User electricity consumption data analysis modeling method and system based on semi-supervised learning

Abstract

The invention provides a user electricity consumption data analysis modeling method and system based on semi-supervised learning, and belongs to the technical field of power system data analysis and prediction. The problems of low data inspection accuracy, large characteristic correlation degree deviation and the like in the prior art are solved. The method comprises the following steps: on the basis of user electricity consumption characteristic data and climate characteristic data, constructing a semi-supervised learning driven electricity consumption data inspection and tracking feedback model by adopting a graph convolutional network; calculating the Pearson's correlation coefficient among the electricity consumption characteristics of different types of users, constructing a characteristic correlation model based on a deep belief network, and analyzing the nonlinear relationship among the user types, the user electricity consumption time sequence characteristics and the climate characteristics; based on a grey correlation analysis method, capturing distribution characteristics of electricity consumption changes; the method is suitable for power system resource optimization, and especially improves the operation efficiency of a power grid under climate variability.

Description

Technical Field

[0001] The present application relates to the technical field of power system data analysis and prediction, and in particular to a user electricity consumption data analysis modeling method and system based on semi-supervised learning. Background Art

[0002] In the power system, the analysis of user electricity consumption data faces many challenges. Existing technologies ignore the diversity of user behavior, and single feature analysis leads to incomplete classification. In addition, since user electricity consumption faces a variety of complex factors, it is difficult to establish an accurate correlation model between user types, user electricity consumption time series characteristics and climate characteristics. Traditional linear methods cannot handle nonlinear relationships, resulting in reduced short-term prediction accuracy. Especially under the influence of climate variability, such as extreme weather and seasonal changes, these problems exacerbate the uncertainty of electricity consumption forecasts. Some current analysis methods, such as pure supervised learning, need to rely on a large amount of labeled data, but when labeled data is scarce (such as special periods or extreme climates), the model generalization ability is poor, the data verification accuracy is low, and the feature correlation is low. To address the above pain points, this application provides an efficient and robust data analysis modeling framework. Summary of the Invention

[0003] In order to solve the problems of low data verification accuracy and large characteristic correlation deviation in the existing technology, this application proposes a user electricity consumption data analysis modeling method and system based on semi-supervised learning, which can improve the accuracy of electricity consumption prediction.

[0004] The technical solution adopted in this application is: a user electricity consumption data analysis and modeling method based on semi-supervised learning, comprising the following steps:

[0005] S1: Electricity consumption data verification and tracking feedback model construction: Based on user electricity consumption characteristic data and climate characteristic data, a semi-supervised learning-driven electricity consumption data verification and tracking feedback model is constructed using a graph convolutional network. The nodes of the graph convolutional network are feature vectors that integrate user electricity consumption characteristic data and climate characteristic data, and the edges of the graph convolutional network represent the similarity of electricity consumption behavior between users and the relationship between user electricity consumption behavior and climate data.

[0006] S2: Characteristic Correlation Model Construction: Calculate the Pearson correlation coefficient between the electricity consumption characteristics of different types of users, and build a characteristic correlation model based on a deep belief network to analyze the nonlinear relationship between user type, user electricity consumption time series characteristics, and climate characteristics;

[0007] S3: Comprehensive analysis model of factors affecting electricity consumption changes: Based on the grey correlation analysis method, it captures the distribution characteristics of electricity consumption changes.

[0008] Furthermore, the user power consumption characteristic data is the power, voltage and current time series data of different types of users over a period of time obtained through the power system. The obtained data is cleaned based on the power grid quality specifications, and the power consumption intensity index is extracted.

[0009] Furthermore, climate characteristic data is extracted from meteorological data, which is obtained by matching meteorological station data with power grid substations through spatiotemporal alignment. The climate characteristic data includes temperature sensitivity coefficients and extreme weather markers.

[0010] Furthermore, each node of the graph convolutional network represents an electricity user entity, including marked nodes and unmarked nodes. The marked nodes are users with meteorological disaster day labels, and the unmarked nodes are ordinary users.

[0011] Edge Weights in Graph Convolutional Networks for:

[0012] ;

[0013] in, is the normalized scale factor of curve similarity.

[0014] Furthermore, the loss function of the electricity consumption data verification and tracking feedback model adopts the binary cross entropy loss function, which is expressed as follows:

[0015] ;

[0016] in, To monitor losses, is the topological loss; is the regularization coefficient;

[0017] The regularization term is the introduced grid topology constraint term, and its expression is:

[0018] ;

[0019] in, is the distribution network connection relationship, For users The eigenvector of For users The eigenvector of

[0020] Graph convolutional networks are used to aggregate label information of labeled nodes and neighbor information of unlabeled nodes.

[0021] Furthermore, the shared weights of the graph convolutional network are updated through the loss function.

[0022] Furthermore, the deep belief network uses the restricted Boltzmann machine to pre-train weights layer by layer, and then fine-tunes through backpropagation to learn the correlation between user type, user electricity consumption timing characteristics and climate characteristics. The output of the deep belief network is the correlation matrix.

[0023] A user electricity consumption data analysis system based on semi-supervised learning, comprising:

[0024] Data processing module: used to process user electricity consumption data and meteorological data to obtain user electricity consumption characteristic data and climate characteristic data;

[0025] Semi-supervised learning module: used to train graph convolutional networks and aggregate labeled data and unlabeled neighbor information on graph convolutional network nodes;

[0026] Correlation analysis module: used to analyze the distribution characteristics of electricity consumption changes through deep confidence network and grey analysis method based on the Pearson correlation coefficient between electricity consumption characteristics of different types of users.

[0027] Furthermore, the training data of the deep belief network uses unlabeled electricity consumption curves, and the fine-tuning data uses labeled data containing meteorological disaster day labels.

[0028] Furthermore, the node input of the graph neural network is a feature vector composed of electricity consumption data of various types of users and climate characteristics, and the edges of the graph neural network are the similarities in electricity consumption behaviors between users and the relationship between users' electricity consumption behaviors and climate data.

[0029] The beneficial effects of this application compared to the prior art are:

[0030] 1. Improve data verification accuracy: Reduce reliance on labeled data through semi-supervised learning, reducing model error by more than 20% in extreme climates;

[0031] 2. Enhanced feature correlation: The Pearson correlation coefficient deviation is controlled within the range of ±0.1, which improves the robustness of the correlation model;

[0032] 3. Support accurate forecasting: It provides reliable input for short-term forecasting, improving forecast accuracy to over 97%, and optimizing power resource allocation and grid resilience. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present application will be further described below with reference to the accompanying drawings:

[0034] Figure 1 The overall flow chart provided for the embodiment of this application;

[0035] Figure 2This is a diagram of the power consumption data verification and tracking feedback model framework provided in the embodiment of this application, showing the GCN structure and loss function application;

[0036] Figure 3 A framework diagram of the association model between user type, user power consumption time series characteristics and climate characteristics provided in the embodiment of the present application, showing the DBN layer structure;

[0037] Figure 4 The flowchart of the comprehensive analysis model of factors affecting changes in user electricity consumption provided in the embodiment of the present application shows the steps of grey correlation analysis. DETAILED DESCRIPTION

[0038] like Figures 1 to 4 As shown, the present application provides a user electricity consumption data analysis and modeling method based on semi-supervised learning. The method constructs an electricity consumption data verification and tracking feedback model through semi-supervised learning, uses the Pearson correlation coefficient and deep belief network to establish a correlation model between user type, user electricity consumption time series characteristics and climate characteristics, and combines gray correlation analysis to construct a comprehensive analysis model of factors affecting user electricity consumption changes. It can solve the following problems: (1) The low accuracy of the data verification model caused by the small shared weight in semi-supervised learning; (2) The low correlation between user type, user electricity consumption time series characteristics and climate characteristics caused by the large deviation of the Pearson correlation coefficient; (3) The nonlinear relationship between climate variability and electricity consumption is difficult to accurately simulate. Through the method of the present application, the performance of the model in scenarios with scarce labeled data can be effectively improved to ensure prediction accuracy.

[0039] The modeling method of this application includes the following three steps:

[0040] S1: Electricity consumption data verification and tracking feedback model construction: Taking various types of electricity consumption data after user electricity consumption behavior classification as input, a semi-supervised learning method is adopted, using the graph convolutional network (GCN) to aggregate labels and unlabeled neighbor information, and updating the shared weights through the binary cross-entropy loss function.

[0041] The various types of electricity consumption data obtained after classifying user electricity usage behavior in step S1 include time-series data such as power, voltage, and current. In this embodiment, all of this data comes from the power system. 15-minute power, voltage, and current time-series data can be collected from residential, agricultural, industrial, or commercial users. Based on the grid quality specification (GB / T 31960), this data is cleaned of abnormalities, power system-specific indicators are extracted, and the nonlinear relationship between user type and climate characteristics is analyzed. This generates an electricity intensity index: electricity intensity = daily maximum load / transformer rated capacity, resulting in user electricity consumption time-series characteristic data. The electricity intensity index, as a core component of node features, directly participates in graph neural network calculations.

[0042] Weather station data is aligned with power grid stations through spatiotemporal alignment, and climate characteristic data is extracted from the meteorological data. This climate characteristic data includes temperature sensitivity coefficients and extreme weather markers. The temperature sensitivity coefficient, k_temp, is calculated by dividing Δpower consumption by Δtemperature. Extreme weather markers include red alerts for heavy rain or high temperatures. Δpower consumption represents the change in temperature over adjacent time periods, while Δtemperature represents the change in power consumption over the same time period. The temperature sensitivity coefficient quantifies the fluctuation in power consumption caused by a unit temperature change.

[0043] The collaborative principle and implementation mechanism of the semi-supervised learning method and graph convolutional network are as follows:

[0044] 1. Graph structure design (as a semi-supervised information carrier)

[0045] The node input of the graph neural network is a feature vector composed of electricity consumption data of various types of users and climate characteristics. The edges of the graph neural network are the similarities between electricity consumption behaviors of users and the relationship between users' electricity consumption behaviors and climate data.

[0046] Its node definition is as follows:

[0047] Each node represents an electricity user entity, and the feature vector is expressed as: [user type code, current period power, temperature, humidity], and the nodes include marked nodes and unmarked nodes, where:

[0048] Marked nodes: users with the meteorological disaster day label (accounting for <5%);

[0049] Unlabeled nodes: ordinary users (accounting for >95%);

[0050] Unlabeled neighbor information of labeled nodes: This refers to the data of unlabeled nodes connected to labeled nodes. Neighbor relationships are defined using edge weights (based on the similarity of electricity usage curves). Graph convolutional networks utilize this unlabeled neighbor information to expand decision boundaries during aggregation.

[0051] The edge weight in the graph structure is defined as:

[0052] (1);

[0053] in, is the normalized scale factor of curve similarity. In this embodiment =0.5.

[0054] By building adjacency relationships based on the similarity of power curves, more information can be transferred between nodes with similar power patterns.

[0055] In the graph structure design of this application, edge weights are calculated only by the similarity of electricity consumption curves between users and do not directly include climate data. The technical rationality of this is due to the following:

[0056] Functional separation principle: Climate characteristics have been integrated into node feature vectors (such as [user type code, current time period power, temperature, humidity]), which indirectly affect the association strength through graph convolution propagation;

[0057] Computational efficiency optimization: avoids the expansion of edge weight matrices caused by high-dimensional climate data and improves the real-time performance of the model;

[0058] Dynamic coupling mechanism: Under extreme climate conditions, users' electricity consumption behaviors spontaneously converge or diverge (e.g., air conditioning load surges on hot days), which is automatically reflected in edge weights. Changing.

[0059] 2. Semi-supervised learning mechanism (aggregating labels and neighbor information)

[0060] The information dissemination formula is as follows:

[0061] (2);

[0062] in, is the normalized adjacency matrix (including edge weights), For the Layer node features, is the activation function (such as ReLU), is the shared weight.

[0063] Semi-supervised implementation process:

[0064] 1) Label node initialization: The disaster day label is used as a supervisory signal (e.g., 0 / 1 indicates anomaly).

[0065] 2) Neighbor aggregation:

[0066] In each iteration, the labeled node transfers label information to the neighbors of the unlabeled nodes;

[0067] Unlabeled nodes fuse their own features with neighbor labels (weighted average).

[0068] 3) Topology constraint regularization, which is the grid topology constraint, is expressed as:

[0069] (3);

[0070] in, is the topological loss; is the distribution network connection relationship (i.e., the grid substation topology), which is the association relationship defined by the transformer-user physical topology. For example, if substation T1 contains users {A, B, C}, then E = {(A, B), (A, C), (B, C)}; For users The characteristic vector of , which integrates the multi-dimensional representation of electricity intensity + climate response; For users The eigenvector of is the edge weight;

[0071] Topology-constrained regularization can enforce similar feature representations for users in the same substation and enhance the generalization ability of the model by utilizing unlabeled data.

[0072] 4) Loss function and weight update, used to solve the problem of too small shared weights. The loss function is a joint loss function, which is essentially a binary cross entropy loss function, and its expression is as follows:

[0073] (4);

[0074] in, is the binary cross entropy loss (i.e., supervision loss) of the labeled nodes, which is used to update the shared weights , updated via back-propagation and optimized simultaneously via gradient descent; It is the topological loss (i.e., the regularization term), which is used to ensure the consistency of predictions of adjacent nodes. At the same time, it can avoid the degradation of feature expression caused by too small weights, and enable unlabeled data to participate in parameter optimization through topological constraints. It is the regularization coefficient, the default value is 0.7, which is used to balance the weight of supervision loss and topology constraint.

[0075] The output of the graph convolutional network is the normalized data after verification, which is used for tracking feedback. After passing through multiple layers of the graph convolutional network, the feature vector of each user node contains:

[0076] Standardized electricity consumption characteristics (peak hour electricity consumption ratio, load fluctuation rate);

[0077] Climate response intensity (optimized value of temperature sensitivity coefficient);

[0078] Abnormality probability score (0-1 range, such as 0.85 indicates high risk).

[0079] The synergy between semi-supervised learning methods and graph convolutional networks has the following advantages in power grid scenarios:

[0080] 1. Reduce tag dependency:

[0081] Only a small number of disaster day labels (e.g., 5,000) are needed to associate 100,000+ unlabeled users through the station area topology;

[0082] When labeled data is scarce, the anomaly detection accuracy still reaches 93.8%.

[0083] 2. Enhanced robustness:

[0084] Regularization term Can suppress noise effects (such as measurement errors);

[0085] Edge weight Irrelevant users are filtered out (when the difference in electricity consumption curve is greater than the threshold, ≈0).

[0086] 3. Engineering adaptability:

[0087] The grid area topology (E) is directly mapped into a graph structure;

[0088] The 15-minute data granularity can support real-time tracking and feedback.

[0089] Through the above mechanism, semi-supervised learning and graph convolutional networks form a closed loop:

[0090] Labeled node data guides initial training → Unlabeled node data expands the decision boundary through the graph structure → Grid topology constraints improve generalization capabilities → Backpropagation simultaneously optimizes supervised and unsupervised objectives, and finally aggregates node label information and unlabeled node neighbor information.

[0091] S2: Characteristic association model construction: Calculate the Pearson correlation coefficient between the electricity consumption characteristics of different types of users, extract nonlinear relationships as input variables of the deep belief network, and combine the deep belief network to establish a deep association between user type, user electricity consumption time series characteristics and climate characteristics.

[0092] Pearson correlation coefficient between electricity consumption characteristics of different types of users The calculation formula is as follows:

[0093] (4);

[0094] in, For variables No. Observation values ​​can be expressed as the power consumption characteristics of user type 1 (such as the 15-minute power Pres of residential users), is the typical daily benchmark load, For variables No. observations, which can be expressed as the power consumption characteristics of user type 2 (such as the voltage deviation rate ΔVind of industrial users), is the benchmark value, the typical daily benchmark value of user type 2 (such as the industrial voltage qualified benchmark 220V±7%), for Standard deviation of electricity consumption fluctuation intensity of user type 1: The larger it is, the more unstable the electricity consumption becomes (for example, the fluctuation of residential air conditioning load on rainy days can reach ±40%). for Standard deviation of electricity consumption fluctuation intensity of user type 2: The smaller → the more stable the power consumption (for example, the voltage deviation rate in high-end manufacturing industry is <3%).

[0095] The Pearson correlation coefficient can be used as the structured input of the deep belief network. The input vector of the deep belief network is constructed as: [resident code, industrial voltage deviation, ], the input vector is directly input into the first layer restricted Boltzmann machine (RBM) of the deep belief network.

[0096] The Pearson correlation coefficient can also guide the pre-training direction of the deep belief network. |Value association can obtain higher connection weight.

[0097] The deep belief network consists of an input layer, a hidden layer, and an output layer. The input layer receives data on the time series characteristics of user electricity consumption and climate characteristics. The hidden layer is composed of multiple stacked restricted Boltzmann machines (RBMs). Each layer learns a higher-level feature representation of the input data and outputs the relationship between user type, user electricity consumption time series characteristics, and climate characteristics through the output layer.

[0098] In this embodiment, the deep belief network uses restricted Boltzmann machine pre-training and back-propagation fine-tuning. The weights are pre-trained layer by layer using the restricted Boltzmann machine, and then fine-tuned through back-propagation to learn the correlation between user type, the temporal characteristics of user electricity consumption, and climate characteristics (such as temperature and humidity). The output of the deep belief network is a correlation matrix, which is used for influencing factor analysis and quantifies the regulatory effect of user type on climate response. The correlation matrix serves as input for grey relational analysis.

[0099] In this embodiment, the training data of the deep belief network can be the unlabeled electricity consumption curve of a certain power grid in a certain year, and the fine-tuning data can be the labeled data containing the meteorological disaster day label. The input layer data includes power, temperature, and humidity. After being downsampled layer by layer by the multi-layer restricted Boltzmann machine, the output power consumption pattern code can be expressed as:

[0100] Input layer (power + temperature and humidity) → RBM1 (128 nodes) → RBM2 (64 nodes) → Output layer (power pattern encoding). The 128 nodes and 64 nodes represent the number of neurons in the RBM1 and RBM2 layers, respectively. Each node processes the features of the input data, gradually abstracting higher-order feature representations. The output power pattern encoding is the structured feature vector of the DBN output layer, generated through two levels of compression:

[0101] Dimensional Condensation:

[0102] Input layer (power + temperature and humidity) → Original dimension: 15 minutes × 24 hours × 3 indicators = 288 dimensions;

[0103] RBM2 output layer → compressed to 64-dimensional feature vector.

[0104] In feature association modeling, the core nonlinear relationships occur between the following two groups of elements:

[0105] 1. User electricity consumption time series characteristics vs. climate characteristics:

[0106] Time series characteristics: Dynamic characteristics of electricity consumption by user type (residential / agricultural / industrial and commercial), including:

[0107] Peak-to-valley ratio of electricity consumption (daily load curve fluctuation);

[0108] Load fluctuation rate (standard deviation of power variation at 15-minute level);

[0109] Daily load rate (ratio of average load to maximum load);

[0110] Climate characteristics: Environmental characteristics driven by meteorological data, including:

[0111] Temperature sensitivity coefficient (Δpower consumption / Δtemperature);

[0112] Extreme weather mark (red alert for heavy rain / high temperature).

[0113] Non-linear performance: The impact of climate factors on electricity consumption varies non-proportionally. For example:

[0114] When the temperature exceeds 30°C, the air conditioning load increases exponentially;

[0115] On rainstorm days, residential electricity consumption drops sharply, but agricultural irrigation electricity consumption surges (user type coupling effect).

[0116] 2. User Type vs. Electricity Consumption-Climate Correlation Pattern:

[0117] User type: Classification labels for electricity usage by residential, agricultural, industrial and commercial;

[0118] Correlation patterns: Hidden patterns mined by deep belief networks, such as:

[0119] Industrial and commercial users: For every 1°C increase in temperature, the air conditioning load increases by 8% (mainly linear);

[0120] Agricultural users: There is a logarithmic relationship between precipitation and electricity consumption for pumping irrigation (significant nonlinearity).

[0121] S3: Comprehensive analysis model of factors affecting electricity consumption changes: Based on the grey correlation analysis method, it captures the distribution characteristics of electricity consumption changes and lays the foundation for short-term forecasts.

[0122] The comprehensive analysis model of factors affecting electricity consumption changes is based on the output of the characteristic correlation model, namely the correlation matrix, and is implemented using the grey correlation analysis method. The main steps are as follows:

[0123] Select the reference series (main electricity consumption change characteristics of users) and the comparison series (climate and other factors) to calculate the grey correlation coefficient :

[0124] (5);

[0125] in, The deviation of the power consumption change characteristic sequence, which indicates the difference between the user's baseline power consumption behavior (such as the typical daily load curve) and the actual value in the first Example: On a rainy day at 08:00, the residential load suddenly increases by +120kW (baseline value 300kW → actual 420kW). =120. Climate factors The deviation of The synergistic deviation between a climate characteristic (such as temperature) and the change in electricity consumption. Calculation: ∥Temperature series value − Electricity consumption change series value ∥. Example: When the temperature suddenly changes by 5°C, ΔTemperature(k) = 120 (with of the same magnitude). middle, is the resolution coefficient ( ∈(0,1)), used to adjust the sensitivity of the correlation coefficient to discrete data. Grid value: = 0.3~0.5 (0.3 on rainstorm days, 0.4 on high temperature days). Function: Suppress noise interference and strengthen the dominant climate influence. is the global maximum deviation, indicating that and The maximum value of . is a time series index (discrete time point). The time point of 15-minute electricity consumption data (such as =32 means 08:00 on the same day). Corresponding to: granularity of the electricity consumption information collection system.

[0126] Calculate grey relational degree , and assign weights , get the comprehensive score .

[0127] The output of the comprehensive analysis model of factors affecting electricity consumption changes is the distribution characteristics of electricity consumption changes, which can support short-term forecasts.

[0128] In this embodiment, the reference sequence can be selected = Typical power consumption curve of the substation, the comparison series can be selected =[temperature, precipitation, wind speed].

[0129] This application's method innovatively integrates semi-supervised learning, deep belief networks, and grey relational analysis to better handle data noise and nonlinearity caused by climate variability. All algorithm parameters are optimized according to grid operation standards (such as GB / T 31960 and Q / GDW 12072).

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A user electricity consumption data analysis and modeling method based on semi-supervised learning, characterized by: The following steps are involved: S1: Electricity consumption data verification and tracking feedback model construction: Based on user electricity consumption characteristic data and climate characteristic data, a semi-supervised learning-driven electricity consumption data verification and tracking feedback model is constructed using a graph convolutional network. The nodes of the graph convolutional network are feature vectors that integrate user electricity consumption characteristic data and climate characteristic data, and the edges of the graph convolutional network represent the similarity of electricity consumption behavior between users and the relationship between user electricity consumption behavior and climate data. S2: Characteristic Correlation Model Construction: Calculate the Pearson correlation coefficient between the electricity consumption characteristics of different types of users, and build a characteristic correlation model based on a deep belief network to analyze the nonlinear relationship between user type, user electricity consumption time series characteristics, and climate characteristics; S3: Comprehensive analysis model of factors affecting electricity consumption changes: Based on the grey correlation analysis method, it captures the distribution characteristics of electricity consumption changes.

2. The method for analyzing and modeling user electricity consumption data based on semi-supervised learning according to claim 1, characterized in that: The user electricity consumption characteristic data is the power, voltage and current time series data of different types of users over a period of time obtained through the power system. The acquired data is cleaned based on the grid quality specifications, and the electricity consumption intensity index is extracted.

3. The user electricity consumption data analysis and modeling method based on semi-supervised learning according to claim 1 is characterized by: Climate characteristic data is extracted from meteorological data. The meteorological data is obtained by matching meteorological station data with power grid substations through spatiotemporal alignment. Climate characteristic data includes temperature sensitivity coefficients and extreme weather markers.

4. The method for analyzing and modeling user electricity consumption data based on semi-supervised learning according to claim 1, characterized in that: Each node in the graph convolutional network represents an electricity user entity, including marked nodes and unmarked nodes. The marked nodes are users with meteorological disaster day labels, and the unmarked nodes are ordinary users. Edge Weights in Graph Convolutional Networks for: ； in, is the normalized scale factor of curve similarity.

5. The method for analyzing and modeling user electricity consumption data based on semi-supervised learning according to claim 4, characterized in that: The loss function of the electricity consumption data inspection and tracking feedback model adopts the binary cross entropy loss function, which is expressed as follows: ； in, To monitor losses, is the topological loss; is the regularization coefficient; The regularization term is the introduced grid topology constraint term, and its expression is: ； in, is the distribution network connection relationship, For users The eigenvector of For users The eigenvector of Graph convolutional networks are used to aggregate label information of labeled nodes and neighbor information of unlabeled nodes.

6. The method for analyzing and modeling user electricity consumption data based on semi-supervised learning according to claim 5, characterized in that: Update the shared weights of the graph convolutional network through the loss function.

7. The method for analyzing and modeling user electricity consumption data based on semi-supervised learning according to claim 1, characterized in that: The deep belief network uses restricted Boltzmann machines to pre-train weights layer by layer, and then fine-tunes through backpropagation to learn the correlation between user type, user electricity consumption timing characteristics and climate characteristics. The output of the deep belief network is a correlation matrix.

8. A user electricity consumption data analysis system based on semi-supervised learning, characterized by: A user electricity consumption data analysis and modeling method based on semi-supervised learning according to any one of claims 1 to 7 is adopted, comprising: Data processing module: used to process user electricity consumption data and meteorological data to obtain user electricity consumption characteristic data and climate characteristic data; Semi-supervised learning module: used to train graph convolutional networks and aggregate labeled data and unlabeled neighbor information on graph convolutional network nodes; Correlation analysis module: used to analyze the distribution characteristics of electricity consumption changes through deep confidence network and grey analysis method based on the Pearson correlation coefficient between electricity consumption characteristics of different types of users.

9. The user electricity consumption data analysis system based on semi-supervised learning according to claim 8, characterized in that: The training data of the deep belief network uses unlabeled electricity consumption curves, and the fine-tuning data uses labeled data containing meteorological disaster day labels.

10. The user electricity consumption data analysis system based on semi-supervised learning according to claim 8, characterized in that: The node input of the graph neural network is a feature vector composed of electricity consumption data of various types of users and climate characteristics. The edges of the graph neural network are the similarities between electricity consumption behaviors of users and the relationship between users' electricity consumption behaviors and climate data.

Images