A power data recognition method and system

Abstract

This invention provides a method and system for identifying electricity data. It acquires time-series electricity consumption data from a target enterprise over multiple consecutive analysis periods, transforms the data into a basic electricity consumption feature vector, constructs a three-dimensional tensor based on the historical period's basic electricity consumption feature vector, performs Tucker decomposition, forms an electricity consumption correlation network, and calculates the structural potential time series of each enterprise node. A nonlinear autoregressive model is used to fit the structural potential time series, and the absolute value of the predicted residual is defined as the node's evolutionary activity. During the identification phase, the feature vector of the current period is weighted according to the evolutionary activity, the electricity consumption correlation network is reconstructed, and the node structural potential is calculated. When the average evolutionary activity of the internal nodes of a certain enterprise cluster exceeds a first threshold, and the average structural potential of the cluster shows a monotonically increasing trend over the most recent N periods with a total increase exceeding a second threshold, the enterprise set corresponding to the cluster is identified as an advanced enterprise unit.

Description

Technical Field

[0001] This application belongs to the field of data processing, and in particular relates to a method and system for identifying power data. Background Technology

[0002] Technologically advanced enterprises represent contemporary advanced productivity spurred by revolutionary technological breakthroughs, innovative allocation of production factors, and profound industrial transformation and upgrading. Traditional methods for identifying high-growth enterprises primarily rely on macroeconomic indicators, corporate financial reports, industry research and analysis, or patent application data. However, financial reports and annual reports have long release cycles and are subject to delays, making it difficult to detect early signals of industrial development. Qualitative analysis methods based on expert interviews or industry reports are highly subjective and have limited coverage, hindering large-scale assessments. With the development of the energy Internet of Things and smart grid technologies, high-frequency, high-dimensional enterprise electricity consumption data offers new possibilities for real-time and objective insights into economic operations. Electricity data, as a "barometer" of enterprise production and operation activities, possesses unique advantages such as strong real-time performance, objectivity, and broad coverage. Some studies have already attempted to utilize electricity data for economic analysis, such as predicting macroeconomic conditions by analyzing changes in total electricity consumption or assessing operational status by analyzing the electricity consumption patterns of individual enterprises. Most existing studies have neglected the correlation between enterprises’ electricity consumption behaviors, making it difficult to construct a correlation model that can reflect the collaborative relationship of the industrial chain from the perspective of network evolution. Furthermore, there is a lack of a useful evaluation system to comprehensively represent the structural importance and evolutionary activity of enterprises or enterprise clusters in the industrial network, making it difficult to accurately identify advanced enterprise units in the nascent or rapidly growing stages. Summary of the Invention

[0003] This invention proposes a method for identifying power data, comprising the following steps:

[0004] Acquire time-series electricity consumption data of multiple target enterprises over multiple consecutive analysis periods, and transform the time-series electricity consumption data of each enterprise in each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate and instantaneous frequency deviation.

[0005] For each historical analysis period, a three-dimensional tensor is constructed based on the basic electricity consumption feature vector and Tucker decomposition is performed to construct an electricity consumption correlation network and calculate the structural potential of each enterprise node in the network, thereby obtaining the time series of the structural potential of each enterprise node.

[0006] Based on the time series of the structural potential of each enterprise node, a nonlinear autoregressive model is used for fitting, and the absolute value of the prediction residual of the model to the structural potential of the most recent historical period is used as the evolutionary activity of the node; according to the evolutionary activity, the basic electricity consumption feature vector of the current analysis period is weighted, and the same Tucker decomposition method as the historical period is used to construct the current electricity consumption correlation network, and the structural potential of each enterprise node in the current network is calculated.

[0007] In the current electricity consumption network, when the average evolutionary activity of the internal nodes of a certain enterprise cluster exceeds a first threshold, and the average structural potential of the cluster is monotonically increasing over the most recent N periods including the current period and the total increase exceeds a second threshold, then the set of enterprises corresponding to the cluster is identified as an advanced enterprise unit.

[0008] Optionally, the step of converting the time-series electricity consumption data of each enterprise within each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate, and instantaneous frequency deviation includes:

[0009] Perform a fast Fourier transform on the current and voltage time-series data within the period to obtain the power spectral density function P(f); normalize the power spectral density to a probability distribution p(f) and calculate the power entropy;

[0010] Based on the Fourier transform results, the amplitude of the fundamental component is extracted. and the amplitude of each harmonic component Calculate the harmonic distortion rate for i = 2, 3, ..., k. ;

[0011] The instantaneous frequency sequence is extracted using a third-order generalized integrator phase-locked loop, and the standard deviation of the instantaneous frequency sequence within one period is calculated as the instantaneous frequency deviation.

[0012] Optionally, the step of constructing a three-dimensional tensor based on the basic electricity consumption feature vector and performing Tucker decomposition for each historical analysis period to build an electricity consumption correlation network includes:

[0013] The K basic electricity consumption characteristics (K=3) of M enterprises within L consecutive historical analysis periods constitute a... 3D tensor ;

[0014] tensor Perform Tucker decomposition to obtain the core tensor G and Enterprise factor matrix A;

[0015] Based on the enterprise factor matrix Constructing the adjacency matrix of the electricity consumption association network The connection weights between firm u and firm v By calculating the matrix The corresponding row vector and row vector The cosine similarity is obtained.

[0016] Optionally, the structural potential of each enterprise node in the computing network includes:

[0017] The structural potential of an enterprise node is defined as its PageRank value in an electricity consumption network.

[0018] The PageRank value of each node is calculated using an iterative algorithm. For a weighted undirected network constructed from cosine similarity, the iterative formula is:

[0019]

[0020] Where PR(u) is the PageRank value of node u, d is the damping coefficient, N is the total number of enterprise nodes, and N(u) is the set of neighboring nodes of node u. It is the weight of the edge connecting nodes u and v. It is the sum of the weights of all edges connected to node v.

[0021] Optionally, the structural potential time series based on each enterprise node is fitted using a nonlinear autoregressive model, including:

[0022] A nonlinear autoregressive neural network model is adopted. The structure of the nonlinear autoregressive neural network model is as follows: the input layer contains the structural potential data of the first 4 historical cycles as input, a hidden layer containing 10 neurons, and an output layer.

[0023] The activation function for the hidden layer is the hyperbolic tangent function, and the activation function for the output layer is a linear function.

[0024] Train the model until the mean squared error is less than 100%. .

[0025] Optionally, the weighting of the basic electricity consumption feature vector for the current analysis period based on the evolutionary activity includes:

[0026] For enterprise node i, the evolutionary activity is The basic electricity consumption feature vector for the current analysis period is: ;

[0027] Evolutionary activity As a scalar multiplier, for the eigenvector We perform weighting to obtain the weighted feature vector. .

[0028] Optionally, identifying the set of enterprises corresponding to the cluster as advanced enterprise units includes:

[0029] A community detection algorithm is used to divide the current electricity consumption network, resulting in multiple enterprise clusters;

[0030] For any cluster C, calculate the average evolutionary activity of all nodes within it. ,like Exceeding the first threshold;

[0031] Furthermore, the average structural potential time series of the cluster C over the most recent four periods (N=4) including the current period is calculated. This verifies that the sequence is strictly monotonically increasing;

[0032] Furthermore, calculate the total increase in the average structural potential of the current cycle relative to the four cycles prior, if the following conditions are met... Then the cluster is identified as an advanced enterprise unit.

[0033] Furthermore, the present invention also relates to an electricity data identification system, comprising the following modules:

[0034] The conversion module is used to acquire time-series electricity consumption data of multiple target enterprises over multiple consecutive analysis periods, and convert the time-series electricity consumption data of each enterprise in each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate and instantaneous frequency deviation.

[0035] The module is used to construct a three-dimensional tensor based on the basic electricity consumption feature vector for each historical analysis period and perform Tucker decomposition to construct an electricity consumption correlation network and calculate the structural potential of each enterprise node in the network, thereby obtaining the time series of the structural potential of each enterprise node.

[0036] The calculation module is used to fit the time series of the structural potential of each enterprise node using a nonlinear autoregressive model, and to take the absolute value of the prediction residual of the model to the structural potential of the most recent historical period as the evolutionary activity of the node; to weight the basic electricity consumption feature vector of the current analysis period according to the evolutionary activity, to construct the current electricity consumption correlation network using the same Tucker decomposition method as the historical period, and to calculate the structural potential of each enterprise node in the current network.

[0037] The identification module is used to identify the set of enterprises corresponding to a certain enterprise cluster as an advanced enterprise unit when the average evolutionary activity of the internal nodes of a certain enterprise cluster exceeds a first threshold and the average structural potential of the cluster is monotonically increasing in the most recent N periods including the current period and the total increase exceeds a second threshold in the current power consumption association network.

[0038] Preferably, the step of converting the time-series electricity consumption data of each enterprise within each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate, and instantaneous frequency deviation includes:

[0039] Perform a fast Fourier transform on the current and voltage time-series data within the period to obtain the power spectral density function P(f); normalize the power spectral density to a probability distribution p(f) and calculate the power entropy;

[0040] Based on the Fourier transform results, the amplitude of the fundamental component is extracted. and the amplitude of each harmonic component Calculate the harmonic distortion rate for i = 2, 3, ..., k. ;

[0041] The instantaneous frequency sequence is extracted using a third-order generalized integrator phase-locked loop, and the standard deviation of the instantaneous frequency sequence within one period is calculated as the instantaneous frequency deviation.

[0042] Preferably, the step of constructing a three-dimensional tensor based on the basic electricity consumption feature vector and performing Tucker decomposition for each historical analysis period to build an electricity consumption correlation network includes:

[0043] The K basic electricity consumption characteristics (K=3) of M enterprises within L consecutive historical analysis periods are used to construct a three-dimensional tensor of M×L×K. ;

[0044] tensor Perform Tucker decomposition to obtain the core tensor G and Enterprise factor matrix A;

[0045] Construct an adjacency matrix W for the electricity consumption association network based on the enterprise factor matrix A, where the connection weights between enterprises u and v are... By calculating the corresponding row vectors in matrix A and row vector The cosine similarity is obtained.

[0046] Preferably, the structural potential of each enterprise node in the computing network includes:

[0047] The structural potential of an enterprise node is defined as its PageRank value in an electricity consumption network.

[0048] The PageRank value of each node is calculated using an iterative algorithm. For a weighted undirected network constructed from cosine similarity, the iterative formula is:

[0049]

[0050] Where PR(u) is the PageRank value of node u, d is the damping coefficient, N is the total number of enterprise nodes, and N(u) is the set of neighboring nodes of node u. It is the weight of the edge connecting nodes u and v. It is the sum of the weights of all edges connected to node v.

[0051] Preferably, the structural potential time series based on each enterprise node is fitted using a nonlinear autoregressive model, including:

[0052] A nonlinear autoregressive neural network model is adopted. The structure of the nonlinear autoregressive neural network model is as follows: the input layer contains the structural potential data of the first 4 historical cycles as input, a hidden layer containing 10 neurons, and an output layer.

[0053] The activation function for the hidden layer is the hyperbolic tangent function, and the activation function for the output layer is a linear function.

[0054] Train the model until the mean squared error is less than 10. -5 .

[0055] Preferably, the weighting of the basic electricity consumption feature vector for the current analysis period based on the evolutionary activity includes:

[0056] For enterprise node i, the evolutionary activity is The basic electricity consumption feature vector for the current analysis period is: ;

[0057] Evolutionary activity As a scalar multiplier, for the eigenvector We perform weighting to obtain the weighted feature vector. .

[0058] Preferably, identifying the set of enterprises corresponding to the cluster as advanced enterprise units includes:

[0059] A community detection algorithm is used to divide the current electricity consumption network, resulting in multiple enterprise clusters;

[0060] For any cluster C, calculate the average evolutionary activity of all nodes within it. ,like Exceeding the first threshold;

[0061] Furthermore, the average structural potential time series of the cluster C over the most recent four periods (N=4) including the current period is calculated. This verifies that the sequence is strictly monotonically increasing;

[0062] Furthermore, calculate the total increase in the average structural potential of the current cycle relative to the four cycles prior, if the following conditions are met... Then the cluster is identified as an advanced enterprise unit.

[0063] This invention utilizes high-frequency, multi-dimensional power data to analyze not only electricity consumption but also power quality data such as Taiwan's power entropy and harmonic distortion rate, which reflect the sophistication of enterprise production equipment and the stability of processes. By constructing an enterprise power consumption correlation network, it elevates the analysis of isolated enterprises to the level of the industrial ecosystem. This invention proposes two core evaluation indicators: structural potential and evolutionary activity. The former represents the importance of an enterprise in the network, while the latter, through nonlinear model prediction of residuals, detects the abruptness and growth potential of enterprise behavior, enabling earlier identification of advanced enterprise units in enterprise clusters with continuously rising influence and in a rapid growth phase. Attached Figure Description

[0064] Figure 1 A flowchart of the first embodiment;

[0065] Figure 2 A schematic diagram of the basic electricity consumption characteristic vector;

[0066] Figure 3 This is a schematic diagram of an electricity-related network. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application. It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with relevant laws, regulations, and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0068] In the first embodiment, the present invention proposes a method for identifying power data, such as... Figure 1 This includes the following steps:

[0069] S1, acquire time-series electricity consumption data of multiple target enterprises in multiple consecutive analysis periods, and convert the time-series electricity consumption data of each enterprise in each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate and instantaneous frequency deviation.

[0070] Electricity consumption data was collected for 12 consecutive months, with each month defined as an analysis period. The active power sequence of each enterprise within a single analysis period was divided into multiple intervals. The frequency of power values ​​falling into each interval was statistically analyzed, and power entropy was calculated based on the information entropy formula. Power entropy represents the complexity and stability of the enterprise's electricity consumption pattern. Voltage and current waveform data were collected, and the amplitude of each harmonic component was calculated using Fourier transform. The harmonic distortion rate was obtained according to the formula for total harmonic distortion, reflecting the usage of nonlinear equipment and the advancement of production processes. The instantaneous frequency of the power grid was recorded, and the root mean square error of the instantaneous frequency relative to a standard industrial frequency (e.g., 50Hz) within an analysis period was calculated to obtain the instantaneous frequency deviation, reflecting the impact of the start-up and shutdown of high-power equipment on the power grid. Each enterprise, in each analysis period, has a three-dimensional basic electricity consumption feature vector composed of the above three indicators, such as... Figure 2 .

[0071] S2, For each historical analysis period, a three-dimensional tensor is constructed based on the basic electricity consumption feature vector and Tucker decomposition is performed to construct an electricity consumption correlation network and calculate the structural potential of each enterprise node in the network, thereby obtaining the time series of the structural potential of each enterprise node.

[0072] For any given historical analysis period, the feature similarity between all enterprise pairs is calculated. For example, the similarity between enterprise i and enterprise j in terms of power entropy can be represented by the reciprocal of the difference between their power entropy values. Using enterprises, enterprise characteristics, and electricity consumption characteristics as three dimensions, an N×N×3 three-dimensional tensor is constructed, where N is the total number of enterprises, 3 is the number of feature dimensions, and each element in the tensor represents the correlation strength between two enterprises on a specific feature. A Tucker decomposition is performed on this tensor, yielding a core tensor and three factor matrices. Each row of the factor matrix corresponding to the enterprise dimension can be considered as the coordinate vector of that enterprise in the potential industry association space. By calculating the inner product of the coordinate vectors of any two enterprises, an N×N electricity consumption association network adjacency matrix can be constructed, where the values ​​in the matrix represent the degree of association between enterprises, such as... Figure 3 As shown, the importance score of each enterprise node in the network is calculated using a webpage ranking algorithm or an eigenvector centrality algorithm, and this score is used as the structural potential of the enterprise in that period. Repeating the above process for all historical analysis periods yields a structural potential time series for each enterprise.

[0073] S3. Based on the time series of the structural potential of each enterprise node, a nonlinear autoregressive model is used for fitting, and the absolute value of the prediction residual of the model to the structural potential of the most recent historical period is used as the evolutionary activity of the node; according to the evolutionary activity, the basic electricity consumption feature vector of the current analysis period is weighted, and the same Tucker decomposition method as the historical period is used to construct the current electricity consumption correlation network, and the structural potential of each enterprise node in the current network is calculated.

[0074] For each enterprise, the structural potential time series of the past 11 historical cycles is extracted as training data. A multilayer perceptron containing an input layer, hidden layer, and output layer is selected as the implementation method for the nonlinear autoregressive model. The input of the model is the structural potential of the previous few cycles, such as the previous 3 cycles, and the output is the prediction of the structural potential of the next cycle. The model is trained using historical data until it converges. Using the trained model, the structural potential data of the 9th, 10th, and 11th cycles are input to predict the structural potential of the 12th cycle, i.e., the most recent historical cycle. The difference between the predicted value and the actual structural potential of the 12th cycle is calculated to obtain the prediction residual. The absolute value of the residual is defined as the evolutionary activity of the enterprise node. The magnitude of the obtained value reflects the drastic change or unpredictability of the enterprise's recent position in the industrial network. Entering the current analysis cycle, i.e., the 13th month, the three-dimensional basic electricity consumption feature vector of each enterprise in the current cycle is calculated using the above method. The evolutionary activity of each enterprise is used as a weight coefficient and multiplied by the basic electricity consumption feature vector of the enterprise in the current cycle to obtain the weighted feature vector. Firms with higher evolutionary activity have larger eigenvector magnitudes, indicating amplified influence in constructing the current network. Based on the weighted eigenvectors of all firms, the entire process of tensor construction, Tucker decomposition, interconnection network construction, and structural potential calculation is repeated to obtain the latest structural potential of each firm node in the current analysis period.

[0075] S4. In the current power consumption network, when the average evolutionary activity of the internal nodes of a certain enterprise cluster exceeds the first threshold, and the average structural potential of the cluster is monotonically increasing in the most recent N periods including the current period and the total increase exceeds the second threshold, then the set of enterprises corresponding to the cluster is identified as an advanced enterprise unit.

[0076] Community detection algorithms, such as the Louvain algorithm, are run on the current electricity consumption network to group closely related enterprise nodes into different enterprise clusters. For each identified cluster, the arithmetic mean of the evolutionary activity of all enterprise nodes within it is calculated. If this mean is greater than a preset first threshold, such as the top 10% after ranking the evolutionary activity of all enterprises, the cluster passes the activity test. For clusters that pass the activity test, the average structural potential over the most recent N periods, such as the most recent 4 periods, is calculated, which is the average structural potential of all members of the cluster in each period. This 4-period average structural potential sequence is checked to see if it is strictly monotonically increasing, and the total increase of the last period relative to the first period is calculated to see if it exceeds a preset second threshold, such as 30%. When an enterprise cluster simultaneously satisfies both the conditions of high average evolutionary activity and continuous growth in average structural potential, all enterprises within the cluster constitute an identified advanced enterprise unit.

[0077] In an optional embodiment, the step of converting the time-series electricity consumption data of each enterprise within each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate, and instantaneous frequency deviation includes:

[0078] Perform a fast Fourier transform on the current and voltage time-series data within the period to obtain the power spectral density function P(f); normalize the power spectral density to a probability distribution p(f) and calculate the power entropy;

[0079] Based on the Fourier transform results, the amplitude of the fundamental component is extracted. and the amplitude of each harmonic component Calculate the harmonic distortion rate for i = 2, 3, ..., k. ;

[0080] The instantaneous frequency sequence is extracted using a third-order generalized integrator phase-locked loop, and the standard deviation of the instantaneous frequency sequence within one period is calculated as the instantaneous frequency deviation.

[0081] Taking a company's data within a one-hour analysis period as an example, current and voltage time-series data are collected at a frequency of 5kHz within this period. A Fast Fourier Transform (FFT) is performed on these two sets of time-series data to obtain the power spectral density function, which represents how signal power is distributed with frequency. This power spectral density function is then normalized to make it a probability distribution, for example, by calculating the power proportions at different frequency points as 0.2, 0.15, etc., and further utilizing the information entropy formula... The power entropy value is calculated, assuming it to be 2.1. Using the Fourier transform, the voltage amplitude at the fundamental frequency (50Hz), for example, 220 volts, is extracted. The voltage amplitudes of the second, third, and even fiftieth harmonics are also extracted, for example, 5 volts, 3 volts, etc. According to the formula for calculating the harmonic distortion rate, the square root of the sum of the squares of each harmonic amplitude is divided by the fundamental amplitude to obtain a ratio value, for example, 0.04. The voltage signal is processed using a third-order generalized integrator phase-locked loop (PLL) technique to obtain an instantaneous frequency sequence fluctuating around 50Hz, and the standard deviation of this sequence is calculated, for example, 0.02Hz. The company's basic electricity consumption characteristic vector for this period then constitutes a vector containing three values: 2.1, 0.04, and 0.02.

[0082] In an optional embodiment, the step of constructing a three-dimensional tensor based on the basic electricity consumption feature vector and performing Tucker decomposition for each historical analysis period to build an electricity consumption correlation network includes:

[0083] The K basic electricity consumption characteristics (K=3) of M enterprises within L consecutive historical analysis periods are used to construct a three-dimensional tensor of M×L×K. ;

[0084] tensor Perform Tucker decomposition to obtain the core tensor G and Enterprise factor matrix A;

[0085] Construct an adjacency matrix W for the electricity consumption association network based on the enterprise factor matrix A, where the connection weights between enterprises u and v are... By calculating the corresponding row vectors in matrix A and row vector The cosine similarity is obtained.

[0086] Suppose we analyze the electricity consumption of 100 enterprises over the past 12 months. Each enterprise has a three-dimensional feature vector for each month, consisting of power entropy, harmonic distortion rate, and instantaneous frequency deviation. This forms a data tensor with dimensions of 100 enterprises × 12 periods × 3 features. This tensor records all electricity consumption characteristics of all enterprises at all time points. We perform Tucker decomposition on this three-dimensional tensor, setting the rank of the enterprise dimension to 10, the rank of the time dimension to 4, and the rank of the feature dimension to 2. The decomposition yields a core tensor and three factor matrices, in the form of... This requires a 100×10 enterprise factor matrix A. Each row of the matrix represents a ten-dimensional latent feature of the corresponding enterprise, including its relationships with other enterprises, time, and electricity consumption characteristics. To construct the electricity consumption association network among enterprises, the cosine similarity between any two row vectors in enterprise factor matrix A is calculated. For example, if the cosine similarity between the fifth and eighth row vectors is 0.85, then the connection weight between enterprise five and enterprise eight in the network's adjacency matrix is ​​0.85. By calculating the cosine similarity pairwise for all enterprises, a 100×100 weighted adjacency matrix is ​​obtained, which represents the electricity consumption association network among enterprises.

[0087] In an optional embodiment, the structural potential of each enterprise node in the computing network includes:

[0088] The structural potential of an enterprise node is defined as its PageRank value in an electricity consumption network.

[0089] The PageRank value of each node is calculated using an iterative algorithm. For a weighted undirected network constructed from cosine similarity, the iterative formula is:

[0090]

[0091] Where PR(u) is the PageRank value of node u, d is the damping coefficient, N is the total number of enterprise nodes, and N(u) is the set of neighboring nodes of node u. It is the weight of the edge connecting nodes u and v. It is the sum of the weights of all edges connected to node v.

[0092] In a power consumption network containing 100 enterprise nodes, each enterprise node is assigned an initial PageRank value, preferably the reciprocal of the total number of nodes, i.e., 1%. The damping coefficient is set to 0.85. In the first iteration, for enterprise node A, it is necessary to consider all its directly connected neighboring nodes, such as B and C. The current PageRank values ​​of B and C are known, both at 1% in the first iteration. The weighted influence of the flow from B and C to A needs to be calculated. This is obtained by dividing the edge weight connecting A and B (e.g., 0.9) by the sum of the weights of all edges connected to B (e.g., 3.5), and then multiplying by B's PageRank value. The same calculation is performed for C. The weighted influence of all neighboring nodes flowing to A is summed, multiplied by the damping coefficient 0.85, and then added to a base value, 0.15 / 100, to obtain A's new PageRank value after the first iteration. This process is performed simultaneously on all one hundred nodes in the network, and the entire iteration process is repeated until the PageRank values ​​of all nodes stabilize and no longer show significant changes. The obtained stable values ​​yield structural potential scores for each enterprise node.

[0093] In an optional embodiment, the fitting of the structural potential time series based on each enterprise node using a nonlinear autoregressive model includes:

[0094] A nonlinear autoregressive neural network model is adopted. The structure of the nonlinear autoregressive neural network model is as follows: the input layer contains the structural potential data of the first 4 historical cycles as input, a hidden layer containing 10 neurons, and an output layer.

[0095] The activation function for the hidden layer is the hyperbolic tangent function, and the activation function for the output layer is a linear function.

[0096] Train the model until the mean squared error is less than 10. -5 .

[0097] Taking a specific company as an example, its structural potential score time series for the past 24 consecutive analysis periods was obtained, such as 0.021, 0.023, etc. To train a nonlinear autoregressive model to predict future trends, the series was transformed into a series of input-output pairs. The input to the first training sample was the structural potential score for the first to fourth periods, and the output was the score for the fifth period; the input to the second sample was the score for the second to fifth periods, and the output was the score for the sixth period, and so on, generating a total of 20 training samples.

[0098] The training samples are input into a pre-defined neural network. This network has four input nodes, each receiving structural potential data from the previous four cycles; a hidden layer containing ten neurons; and an output node for predicting the structural potential of the next cycle. During training, an optimization algorithm continuously adjusts the weight parameters within the network, aiming to minimize the mean squared error between the network's predicted output and the actual output. When the error is successfully reduced to less than 10... -5 Training stops when the time is up. After training is complete, the model can be used to predict the structural potential of the next cycle.

[0099] In an optional embodiment, the weighting of the basic electricity consumption feature vector for the current analysis period based on the evolutionary activity includes:

[0100] For enterprise node i, the evolutionary activity is The basic electricity consumption feature vector for the current analysis period is: ;

[0101] Evolutionary activity As a scalar multiplier, for the eigenvector We perform weighting to obtain the weighted feature vector. .

[0102] Specifically, for Company A, in the current analysis period, the structural potential data from the past four periods is input into the nonlinear autoregressive model trained in the previous stage to predict the structural potential of the current period. Then, the predicted value is compared with the actual calculated structural potential of the current period, and the difference between the two is defined as the evolutionary activity of the company. Assume that the calculated evolutionary activity of Company A is 1.3. The basic electricity consumption characteristic vector of Company A in the current period is calculated normally, assuming a power entropy of 2.0, a harmonic distortion rate of 0.05, and an instantaneous frequency deviation of 0.015, resulting in a vector of 2.0, 0.05, and 0.015. The scalar value of the evolutionary activity, i.e., 1.3, is multiplied by each component of the three-dimensional characteristic vector. The resulting new weighted characteristic vector is 2.6, 0.065, and 0.0195. This new vector not only reflects the current basic electricity consumption state of Company A but also incorporates the changing trend of the structural potential, making the characteristics more timely.

[0103] In an optional embodiment, identifying the set of enterprises corresponding to the cluster as advanced enterprise units includes:

[0104] A community detection algorithm is used to divide the current electricity consumption network, resulting in multiple enterprise clusters;

[0105] For any cluster C, calculate the average evolutionary activity of all nodes within it. ,like Exceeding the first threshold of 1.5;

[0106] Furthermore, the average structural potential time series of the cluster C over the most recent four periods (N=4) including the current period is calculated. This verifies that the sequence is strictly monotonically increasing;

[0107] Furthermore, calculate the total increase in the average structural potential of the current cycle relative to the four cycles prior, if the following conditions are met... Then the cluster is identified as an advanced enterprise unit.

[0108] Specifically, a community detection algorithm is used to process the electricity consumption network for the current period, automatically dividing the enterprise nodes in the network into several closely connected clusters. One cluster containing 10 enterprises is selected for analysis. The evolutionary activity of each of these 10 enterprises is calculated, and the average value is taken. If the calculated average evolutionary activity is 1.6, and this value is higher than a preset first threshold of 1.5, then the first condition is met.

[0109] The cluster's performance over the most recent four periods was traced. The average structural potential of the 10 firms in each period was calculated, resulting in a time series of length four, with values ​​of, for example, 0.030, 0.035, 0.042, and 0.048. Since each subsequent value is greater than the previous one, the series is strictly monotonically increasing, satisfying the second condition. The overall growth rate of the structural potential over the four periods was calculated, yielding a result of 0.6. This growth rate exceeds the set second threshold of 0.5, thus satisfying the third condition. Because the cluster passed all three conditions, the set of 10 firms was identified as an advanced firm unit.

[0110] In a second embodiment, a power data identification system is provided, comprising the following modules:

[0111] The conversion module is used to acquire time-series electricity consumption data of multiple target enterprises over multiple consecutive analysis periods, and convert the time-series electricity consumption data of each enterprise in each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate and instantaneous frequency deviation.

[0112] The module is used to construct a three-dimensional tensor based on the basic electricity consumption feature vector for each historical analysis period and perform Tucker decomposition to construct an electricity consumption correlation network and calculate the structural potential of each enterprise node in the network, thereby obtaining the time series of the structural potential of each enterprise node.

[0113] The calculation module is used to fit the time series of the structural potential of each enterprise node using a nonlinear autoregressive model, and to take the absolute value of the prediction residual of the model to the structural potential of the most recent historical period as the evolutionary activity of the node; to weight the basic electricity consumption feature vector of the current analysis period according to the evolutionary activity, to construct the current electricity consumption correlation network using the same Tucker decomposition method as the historical period, and to calculate the structural potential of each enterprise node in the current network.

[0114] The identification module is used to identify the set of enterprises corresponding to a certain enterprise cluster as an advanced enterprise unit when the average evolutionary activity of the internal nodes of a certain enterprise cluster exceeds a first threshold and the average structural potential of the cluster is monotonically increasing in the most recent N periods including the current period and the total increase exceeds a second threshold in the current power consumption association network.

[0115] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0116] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A method for identifying power data, characterized in that, Includes the following steps: Acquire time-series electricity consumption data of multiple target enterprises over multiple consecutive analysis periods, and transform the time-series electricity consumption data of each enterprise in each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate and instantaneous frequency deviation. For each historical analysis period, a three-dimensional tensor is constructed based on the basic electricity consumption feature vector and Tucker decomposition is performed to construct an electricity consumption correlation network and calculate the structural potential of each enterprise node in the network, thereby obtaining the time series of the structural potential of each enterprise node. Based on the time series of the structural potential of each enterprise node, a nonlinear autoregressive model is used for fitting, and the absolute value of the prediction residual of the model to the structural potential of the most recent historical period is used as the evolutionary activity of the node; according to the evolutionary activity, the basic electricity consumption feature vector of the current analysis period is weighted, and the same Tucker decomposition method as the historical period is used to construct the current electricity consumption correlation network, and the structural potential of each enterprise node in the current network is calculated. In the current electricity consumption network, when the average evolutionary activity of the internal nodes of a certain enterprise cluster exceeds the first threshold, and the average structural potential of the cluster is monotonically increasing in the most recent N periods including the current period and the total increase exceeds the second threshold, then the set of enterprises corresponding to the cluster is identified as an advanced enterprise unit. The structural potential of each enterprise node in the computing network includes: The structural potential of an enterprise node is defined as its PageRank value in an electricity consumption network. The PageRank value of each node is calculated using an iterative algorithm. For a weighted undirected network constructed from cosine similarity, the iterative formula is: Where PR(u) is the PageRank value of node u, d is the damping coefficient, N is the total number of enterprise nodes, and N(u) is the set of neighboring nodes of node u. It is the weight of the edge connecting nodes u and v. It is the sum of the weights of all edges connected to node v; The weighting of the basic electricity consumption feature vector for the current analysis period based on the evolutionary activity includes: For enterprise node i, the evolutionary activity is The basic electricity consumption feature vector for the current analysis period is: ; Evolutionary activity As a scalar multiplier, for the eigenvector We perform weighting to obtain the weighted feature vector. .

2. The method according to claim 1, characterized in that, The process of converting the time-series electricity consumption data of each enterprise within each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate, and instantaneous frequency deviation includes: Perform a fast Fourier transform on the current and voltage time-series data within the period to obtain the power spectral density function P(f); normalize the power spectral density to a probability distribution p(f) and calculate the power entropy; Based on the Fourier transform results, the amplitude of the fundamental component is extracted. and the amplitude of each harmonic component Calculate the harmonic distortion rate for i = 2, 3, ..., k. ; The instantaneous frequency sequence is extracted using a third-order generalized integrator phase-locked loop, and the standard deviation of the instantaneous frequency sequence within one period is calculated as the instantaneous frequency deviation.

3. The method according to claim 1, characterized in that, For each historical analysis period, a three-dimensional tensor is constructed based on the basic electricity consumption feature vector, and Tucker decomposition is performed to build an electricity consumption correlation network, including: The K basic electricity consumption characteristics (K=3) of M enterprises within L consecutive historical analysis periods are used to construct a three-dimensional tensor of M×L×K. ; tensor Perform Tucker decomposition to obtain the core tensor G and Enterprise factor matrix A; Construct an adjacency matrix W for the electricity consumption association network based on the enterprise factor matrix A, where the connection weights between enterprises u and v are... By calculating the corresponding row vectors in matrix A and row vector The cosine similarity is obtained.

4. The method according to claim 1, characterized in that, The structural potential time series based on each enterprise node is fitted using a nonlinear autoregressive model, including: A nonlinear autoregressive neural network model is adopted. The structure of the nonlinear autoregressive neural network model is as follows: the input layer contains the structural potential data of the first 4 historical cycles as input, a hidden layer containing 10 neurons, and an output layer. The activation function for the hidden layer is the hyperbolic tangent function, and the activation function for the output layer is a linear function. Train the model until the mean squared error is less than 10. -5 .

5. The method according to claim 1, characterized in that, The step of identifying the set of enterprises corresponding to the cluster as advanced enterprise units includes: A community detection algorithm is used to divide the current electricity consumption network, resulting in multiple enterprise clusters; For any cluster C, calculate the average evolutionary activity of all nodes within it. ,like Exceeding the first threshold; Furthermore, the average structural potential time series of the cluster C over the most recent four periods (N=4) including the current period is calculated. This verifies that the sequence is strictly monotonically increasing; Furthermore, calculate the total increase in the average structural potential of the current cycle relative to the four cycles prior, if the following conditions are met... Then the cluster is identified as an advanced enterprise unit.

6. A power data identification system, characterized in that, Includes the following modules: The conversion module is used to acquire time-series electricity consumption data of multiple target enterprises over multiple consecutive analysis periods, and convert the time-series electricity consumption data of each enterprise in each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate and instantaneous frequency deviation. The module is used to construct a three-dimensional tensor based on the basic electricity consumption feature vector for each historical analysis period and perform Tucker decomposition to construct an electricity consumption correlation network and calculate the structural potential of each enterprise node in the network, thereby obtaining the time series of the structural potential of each enterprise node. The calculation module is used to fit the time series of the structural potential of each enterprise node using a nonlinear autoregressive model, and to take the absolute value of the prediction residual of the model to the structural potential of the most recent historical period as the evolutionary activity of the node; to weight the basic electricity consumption feature vector of the current analysis period according to the evolutionary activity, to construct the current electricity consumption correlation network using the same Tucker decomposition method as the historical period, and to calculate the structural potential of each enterprise node in the current network. The identification module is used to identify the set of enterprises corresponding to a certain enterprise cluster as an advanced enterprise unit when the average evolutionary activity of the internal nodes of a certain enterprise cluster exceeds a first threshold and the average structural potential of the cluster is monotonically increasing in the most recent N periods including the current period and the total increase exceeds a second threshold in the current power consumption association network. The structural potential of each enterprise node in the computing network includes: The structural potential of an enterprise node is defined as its PageRank value in an electricity consumption network. The PageRank value of each node is calculated using an iterative algorithm. For a weighted undirected network constructed from cosine similarity, the iterative formula is: Where PR(u) is the PageRank value of node u, d is the damping coefficient, N is the total number of enterprise nodes, and N(u) is the set of neighboring nodes of node u. It is the weight of the edge connecting nodes u and v. It is the sum of the weights of all edges connected to node v; The weighting of the basic electricity consumption feature vector for the current analysis period based on the evolutionary activity includes: For enterprise node i, the evolutionary activity is The basic electricity consumption feature vector for the current analysis period is: ; Evolutionary activity As a scalar multiplier, for the eigenvector We perform weighting to obtain the weighted feature vector. .

7. The system according to claim 6, characterized in that, The process of converting the time-series electricity consumption data of each enterprise within each period into a basic electricity consumption feature vector containing power entropy, harmonic distortion rate, and instantaneous frequency deviation includes: Perform a fast Fourier transform on the current and voltage time-series data within the period to obtain the power spectral density function P(f); normalize the power spectral density to a probability distribution p(f) and calculate the power entropy; Based on the Fourier transform results, the amplitude of the fundamental component is extracted. and the amplitude of each harmonic component Calculate the harmonic distortion rate for i = 2, 3, ..., k. ; The instantaneous frequency sequence is extracted using a third-order generalized integrator phase-locked loop, and the standard deviation of the instantaneous frequency sequence within one period is calculated as the instantaneous frequency deviation.

8. The system according to claim 6, characterized in that, For each historical analysis period, a three-dimensional tensor is constructed based on the basic electricity consumption feature vector, and Tucker decomposition is performed to build an electricity consumption correlation network, including: The K basic electricity consumption characteristics (K=3) of M enterprises within L consecutive historical analysis periods are used to construct a three-dimensional tensor of M×L×K. ; tensor Perform Tucker decomposition to obtain the core tensor G and Enterprise factor matrix A; Construct an adjacency matrix W for the electricity consumption association network based on the enterprise factor matrix A, where the connection weights between enterprises u and v are... By calculating the corresponding row vectors in matrix A and row vector The cosine similarity is obtained.

Images