based on static tables and simple charts, which are difficult to intuitively and comprehensively present complex load situations, and are not conducive to decision makers quickly grasping the overall situation and identifying potential problems. At the same time, many existing systems lack real-time analysis and display capabilities, resulting in a lag in decision information and failure to meet the needs of rapid response of power systems. In terms of early warning mechanisms, most systems still remain in a simple early warning mode based on fixed thresholds, which cannot adapt to complex and changeable load conditions and is difficult to effectively prevent and respond to potential power safety risks. In recent years, academia and industry have begun to try to apply advanced technologies such as data mining and machine learning to power load analysis. These studies have made certain progress in specific areas, such as improving the accuracy of load forecasting or improving the effect of data visualization. However, these attempts are often limited to a specific link and lack systematic and holistic considerations. For example, some studies focus on the optimization of load forecasting algorithms, but ignore the visualization of forecasting results; some work focuses on the development of data visualization platforms, but is relatively weak in data mining and feature extraction; and some studies have proposed new load anomaly detection algorithms, but lack supporting early warning mechanisms and decision support systems.

[0004] In view of the above problems and limitations, a data mining visualization method and system for power load data situation awareness is proposed to overcome the limitations caused by the single function in the existing technology, open up new ways for the in-depth utilization of power load data, and promote the entire power industry to develop in a more efficient, safer and smarter direction, which has become an urgent problem to be solved. Summary of the invention

[0005] The purpose of the present invention is to provide a data mining visualization method and system for power load data situation awareness. The method deeply analyzes load data through data mining technology, and intuitively presents load situation in combination with visualization technology, which can assist in realizing intelligent monitoring and early warning of power systems.

[0006] On the one hand, the present invention provides a data mining visualization method for power load data situation awareness, comprising:

[0007] Collecting power load data and preprocessing the power load data, wherein the preprocessing includes data cleaning, normalization and outlier processing;

[0008] Extracting multidimensional features of the preprocessed power load data, wherein the multidimensional features are extracted using statistical analysis and machine learning algorithms;

[0009] Constructing a comprehensive situation index based on the extracted multi-dimensional features, wherein the comprehensive situation index is used to reflect the load situation;

[0010] The load status information is intuitively displayed through a graphical interface.

[0011] Preferably, the data cleaning includes removing duplicate data and filling missing values; normalization is used to unify data of different dimensions to the same scale; and outlier processing is used to identify and process abnormal data points through statistical methods.

[0012] Further preferably, the multi-dimensional features of the pre-processed power load data are extracted by combining signal decomposition, statistical analysis and machine learning algorithms.

[0013] Further preferably, constructing a comprehensive situation index based on the extracted multi-dimensional features includes:

[0014] Designing an index system, wherein the indexes in the index system are divided into load characteristic indexes, power quality indexes and load prediction indexes;

[0015] The multi-dimensional features are integrated and synthesized to form a comprehensive situation index that can reflect the power load situation;

[0016] The comprehensive situation indicators of the design are calculated and evaluated to ensure that they can effectively reflect the actual situation and changing trend of the power load.

[0017] Further preferably, the graphical interface includes a time series graph, a heat map, a bar graph and a pie chart.

[0018] The present invention also provides a data mining and visualization system for power load data situation awareness, comprising:

[0019] A power load data collection and preprocessing module, used to collect power load data and preprocess the power load data, wherein the preprocessing includes data cleaning, normalization and outlier processing;

[0020] A multidimensional feature extraction module, used to extract multidimensional features of the preprocessed power load data, wherein the multidimensional features are extracted using statistical analysis and machine learning algorithms;

[0021] A situation indicator construction module, used to construct a comprehensive situation indicator based on the extracted multi-dimensional features, wherein the comprehensive situation indicator is used to reflect the load situation;

[0022] The graphical interface display module is used to intuitively display load status information through a graphical interface.

[0023] Preferably, the data cleaning includes removing duplicate data and filling missing values; normalization is used to unify data of different dimensions to the same scale; and outlier processing is used to identify and process abnormal data points through statistical methods.

[0024] Further preferably, the multi-dimensional features of the pre-processed power load data are extracted by combining signal decomposition, statistical analysis and machine learning algorithms.

[0025] Further preferably, constructing a comprehensive situation index based on the extracted multi-dimensional features includes:

[0026] Designing an index system, wherein the indexes in the index system are divided into load characteristic indexes, power quality indexes and load prediction indexes;

[0027] The multi-dimensional features are integrated and synthesized to form a comprehensive situation index that can reflect the power load situation;

[0028] The comprehensive situation indicators of the design are calculated and evaluated to ensure that they can effectively reflect the actual situation and changing trend of the power load.

[0029] Further preferably, the graphical interface includes a time series graph, a heat map, a bar graph and a pie chart.

[0030] The data mining visualization method and system for power load data situation awareness provided by the present invention can not only improve the monitoring, analysis and management capabilities of the power system, but also provide important support for the safe and stable operation of the power system. First, through data collection and preprocessing, data quality and consistency can be ensured, providing a reliable basis for subsequent analysis; secondly, in terms of multidimensional feature extraction, statistical analysis and machine learning algorithms are used to deeply mine the complex characteristics of power load data, such as EMD, EEMD, VMD, SVM and neural network methods, so as to provide more comprehensive and accurate situation awareness information; then, a situation indicator system is constructed, which can fully reflect the actual situation and change trend of power load, and provide important reference for the operation and management of the power system; finally, the visualization design intuitively displays the load situation through a variety of technologies such as time series diagrams, heat maps, bar charts and pie charts, and users can dynamically adjust and view data in different time periods and regions to improve analysis and decision-making efficiency. The constructed real-time visualization system can dynamically update the situation indicators, and automatically generate warning information when exceeding the preset threshold, timely discover and handle abnormal situations, and ensure the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A flow chart of a data mining visualization method for power load data situation awareness provided by the present invention; DETAILED DESCRIPTION

[0032] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0033] like Figure 1 As shown, the present invention provides a data mining visualization method for power load data situation awareness, comprising:

[0034] S1: Collecting power load data and preprocessing the power load data, wherein the preprocessing includes data cleaning, normalization and outlier processing;

[0035] Among them, data cleaning includes removing duplicate data and filling missing values; normalization is used to unify data of different dimensions to the same scale; outlier processing is used to identify and process abnormal data points through statistical methods;

[0036] The relevant preprocessing methods can adopt the following methods:

[0037] 1. Missing data processing

[0038] Network delays, system freezes, and other factors increase the possibility of missing individual data. In this case, the average value of the adjacent sampling points before and after the missing data is usually used to fill in the missing data. If the missing data is continuous, it is often caused by power outages and unexpected tripping. At this time, the situation is relatively cumbersome, which limits the scope of use of the classic interpolation method and weighted average method, and requires reference to expert experience. Referring to the load day data similar to the missing data, it is necessary to consider the external environmental factors at the time, and to make corrections and fill in the data with human experience;

[0039] 2. Normalization

[0040] Normalization of data does not mean scaling it by a specific percentage within a set area, so the negative impact of the data's own attributes on the estimation accuracy can also be avoided. The collected raw data of power load, temperature, week type, etc. belong to different data types, and the standard model cannot be directly estimated during data processing. In view of this, the data needs to be normalized. The three commonly used normalization methods include: extreme value method, standard deviation, and mean normalization method.

[0041] (1) Normalization of historical load data

[0042] Using the extreme value method, normalized formula:

[0043]

[0044] Denormalization formula:

[0045] Y d =(Y max -Y min )×Y d +Y min (6)

[0046] Where: Y max and Y min are the maximum and minimum values ​​in the sequence Y;

[0047] Y d ——historical load data;

[0048] Y d ′——normalized historical load data, the value is [0,1].

[0049] (2) Normalization of meteorological data

[0050] Meteorological data mainly include maximum temperature, minimum temperature, humidity, etc. To ensure the accuracy of model training and prediction, meteorological data also needs to be normalized. By analogy with the normalization method of historical load data, the temperature and humidity normalization formula is obtained:

[0051]

[0052] Where: T max and T min are the maximum and minimum values ​​of temperature and humidity respectively;

[0053] T m ′—normalized temperature data, value is [0,1];

[0054] T m —Original temperature and humidity data.

[0055] (3) Time data normalization processing

[0056] After the above analysis of the characteristic point data of the power load, the time type can determine the scale of the power load. Among them, we focus on working days and weekends, and major holidays are not included in the investigation. For comprehensive urban power systems, measures with larger values ​​are usually used on working days to adjust the pressure of the power system, while smaller values ​​are used on weekends to relieve the pressure of the power system. In order to better predict the power load on working days and non-working days, the expected deviation can be reduced and the expected accuracy can be increased by quantifying the week type. The specific quantification scale is shown in Table 1.

[0057] Table 1 Selection of time type

[0058] Time Type on Monday Tuesday Wednesday Thursday Friday Saturday Sunday Value 0.5 0.6 0.6 0.6 0.6 0.8 0.8

[0059] 3. Abnormal data processing

[0060] Under normal circumstances, the power load data changes continuously. If a sudden change occurs, two methods, horizontal and vertical, can be used to process abnormal data:

[0061] Horizontal processing: In the horizontal direction, due to the continuity of data, the load mutation amplitude will not be too large. Formula (1) is used to determine whether the data is abnormal, and formula (2) is used to correct it;

[0062] The discriminant formula is as follows:

[0063] max[W(n,t)-W(n,t-1),∣W(n,t)-W(n,t+1)∣]>ε(t) (1)

[0064] The correction formula is as follows:

[0065]

[0066] Where: ε(t)—threshold;

[0067] W(n,t)—power load value at time t on day n;

[0068] Vertical processing: The changes in load have their inherent regularity and require a certain process. The two adjacent sampling nodes are not much different in the vertical direction. If the data of a node is significantly different from the data of the adjacent node, then the point must be abnormal data.

[0069] The discriminant formula is as follows:

[0070] |W(n,t)-m(t)|>r(t) (3)

[0071] The correction formula is as follows:

[0072]

[0073] Where: t—sampling node; r(t)—threshold;

[0074] W(n,t)—power load value at time t on day n;

[0075] m(t)—the average load value of the sampling point in recent days.

[0076] S2: Extracting multidimensional features of the preprocessed power load data, wherein the multidimensional features can be mined using statistical analysis and machine learning algorithms, such as wavelet decomposition, empirical mode decomposition (EMD), ensemble empirical mode decomposition (EEMD) and variational mode decomposition (VMD). These methods can effectively extract the main features of the power load data and identify the key factors affecting load changes;

[0077] Multidimensional feature extraction is a key step in power load data situation awareness. The use of statistical analysis and machine learning algorithms to mine data features specifically includes the following parts:

[0078] 1. Application of signal decomposition technology

[0079] To extract the main features from the power load data, the following signal decomposition techniques are used:

[0080] (1) Empirical Mode Decomposition (EMD): Decomposes complex power load signals into several intrinsic mode functions (IMFs), each of which represents a different frequency component of the original signal. EMD can adaptively decompose nonlinear and non-stationary signals and is particularly suitable for feature extraction of power load data.

[0081] (2) Ensemble empirical mode decomposition (EEMD): Based on EMD, EEMD overcomes the modal aliasing problem of EMD by adding Gaussian white noise and performing multiple decompositions to obtain the average, thus further improving the stability and accuracy of the decomposition.

[0082] (3) Variational mode decomposition (VMD): VMD decomposes the signal into several intrinsic mode functions through variational optimization method, which can better handle the noise and instability in the signal and is particularly suitable for the decomposition and feature extraction of complex power load signals.

[0083] 2. Application of statistical analysis techniques

[0084] Statistical analysis technology can extract important features from data by calculating and analyzing the statistical characteristics of data, such as:

[0085] (1) Mean and variance: Calculate the mean and variance of power load data to reflect the central tendency and dispersion of the data;

[0086] (2) Kurtosis and skewness: By calculating the kurtosis and skewness of the data, we can understand the shape and symmetry of the data distribution, which is very important for identifying abnormal load data;

[0087] (3) Autocorrelation and cross-correlation: Autocorrelation and cross-correlation are used to analyze the time series characteristics of power load data and identify periodicity and correlation characteristics in the data;

[0088] 3. Application of machine learning algorithms

[0089] Through machine learning algorithms, more complex and in-depth features can be automatically extracted from power load data, including:

[0090] (1) Support vector machine (SVM): The power load data is classified and regressed using the SVM algorithm to extract features that can distinguish different load patterns.

[0091] (2) Neural Network (NN): The neural network model is used to perform nonlinear mapping and feature learning on the power load data to extract the potential features and patterns in the data;

[0092] (3) Decision tree and random forest: Decision tree and random forest algorithms are used to extract features with high information gain from power load data to help identify key influencing factors.

[0093] 4. Feature selection and dimensionality reduction

[0094] After extracting multidimensional features, feature selection and dimensionality reduction are required to reduce computational complexity and improve the generalization ability of the model. Common methods include:

[0095] (1) Principal component analysis (PCA): PCA is used to map high-dimensional data into a low-dimensional space, retaining the main information components in the data and removing redundancy and noise.

[0096] (2) Linear Discriminant Analysis (LDA): LDA method is used to reduce the dimension while retaining the differences between data classes, thereby improving the discriminative ability of features;

[0097] (3) Recursive feature elimination (RFE): Recursively select features through the RFE method, remove features that contribute less to the model, and optimize the feature set;

[0098] 5. Fusion and construction of multi-dimensional features

[0099] After completing feature extraction, selection and dimensionality reduction, multi-dimensional features need to be fused and constructed to form a comprehensive feature set for subsequent situation indicator construction and model training. These comprehensive feature sets can more comprehensively and accurately reflect the situation of power load data.

[0100] By combining signal decomposition, statistical analysis and machine learning algorithms, rich and diverse features can be extracted from power load data, providing a solid foundation for subsequent situation indicator construction and prediction.

[0101] S3: constructing a comprehensive situation index based on the extracted multi-dimensional features, wherein the comprehensive situation index is used to reflect the load situation;

[0102] Among them, a comprehensive situation index reflecting the power load situation is constructed based on the extracted multi-dimensional features. The specific steps include:

[0103] S31: Design an indicator system, wherein the indicators in the indicator system are divided into the following categories:

[0104] (1) Load characteristic indicators: including the average value, maximum value, minimum value, standard deviation, kurtosis and skewness of the load. These indicators can reflect the basic statistical characteristics and distribution pattern of the power load and help understand the fluctuation and concentration trend of the load. For example, the standard deviation of the load can be used to evaluate the volatility of the load, and the kurtosis and skewness can be used to analyze the distribution pattern of the load data.

[0105] (2) Power quality indicators: including parameters such as voltage, current, frequency and power. These indicators can reflect the operating status and quality of the power system and are crucial to ensuring the safe and stable operation of the power system. For example, the stability of voltage and frequency is an important guarantee for the normal operation of the power system. By monitoring these indicators, abnormal conditions in the power system can be discovered and handled in a timely manner.

[0106] (3) Load forecasting indicators: Based on historical data and forecasting models, they provide forecast values ​​and confidence intervals for future load changes. These indicators can help power system operators plan and dispatch in advance. For example, by evaluating the output results of the load forecasting model, the trend and uncertainty of future load changes can be obtained, so as to formulate corresponding dispatching strategies.

[0107] S32: Multi-dimensional feature fusion and synthesis

[0108] The multi-dimensional features are integrated and synthesized to form a comprehensive situation index that can reflect the power load situation. The specific method includes:

[0109] (1) Weighted average: Different weights are assigned to different features, and weighted averages are performed according to their importance to form a comprehensive situation index. For example, the volatility, periodicity, and randomness characteristics of the load can be weighted and integrated to form a load change situation index. The weighted average method can make full use of the information of each feature and improve the accuracy and reliability of the comprehensive situation index.

[0110]

[0111] Among them, w i is the weight of the i-th feature, f i is the value of the ith feature.

[0112] (2) Feature selection: Through feature selection algorithms, the most important features for situation awareness are screened out, redundant features are removed, and the effectiveness and simplicity of the comprehensive situation indicators are improved. For example, the recursive feature elimination (RFE) method can be used to optimize the feature set by recursively training the model and removing features that contribute less to the model.

[0113] (3) Principal Component Analysis (PCA): PCA is used to map high-dimensional data to low-dimensional space, retain the main information components in the data, and remove redundancy and noise. PCA can effectively reduce the dimension of data and improve the effectiveness and simplicity of features. For example, through the PCA method, multidimensional features can be reduced to a few principal components, which can retain the information of the original features to the greatest extent and form a comprehensive situation indicator.

[0114] The formula for PCA is as follows:

[0115] Z = XW;

[0116] Among them, Z is the data after dimensionality reduction, X is the original data matrix, and W is the eigenvector matrix.

[0117] S33: Indicator calculation and evaluation

[0118] Calculate and evaluate the designed comprehensive situation indicators to ensure that they can effectively reflect the actual situation and change trend of power load. The specific steps include:

[0119] Index calculation: Calculate the values ​​of various comprehensive indicators based on the extracted features and fusion methods. For example, calculate the mean and variance of the load through statistical analysis, and calculate the predicted value and confidence interval of the future load through the prediction model.

[0120] Indicator evaluation: Evaluate the effectiveness and accuracy of situation indicators through actual data verification and simulation analysis. Compare the indicator values ​​calculated by different methods and select the optimal calculation method and parameter settings. For example, evaluate the accuracy of load forecast indicators through error analysis, and verify the effectiveness of power quality indicators through comparison with measured data.

[0121] S4: intuitively display load status information through a graphical interface;

[0122] Among them, in the present invention, a dynamically updated visualization system is constructed and an early warning threshold is set. The power load data is monitored in real time. When the load exceeds the preset threshold, the system can automatically generate early warning information to remind the staff to take corresponding measures. The visualization system also supports historical data query and analysis to help load trend prediction and anomaly detection.

[0123] The design principles of the graphical interface are as follows:

[0124] Simplicity: The interface design should be concise and clear, avoiding excessive unnecessary information that distracts the user's attention. Through concise design, users can quickly find the information they need and improve the efficiency of data analysis.

[0125] Intuitiveness: Use visual elements such as graphics, charts and colors to intuitively display the power load situation, which is convenient for users to quickly understand. Intuitive display can help users better grasp the changing trends and characteristics of the load.

[0126] Interactivity: Provide interactive functions to allow users to dynamically adjust and view the load status of different time periods and different areas. Through interactive design, users can flexibly operate the interface to obtain the required detailed information.

[0127] Among them, the selection and application of data visualization technology is crucial to intuitively display the power load situation. It mainly includes the following technologies:

[0128] Time series graph: used to display the historical data and change trend of power load. For example, you can draw a daily load curve, a weekly load trend graph, etc. Time series graphs can show the dynamic process of load changes over time, helping users understand the periodicity and volatility of load.

[0129] Heat map: reflects the load situation in different areas or time periods through color depth. For example, a heat map can be used to show the load distribution in different time periods within 24 hours. Heat maps can intuitively show the concentration of load at different times and regions.

[0130] Bar charts and pie charts: used to display the statistical results of load characteristics. For example, a bar chart can be used to display the average load of different months, and a pie chart can be used to display the proportion of different load types. Bar charts and pie charts can help users quickly understand the distribution and composition of the load.

[0131] Among them, choosing the right visualization tool can improve the effect of data display and development efficiency. Commonly used visualization tools include:

[0132] Matplotlib: A commonly used drawing library in Python, suitable for generating various static, dynamic and interactive charts. Matplotlib is powerful and suitable for handling various complex visualization needs.

[0133] D3.js: A JavaScript library for generating dynamic, interactive charts based on data, suitable for web-based presentation. D3.js is highly flexible and can create a variety of complex visualization effects.

[0134] Tableau: A commercial data visualization tool that provides a wide range of chart types and interactive functions, suitable for visualizing complex data. Tableau is easy to use and suitable for quickly creating professional visualization reports.

[0135] In order to realize real-time monitoring and early warning functions, the visualization system includes:

[0136] Real-time data collection and processing module: collects power load data in real time through smart meters and sensors, and performs preprocessing and feature extraction. The real-time collected data can reflect the current status of the power system in a timely manner.

[0137] Real-time data update module: Utilize real-time data stream technology to dynamically update the data in the visualization interface, ensuring that users can view the latest load situation at any time. The real-time data update function ensures that the information obtained by users is always the latest.

[0138] Warning information display module: When the load status indicator exceeds the preset threshold, the system generates warning information in the visual interface to prompt users to take corresponding measures. Warning information can help users discover and deal with potential power system problems in a timely manner.

[0139] In order to improve user experience and data analysis efficiency, the visualization interface can be designed with friendly interactive functions, including:

[0140] Timeline selection: Users can select a specific time period through the timeline to view load data, supporting time views of different granularities such as day, week, month, etc. The timeline selection function can help users flexibly browse and analyze data in different time periods.

[0141] Region switching: Users can switch to view the load situation in different regions, and support comparative analysis of multi-region data. The region switching function can help users understand the load characteristics and differences in different regions.

[0142] Indicator screening: Users can select and screen different situation indicators and view the visualization results of specific indicators as needed. The indicator screening function can help users focus on specific analysis dimensions and improve the pertinence of data analysis.

[0143] The present invention also provides a data mining and visualization system for power load data situation awareness, which is used to execute the above-mentioned data mining and visualization method for power load data situation awareness, wherein the data mining and visualization system for power load data situation awareness includes:

[0144] A power load data collection and preprocessing module, used to collect power load data and preprocess the power load data, wherein the preprocessing includes data cleaning, normalization and outlier processing;

[0145] A multidimensional feature extraction module, used to extract multidimensional features of the preprocessed power load data, wherein the multidimensional features are extracted using statistical analysis and machine learning algorithms;

[0146] A situation indicator construction module, used to construct a comprehensive situation indicator based on the extracted multi-dimensional features, wherein the comprehensive situation indicator is used to reflect the load situation;

[0147] The graphical interface display module is used to intuitively display load status information through a graphical interface.

[0148] Specifically, the data cleaning includes removing duplicate data and filling missing values; normalization is used to unify data of different dimensions to the same scale; and outlier processing is used to identify and process abnormal data points through statistical methods.

[0149] Specifically, the multi-dimensional features of the preprocessed power load data are extracted by combining signal decomposition, statistical analysis and machine learning algorithms.

[0150] Specifically, the comprehensive situation indicators constructed based on the extracted multi-dimensional features include:

[0151] Designing an index system, wherein the indexes in the index system are divided into load characteristic indexes, power quality indexes and load prediction indexes;

[0152] The multi-dimensional features are integrated and synthesized to form a comprehensive situation index that can reflect the power load situation;

[0153] The comprehensive situation indicators of the design are calculated and evaluated to ensure that they can effectively reflect the actual situation and changing trend of the power load.

[0154] Specifically, the graphical interface includes time series graphs, heat maps, bar graphs, and pie charts.

[0155] The data mining visualization method and system for power load data situation awareness provided by the present invention can not only improve the monitoring, analysis and management capabilities of the power system, but also provide important support for the safe and stable operation of the power system. First, through data collection and preprocessing, data quality and consistency can be ensured, providing a reliable basis for subsequent analysis; secondly, in terms of multidimensional feature extraction, statistical analysis and machine learning algorithms are used to deeply mine the complex characteristics of power load data, such as EMD, EEMD, VMD, SVM and neural network methods, so as to provide more comprehensive and accurate situation awareness information; then, a situation indicator system is constructed, which can fully reflect the actual situation and change trend of power load, and provide important reference for the operation and management of the power system; finally, the visualization design intuitively displays the load situation through a variety of technologies such as time series diagrams, heat maps, bar charts and pie charts, and users can dynamically adjust and view data in different time periods and regions to improve analysis and decision-making efficiency. The constructed real-time visualization system can dynamically update the situation indicators, and automatically generate warning information when exceeding the preset threshold, timely discover and handle abnormal situations, and ensure the safe and stable operation of the power system.

[0156] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A data mining visualization method for power load data situation awareness, characterized in that: include: Collecting power load data and preprocessing the power load data, wherein the preprocessing includes data cleaning, normalization and outlier processing; Extracting multidimensional features of the preprocessed power load data, wherein the multidimensional features are extracted using statistical analysis and machine learning algorithms; Constructing a comprehensive situation index based on the extracted multi-dimensional features, wherein the comprehensive situation index is used to reflect the load situation; The load status information is intuitively displayed through a graphical interface.

2. The data mining visualization method for power load data situation awareness according to claim 1 is characterized in that: The data cleaning includes removing duplicate data and filling missing values; Normalization is used to unify data of different dimensions to the same scale; outlier processing is used to identify and process abnormal data points through statistical methods.

3. The data mining visualization method for power load data situation awareness according to claim 1 is characterized in that: The multidimensional features of the preprocessed power load data are extracted by combining signal decomposition, statistical analysis and machine learning algorithms.

4. The data mining visualization method for power load data situation awareness according to claim 1 is characterized in that: The comprehensive situation indicators based on the extracted multi-dimensional features include: Designing an index system, wherein the indexes in the index system are divided into load characteristic indexes, power quality indexes and load prediction indexes; The multi-dimensional features are integrated and synthesized to form a comprehensive situation index that can reflect the power load situation; The comprehensive situation indicators of the design are calculated and evaluated to ensure that they can effectively reflect the actual situation and changing trend of the power load.

5. The data mining visualization method for power load data situation awareness according to any one of claims 1 to 4, characterized in that: Graphical interfaces include time series graphs, heat maps, bar graphs, and pie charts.

6. A data mining and visualization system for power load data situation awareness, characterized in that: include: A power load data collection and preprocessing module, used to collect power load data and preprocess the power load data, wherein the preprocessing includes data cleaning, normalization and outlier processing; A multidimensional feature extraction module, used to extract multidimensional features of the preprocessed power load data, wherein the multidimensional features are extracted using statistical analysis and machine learning algorithms; A situation indicator construction module, used to construct a comprehensive situation indicator based on the extracted multi-dimensional features, wherein the comprehensive situation indicator is used to reflect the load situation; The graphical interface display module is used to intuitively display load status information through a graphical interface.

7. The data mining and visualization system for power load data situation awareness according to claim 6 is characterized in that: The data cleaning includes removing duplicate data and filling missing values; Normalization is used to unify data of different dimensions to the same scale; outlier processing is used to identify and process abnormal data points through statistical methods.

8. The data mining and visualization system for power load data situation awareness according to claim 6 is characterized in that: The multidimensional features of the preprocessed power load data are extracted by combining signal decomposition, statistical analysis and machine learning algorithms.

9. The data mining and visualization system for power load data situation awareness according to claim 6 is characterized in that: The comprehensive situation indicators based on the extracted multi-dimensional features include: Designing an index system, wherein the indexes in the index system are divided into load characteristic indexes, power quality indexes and load prediction indexes; The multi-dimensional features are integrated and synthesized to form a comprehensive situation index that can reflect the power load situation; The comprehensive situation indicators of the design are calculated and evaluated to ensure that they can effectively reflect the actual situation and changing trend of the power load.

10. The data mining and visualization system for power load data situation awareness according to any one of claims 6 to 9, characterized in that: Graphical interfaces include time series graphs, heat maps, bar graphs, and pie charts.

Images