Distributed flexible load modeling method and system for power distribution network reliability evaluation

Abstract

The application discloses a kind of distributed flexible load modeling methods and systems for reliability evaluation of distribution network, wherein the method comprises: based on parallel computing and merging classification method, the electricity consumption behavior of residential user is classified, and the classification result is converted into gray image;Based on multi-information fusion, a set of demand curve functions is constructed, and an image training sample is generated;The image training sample is used as input, and a nonlinear demand price elasticity evaluation model based on deep learning method is trained;The gray image is input into the trained model, the polynomial representation of demand curve is extracted, and a set of demand price elasticity curve polynomials is output;Based on historical data and demand price elasticity curve polynomial set, a distributed flexible load model is constructed to evaluate the electricity consumption of a single user under a given electricity price, which can be applied to personalized dynamic pricing and distribution network reliability evaluation models, providing strong support for the optimized operation and sustainable development of power systems.

Description

Technical Field

[0001] This invention relates to the field of power distribution network system technology, and in particular to a distributed flexible load modeling method and system for power distribution network reliability assessment. Background Technology

[0002] With the advancement of global energy structure transformation, demand response (DR) technology in smart distribution networks is considered a key tool for improving power system efficiency, ensuring power supply security, and enhancing the accuracy of distribution network reliability assessments. As a method of regulating electricity demand through user participation, demand response has become an indispensable part of the modern electricity market. By encouraging users to adjust their electricity consumption patterns according to changes in power supply and demand, demand response can not only effectively reduce power system operating costs, minimize peak-valley differences, and improve grid flexibility, but also alleviate the load on power equipment and extend the lifespan of power infrastructure, thereby promoting more sustainable energy development.

[0003] Driven by electricity market reforms and green energy policies, an increasing number of electricity consumers, especially residential users, are participating in demand response programs. Residential users constitute a significant proportion of the electricity system, accounting for approximately 20% of global electricity consumption. Therefore, researching and optimizing residential demand response has become an important direction in electricity system and market research. Especially with the increasing proportion of renewable energy sources (such as wind and solar power), residential demand response not only helps regulate electricity load but also aligns with green electricity markets, further promoting the integration of renewable energy.

[0004] Demand response programs can be broadly categorized into incentive-based and price-based types. Incentive-based demand response programs typically use direct load control or reward mechanisms to encourage users to reduce electricity consumption during periods of high demand or increase it during periods of low demand, thereby regulating electricity consumption. For residential users, incentive-based demand response is often more common, primarily because household electricity consumption is dispersed and relatively small, making direct incentives or load control more effective. However, with the deepening reforms of the electricity market and increased user sensitivity to electricity prices, price-based demand response is gradually becoming an effective regulatory tool. In price-based demand response, retailers set dynamic electricity prices based on real-time electricity supply and demand conditions, encouraging users to adjust their electricity consumption behavior in response to price fluctuations. This model not only helps optimize grid operation but also effectively stimulates flexible consumer responses.

[0005] However, despite the significant potential of price-based demand response (DR) in theory, it also faces many challenges.

[0006] First, the relationship between electricity demand and price is often non-linear, and most existing studies are based on linear price elasticity of demand (PED) models, which limits the understanding of complex electricity consumption behavior. Traditional PED models assume a linear relationship between price and demand, but in reality, household electricity consumption is often influenced by various factors, such as weather changes, appliance type, and timing of electricity use. Therefore, linear models often fail to adequately capture these non-linear characteristics, leading to poor predictive performance.

[0007] Secondly, deep learning-based methods, as a data-driven approach, have gradually been applied in demand response research. Deep learning models can extract implicit patterns from large amounts of historical data, thereby predicting the non-linear relationship between electricity consumption patterns and prices. The advantages of these methods lie in their powerful modeling capabilities and flexibility, enabling them to handle complex and dynamic electricity demands. However, the effectiveness of deep learning models typically relies on large amounts of user data, especially detailed usage information of various appliances within a household. In practical applications, collecting such large-scale data and effectively modeling it is often impractical, especially when it involves the privacy of individual users. Appliance information and user preferences are usually private data, and this data may be incomplete or inaccurate. Therefore, the application of traditional deep learning methods in real-world scenarios faces significant challenges.

[0008] To address these issues, recent research has explored new approaches. For example, some studies have proposed utilizing multi-information fusion techniques and cluster analysis methods to avoid over-reliance on precise household appliance data. By classifying residential users' electricity consumption patterns, researchers can develop simpler and more efficient forecasting methods by comprehensively analyzing the common characteristics of user groups. These methods not only improve the adaptability of models but also reduce dependence on data integrity, making them more feasible in practice. Furthermore, with the increasing complexity of demand response systems, traditional demand response models often struggle to handle large-scale user data or cope with the uncertainty and high dynamism of the electricity market. Summary of the Invention

[0009] To address the aforementioned issues, this invention proposes a distributed flexible load modeling method and system for distribution network reliability assessment. It employs a parallel computing and merging (PC&M) classification method to classify user electricity consumption patterns without assumptions, thereby improving the adaptability of the classification. Furthermore, it utilizes multi-information fusion technology to generate training samples and combines deep learning methods to establish a nonlinear PED model for distributed flexible load modeling. This model can be applied to personalized dynamic pricing, distribution network reliability assessment, and other models.

[0010] The technical solution adopted in this invention is as follows:

[0011] A distributed flexible load modeling method for distribution network reliability assessment includes:

[0012] The electricity consumption behavior of residential users is classified based on parallel computing and merging classification methods, and the classification results are converted into grayscale images.

[0013] A set of demand curve functions is constructed based on multi-information fusion, and image training samples are generated.

[0014] The image training samples are used as input to train a nonlinear demand price elasticity assessment model based on deep learning methods; the grayscale image is input into the trained model to extract the polynomial representation of the demand curve and output a set of polynomials of the demand price elasticity curve.

[0015] Based on historical data and the set of polynomials of demand price elasticity curves, a distributed flexible load model is constructed to evaluate the electricity consumption of a single user at a given electricity price, serving as an input parameter for distribution network reliability assessment.

[0016] Furthermore, the method of classifying residential users' electricity consumption behavior based on parallel computing and merging classification, and converting the classification results into grayscale images, includes:

[0017] Data preprocessing: The collected residential electricity consumption data is normalized and dimensionality is reduced using principal component analysis algorithm. The processed data is then evenly distributed to each data block.

[0018] Parallel computing: Electricity consumption behavior samples are calculated separately in each data block. The cumulative proximity, eccentricity and standardized eccentricity are used for recursive calculation. The category of new electricity consumption behavior samples is determined by Chebyshev's inequality.

[0019] Category merging: Merge the categories collected from each data block by finding the nearest neighbor category and merging when the merging conditions are met, and update the parameters of the merged category;

[0020] Data integration: Pair residential electricity consumption in each category with the corresponding 24-hour electricity price data to form data points;

[0021] Image conversion: Convert hourly data points into grayscale images, generating 24 images for each category.

[0022] Furthermore, the step of constructing a set of demand curve functions based on multi-information fusion and generating image training samples includes:

[0023] Construction of the demand curve function set: Logarithmic demand function and exponential demand function are selected as the basis, and parameters are adjusted through experiments to balance stability and diversity;

[0024] Data point generation: Randomly select the demand curve function to generate data points to form image training samples;

[0025] Image training sample generation: Based on the set of demand curve functions and data generation functions, a large number of data points are synthesized for model training.

[0026] Further, the image training samples are used as input to train a nonlinear demand price elasticity assessment model based on a deep learning method; the grayscale image is input into the trained model to extract the polynomial representation of the demand curve, and the set of polynomials for the demand price elasticity curve is output, including:

[0027] Model training: The nonlinear demand price elasticity assessment model based on the attention U-Net network architecture is trained using generated image training samples to learn the intrinsic relationship between electricity price and electricity consumption;

[0028] Pixel mapping and polynomial fitting: Extracting a polynomial representation of the demand curve from the input grayscale image using a trained model;

[0029] Price elasticity assessment: The set of polynomial curves for demand price elasticity is obtained by pixel mapping and polynomial fitting.

[0030] Furthermore, the construction of the distributed flexible load model based on historical data and the set of polynomials of demand price elasticity curves includes:

[0031] Data segmentation: The test data is segmented into weekdays and rest days, and weather factors are reflected in residents' electricity consumption behavior. Historical data for preset weekdays is collected, including electricity prices and corresponding user electricity consumption.

[0032] Model Construction: A distributed flexible load model is established using the obtained electricity consumption behavior classification results and the nonlinear demand price elasticity assessment model;

[0033] Recalculation and Update: If the assessment date changes, the classification of residential user electricity consumption behavior, the assessment of electricity demand price elasticity, and the modeling of distributed flexible loads will be re-performed.

[0034] A distributed flexible load modeling system for distribution network reliability assessment includes:

[0035] The electricity consumption behavior classification module is configured to classify the electricity consumption behavior of residential users based on parallel computing and merging classification methods, and convert the classification results into grayscale images;

[0036] The training sample generation module is configured to construct a set of demand curve functions based on multi-information fusion and generate image training samples;

[0037] The evaluation model building module is configured to take the image training samples as input to train a nonlinear demand price elasticity evaluation model based on deep learning methods; input the grayscale image into the trained model, extract the polynomial representation of the demand curve, and output a set of polynomials of the demand price elasticity curve.

[0038] The load model building module is configured to construct a distributed flexible load model based on historical data and the set of polynomials of the demand price elasticity curves to evaluate the electricity consumption of a single user at a given electricity price, serving as input parameters for distribution network reliability assessment.

[0039] Furthermore, the electricity consumption behavior classification module includes:

[0040] The data preprocessing unit is configured to normalize the collected residential electricity consumption behavior data and use principal component analysis algorithm to reduce dimensionality, and distribute the processed data evenly to each data block;

[0041] Parallel computing units are configured to perform electricity consumption behavior sample calculations in each data block, using three indicators—cumulative proximity, eccentricity, and standardized eccentricity—for recursive calculations, and determining the category of new electricity consumption behavior samples using Chebyshev's inequality.

[0042] The category merging unit is configured to merge categories collected from various data blocks by finding the nearest neighbor category and merging when the merging conditions are met, and updating the parameters of the merged category.

[0043] The data integration unit is configured to pair residential electricity consumption in each category with the corresponding 24-hour electricity price data to form data points;

[0044] The image conversion unit is configured to convert hourly data points into grayscale images, generating 24 images for each category.

[0045] Furthermore, the training sample generation module includes:

[0046] The demand curve function set building unit is configured to select logarithmic demand function and exponential demand function as the basis, and adjust the parameters through experiments to balance stability and diversity.

[0047] The data point generation unit is configured to randomly select the demand curve function to generate data points, which constitute image training samples.

[0048] The image training sample generation unit is configured to synthesize a large number of data points based on the demand curve function set and the data generation function for model training.

[0049] Furthermore, the evaluation model construction module includes:

[0050] The model training unit is configured to train a nonlinear demand price elasticity assessment model based on an attention U-Net network architecture using generated image training samples, and to learn the intrinsic relationship between electricity price and electricity consumption.

[0051] The pixel mapping and polynomial fitting unit is configured to extract a polynomial representation of the demand curve from the input grayscale image through the trained model.

[0052] The price elasticity assessment unit is configured to obtain a set of polynomials for the demand price elasticity curves through pixel mapping and polynomial fitting.

[0053] Furthermore, the load model construction module includes:

[0054] The data segmentation unit is configured to divide the test data into weekdays and rest days, and to reflect weather factors in residents' electricity consumption behavior, and to collect historical data for preset weekdays, including electricity prices and corresponding user electricity consumption.

[0055] The model building unit is configured to use the obtained electricity consumption behavior classification results and the nonlinear demand price elasticity assessment model to establish a distributed flexible load model.

[0056] The recalculation and update unit is configured to reclassify residential user electricity consumption behavior, assess electricity demand price elasticity, and model distributed flexible loads when the assessment date changes.

[0057] The beneficial effects of this invention are as follows:

[0058] This invention effectively improves the adaptability of residential user electricity consumption pattern classification by introducing a parallel computing and merging (PC&M) classification method, without over-reliance on precise appliance data, thus reducing data collection difficulty and privacy risks. Simultaneously, the use of multi-information fusion technology to generate training samples, combined with a nonlinear PED model built using deep learning methods, can more accurately capture the nonlinear relationship between electricity demand and price, improving prediction accuracy. Furthermore, the established distributed flexible load model can be applied to models such as personalized dynamic pricing and distribution network reliability assessment, providing strong support for the optimized operation and sustainable development of the power system. Attached Figure Description

[0059] Figure 1 This is a flowchart of a distributed flexible load modeling method for distribution network reliability assessment according to Embodiment 1 of the present invention.

[0060] Figure 2This is a schematic diagram of the PC&M classification method in Embodiment 2 of the present invention;

[0061] Figure 3 This is a visualization of the classification results in Embodiment 2 of the present invention;

[0062] Figure 4 This is a flowchart of the training sample generation process in Embodiment 2 of the present invention;

[0063] Figure 5 This is a structural diagram of the Attention U-net in Embodiment 2 of the present invention;

[0064] Figure 6 This is one of the schematic diagrams of the clustering results of the PC&M classification method in Embodiment 2 of the present invention;

[0065] Figure 7 This is the second schematic diagram of the clustering results of the PC&M classification method in Embodiment 2 of the present invention;

[0066] Figure 8 This is the third schematic diagram of the clustering results of the PC&M classification method in Embodiment 2 of the present invention;

[0067] Figure 9 This is a schematic diagram of the distributed flexible load assessment results in Embodiment 2 of the present invention. Detailed Implementation

[0068] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments are now described. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention; that is, the described embodiments are only a part of the embodiments of the invention, not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0069] Example 1

[0070] This embodiment provides a distributed flexible load modeling method for distribution network reliability assessment, such as... Figure 1 As shown, it includes:

[0071] The electricity consumption behavior of residential users is classified based on parallel computing and merging classification methods, and the classification results are converted into grayscale images.

[0072] A set of demand curve functions is constructed based on multi-information fusion, and image training samples are generated.

[0073] The image training samples are used as input to train a nonlinear demand price elasticity assessment model based on deep learning methods; the grayscale image is input into the trained model to extract the polynomial representation of the demand curve and output a set of polynomials of the demand price elasticity curve.

[0074] Based on historical data and the set of polynomials of demand price elasticity curves, a distributed flexible load model is constructed to evaluate the electricity consumption of a single user at a given electricity price, serving as an input parameter for distribution network reliability assessment.

[0075] In this embodiment, user electricity consumption patterns are classified without assumptions by using parallel computing and merging classification methods, thereby improving the adaptability of the classification; training samples are generated by multi-information fusion technology, and a nonlinear PED model is established by combining deep learning methods and used for distributed flexible load modeling, which can be applied to models such as personalized dynamic pricing and distribution network reliability assessment.

[0076] Specifically, the distributed flexible load modeling method in this embodiment can be implemented using the following steps:

[0077] Step 1: A PC&M-based classification method for residential electricity consumption behavior

[0078] Step 1.1: Data Preprocessing Module. The collected residential electricity consumption data is first normalized, and then dimensionality reduction is performed using principal component analysis. The processed data is then evenly distributed into various data blocks.

[0079] Step 1.2: Parallel Computing Module. The processed electricity consumption behavior samples are calculated separately in each data block. Using the PC&M framework, three indices—cumulative proximity, eccentricity, and standardized eccentricity—are recursively calculated to update the cumulative proximity and eccentricity of each sample, and parameters are recursively updated using Euclidean distance. Chebyshev's inequality is used to determine whether a new electricity consumption behavior sample belongs to a certain category. If the condition is met, the sample is assigned to the corresponding category; otherwise, a new category is created.

[0080] Step 1.3: Category Merging Module. The categories collected from each processor are merged. Starting with the smallest category, its nearest neighbor category is found. When the merging conditions are met, the two categories are merged, and the merging operation continues. The merged category updates its parameters.

[0081] Step 2: Visualization of electricity consumption behavior classification results

[0082] Step 2.1: Data Integration. Within each category, residential electricity consumption is paired with the corresponding 24-hour electricity price data, forming a large number of (price, electricity consumption) data points.

[0083] Step 2.2: Image Conversion. Convert the (price, electricity consumption) data points for each hour into grayscale images, generating 24 images for each category.

[0084] Step 2.3: Image Input Model. These grayscale images are used as input to the neural network trained in Step 4 to obtain the complex relationship between user price and electricity consumption.

[0085] Step 3: Generation of training samples based on multi-information fusion

[0086] Step 3.1: Construct a set of demand curve functions. Based on prior knowledge, construct demand curve functions with a monotonically decreasing shape. Select logarithmic and exponential demand functions as the basis, and adjust their parameters through experiments to balance stability and diversity.

[0087] Step 3.2: Data Point Generation. Randomly select a demand curve function and generate data points based on the curve. These data points, after visualization, constitute a pair of image samples.

[0088] Step 3.3: Image Training Sample Generation. Based on the demand curve function set and data generation function, a large number of data points are synthesized, overcoming the shortcomings of traditional methods that rely on historical data. The synthesized training samples can cover a variety of scenarios.

[0089] Step 4: PED evaluation model based on Attention U-Net

[0090] Step 4.1: Model Training. Using the generated image samples, a deep learning model based on the Attention U-Net mechanism is trained to learn the intrinsic relationship between electricity prices and electricity consumption. This model extracts feature information from the image through a convolutional neural network architecture and uses attention gates to focus on the most important regions.

[0091] Step 4.2: Pixel Mapping and Polynomial Fitting. The trained model can extract a polynomial representation of the demand curve from the input grayscale image. First, pixel coordinates in the image are converted to historical data coordinates through pixel mapping, and then a polynomial fitting method is used to obtain the 24-hour demand curve.

[0092] Step 4.3: Price Elasticity Assessment Model. The images of each category at each time point obtained in Step 2 are used as input to the trained model. The output RGB images are converted to grayscale images, and then the final 24-hour demand curve is obtained through pixel mapping and polynomial fitting. This process includes progressive mapping and fitting, ultimately yielding a set of polynomials for the demand price elasticity curves of each electricity consumption pattern category.

[0093] Step 5: Distributed flexible load modeling.

[0094] Step 5.1: Divide the test data into weekdays and rest days, and incorporate weather factors into residents' electricity consumption behavior. Collect historical data from the day before a specific workday (d-1) to workday dD, including electricity prices and corresponding user electricity consumption. Using the user electricity consumption behavior classification method and the electricity demand price elasticity assessment model mentioned in the previous steps, obtain the 24-hour demand price elasticity curve polynomial set for each electricity consumption pattern category. This is used to establish a distributed flexible load model and evaluate the electricity consumption of individual users under a given electricity price, serving as input parameters for distribution network reliability assessment.

[0095] Step 5.2: Recalculation and Update. If the assessment date changes, it is necessary to reclassify residential user electricity consumption behavior, assess the electricity demand price elasticity for each category, and model distributed flexible loads. Since the trained deep learning model does not require retraining, it does not incur additional computational burden.

[0096] Example 2

[0097] This embodiment is based on embodiment 1:

[0098] This embodiment provides a distributed flexible load modeling method for distribution network reliability assessment, including the following steps:

[0099] Step 1: Parallel Computation and Merging (PC&M) Classification Method Based on Typicality and Eccentricity Data Analysis (TEDA). TEDA, as a data analysis framework, relies solely on empirical data samples and does not contain any prior assumptions.

[0100] Step 1.1: Within this framework, three key metrics are proposed, as shown below.

[0101] Cumulative proximity π n (x i x represents the sum of distances between a specific data sample and all other data samples within the group. i The cumulative proximity is expressed by formula (1):

[0102]

[0103] Where, d 2 (x i ,x j ) represents x i and x j Distance metric between.

[0104] The eccentricity and standardized eccentricity are obtained by normalizing the cumulative proximity, and are expressed as formulas (2) and (3) respectively:

[0105]

[0106] in, It is the average cumulative proximity.

[0107] Step 1.2: For the task of classifying residential user electricity consumption behavior, traditional methods typically require pre-defining the number and centers of categories, which is subjective. The method in this embodiment avoids this problem, allowing the number and centers of categories to depend entirely on the original characteristics of historical electricity consumption pattern data. The framework of the PC&M classification method is as follows: Figure 2 As shown, this method comprises three independent modules, which will be explained in detail below.

[0108] Step 1.3: In the data processing module, all collected residential electricity consumption behavior data are first standardized, then principal component analysis algorithm is used for dimensionality reduction, and finally the processed data is evenly placed in each data block.

[0109] Step 1.4: In the parallel computing module, the processed electricity consumption behavior samples are computed on the corresponding processors in each data block. Based on the single-processing characteristic of this method, it can process large amounts of residential electricity consumption behavior data and perform recursive calculations.

[0110] Step 1.4.1: Key metrics in TEDA: Cumulative proximity, eccentricity, and standardized eccentricity, updated recursively using Euclidean distance. New data sample x n The cumulative proximity can be recursively calculated using formulas (4)-(6):

[0111] π n (x n )=n(||x n -μ n || 2 +X n -||μ n || 2 )

[0112]

[0113] Where μ represents the local mean of the recursive update, i.e. the center of each class; X represents the scalar product of the recursive update.

[0114] Step 1.4.2: The sum of cumulative proximity can be updated using formula (7):

[0115]

[0116] Step 1.4.3: Substituting formulas (6) and (7) into formula (1), we can obtain the expression for the standardized eccentricity, as shown in formula (8):

[0117]

[0118] Step 1.4.4: Chebyshev's inequality is used as a criterion to determine whether a new user's electricity consumption behavior sample belongs to a specific category. This criterion is based on standardized eccentricity and is restated as Equation (9):

[0119]

[0120] Where κ represents the distance between the new data sample and the local mean of a specific category being greater than κ times the standard deviation.

[0121] Step 1.4.5: The sum of the eccentricities of all samples within each category equals 2β, where β represents the total number of electricity consumption behaviors in that category. Therefore, the average standardized eccentricity of each category is 2. To balance sensitivity to outliers and tolerance for internal variance, the value of κ is set to 2 in this study. According to formula (9), the association condition between a new electricity consumption behavior sample and a certain category can be expressed as formula (10):

[0122]

[0123] Step 1.4.6: If a new electricity consumption behavior sample meets the conditions of multiple categories, its category will be determined by formula (11):

[0124]

[0125] Where, k * This represents the selected category, and K represents the number of existing categories in the processor. This is used to determine the new ECP data sample x. n After determining the appropriate category, the relevant parameters for that category are then updated according to formula (12).

[0126]

[0127] Step 1.4.7: If a sample does not meet the criteria for any category in formula (12), the formation of a new category will be initiated.

[0128] Step 1.5: In the category merging module, categories from each processor are grouped together. Then, starting with the smallest category, a search is performed for its nearest neighbor category. Once the conditions of formula (13) are met, the two categories are merged into a new category, and a new merging operation begins.

[0129]

[0130] Where μ i and μ j Let represent the mean of class i and class j, respectively. Calculated using formula (14):

[0131]

[0132] After the merging process, the parameters of the new category will be updated according to formula (15):

[0133]

[0134] Step 2: Visualization of Electricity Consumption Behavior Classification Results and Corresponding Electricity Prices. In this embodiment, directly analyzing the relationship between electricity consumption and individual residential prices reveals significant variability, making it challenging to accurately assess the price elasticity of specific users across different time intervals. This situation prompted us to adopt the designed PC&M classification method to classify electricity consumption patterns (ECP) across a broad group of end users, and then utilize machine vision technology for in-depth analysis of each category.

[0135] Step 2.1: After obtaining the classification results, firstly, the residential ECP data in each category is merged with the corresponding 24-hour price data, and then discretized into data points composed of price and consumption. Then, the data points corresponding to each hour are converted into grayscale images, ultimately resulting in 24 images. As a demonstration of the detailed visualization process, category 1 is selected as a representative example, such as... Figure 3 As shown.

[0136] Step 2.2: Assume Category 1 contains β residential ECPs, each paired with its corresponding 24-hour dynamic pricing data, resulting in a total of 24β data points (p, q). These data points are then plotted as hourly grayscale images. Thus, for each category, a set of 24-hour images is obtained, which serves as input data for the neural network model in Step 4.

[0137] Step 3: Generation of training samples based on multi-information fusion. To develop a well-trained model, a large number of training samples are necessary. However, obtaining diverse training samples from historical data, i.e., image pairs (I... x ,I DC This is challenging. Furthermore, extracting real-world data from the collected data is not feasible.

[0138] Therefore, this embodiment proposes a data generation function based on multi-information fusion, used to generate data from a selected demand curve f. DC A large number of data points (p, q) are generated around the curve, which is then treated as real data. These data are then visualized and combined into an image pair, which serves as the training sample. The process for generating the training sample is as follows: Figure 4 As shown.

[0139] Step 3.1: From Figure 4 It can be seen that the first stage is to establish a set of demand curve functions exhibiting a monotonically decreasing shape curvature based on prior knowledge. To ensure simplicity and broad applicability, logarithmic demand functions and exponential demand functions were selected to create the real dataset. These functions are expressed as formula (16):

[0140] q log (p)=a log +b log ln(p)

[0141] q exp (p)=a exp exp(b exp p)

[0142] Where, parameter a log ,b log ,a exp ,b exp The settings were determined through preliminary experiments, aiming to generate f. DC A balance between stability and diversity.

[0143] Step 3.2: Data Point Generation. The second stage involves randomly selecting a demand curve function and generating data points around that curve, as shown in formula (17):

[0144] (p,q)=g ζ (f DC )

[0145] Among them, f DC g represents the actual data function describing the demand curve. ζ Let (p,q) represent the data generation function, where (p,q) represents the composite pair of electricity consumption and price.

[0146] Step 3.3: Generate function g based on the data. ζ It can be based on various predefined f DC By synthesizing sufficiently rich training samples, the shortcomings of traditional methods that rely on historical data are overcome. To observe and analyze the inherent graphical features of real-world collected data points, g... ζ Defined as formula (18):

[0147] Normal Data=(x,δ*derivative(f DC (x)+f DC (x))

[0148] Sparse Outliers = (a, b)

[0149]

[0150] Where δ represents the deviation of normal data from f DC It is proportional to the first derivative of (x).

[0151] Step 3.4: By repeatedly executing the above procedure, a large number of training samples can be obtained. Considering the time cost of model training, a total of 12,000 image pairs were generated, which is considered sufficient to cover various scenarios within the scope of this study.

[0152] Step 4: Evaluation model of PED based on Attention U-Net. This embodiment designs a data-driven deep learning model that performs the inverse process of data generation. This model is able to extract the real demand curve f from simulated data points. DC Furthermore, it can be extended to real-world data consisting of electricity demand and price. The simplified form of the model is shown in equation (19):

[0153]

[0154] Step 4.1: To incorporate the local spatial relationships between data points (p, q) in image visualization, a machine vision-based modeling method was selected. The objective function of this method is shown in equation (20):

[0155]

[0156] in, It is the inverse function, with weight coefficients ω; L is the loss function; I x Visualization of data points (p, q); I DC It is the demand curve f DC Image visualization.

[0157] Step 4.2: This embodiment proposes a framework based on the Attention U-net mechanism to solve the above modeling task. Attention U-net is a convolutional neural network widely used in medical image analysis. It selectively focuses on the regions containing the most information in the input image by using attention gates. Attention U-net has a significant ability to preserve low-level features and capture high-level features, making it a suitable choice for modeling inverse functions. Figure 5 The Attention U-Net architecture used in this study is shown, consisting of four basic layers: convolutional layers, max-pooling layers, upsampling layers, and an output layer. The numbers shown above the figure indicate the configuration of the feature maps.

[0158] From Figure 5As can be seen, the Attention U-Net architecture consists of two parts. In the encoder, the feature size is reduced from 256×256 to 32×32 through four downsampling operations. In the decoder, the original image resolution is restored through four upsampling operations. During each training cycle, each convolutional layer undergoes three consecutive operations (3×3 convolution, normalization, and ReLU activation). Max pooling layers are downsampled using 2×2 pooling operations, while upper convolutional layers are upsampled using 2×2 transposed convolution operations. The output layer maps the feature map to the RGB image space through a single convolution operation. Skip connections allow concatenating low-level feature maps with the upsampled feature maps, thereby recovering the spatial information lost during the downsampling operation. The gating signal extracted from the encoder feature map selectively emphasizes regions in the feature map based on its relevance to the task.

[0159] Step 4.3: The mean squared error (MSE) is used as the loss function, as shown in formula (21):

[0160]

[0161] After model training, the Attention U-net-based model can accurately extract information from I... x Extract I DC The grayscale image obtained in the third part is input into the trained model to generate an RGB image I. DC In order to extract from image I DC Derive the demand curve f DC A polynomial was proposed, and a pixel mapping and polynomial fitting module was put forward.

[0162] It needs to be clarified that the coordinates of the normalized historical data consist of two floating-point numbers, representing the normalized price and normalized electricity consumption, respectively. The pixel coordinates, on the other hand, are represented by two integers, indicating the pixel's relative position in the image, which are then mapped to the coordinates of the normalized data.

[0163] The output RGB image is first converted to a 256×256 grayscale image. Considering that the pixel coordinate range of an image in Python is 32 to 229, there are a total of 198 coordinate pairs (x, y, y) that can be used to map to the data coordinate system. pixel y pixel It is worth noting that the grayscale value of each pixel ranges from 0 to 255, where 0 represents black and 255 represents white.

[0164] Step 4.4: For each column of fixed x-coordinates, identify the pixel with the smallest grayscale value and assign it as the y-coordinate. Then, map this selected pixel to the coordinates of the historical data. The complete pixel mapping process is represented by formula (22):

[0165]

[0166] Where l represents the integer position of the pixel in the column, (x datal ,y datal ) represents the mapped data coordinate pairs.

[0167] Step 4.5: After completing the pixel mapping process, a total of 198 normalized data points are obtained, which are used to derive f. DC This process is accomplished through a polynomial fitting formula (23):

[0168]

[0169] Where λ represents the penalty term; a i This represents the weighting coefficient. Through this process, the polynomial representation of the demand curve is obtained.

[0170] Step 4.6: By repeating the above operations, the 24-hour grayscale image is processed and a set of polynomials containing the hourly demand curve fDC for each category is constructed, as shown in (24):

[0171] F k ={f 1,k ,f 2,k ,...,f 24,k}, k=1,2,...,K

[0172] Step 5: Distributed Flexible Load Modeling. To mitigate the impact of seasonal and date variations on residential ECPs, the historical data used for testing in this embodiment will first be divided into weekdays and rest days according to each season. It is assumed that weather factors are reflected in consumer electricity consumption behavior. To assess the electricity consumption of a specific consumer on a particular weekday d, daily ECPs and corresponding retail prices from weekdays d-1 to dD will be collected as test data.

[0173] Step 5.1: The principle of the dedicated encapsulation model and how to evaluate the daily electricity consumption of a single end user are as follows. Within the retailer's jurisdiction, numerous residents are able to adjust their electricity demand under a given dynamic electricity price. By utilizing the proposed ECP classification method and the PED evaluation model based on multi-information fusion, a set of polynomials delineating the 24-hour demand curves for each category is obtained. Combining the percentage of ECP distribution for a specific consumer in each category, an encapsulation model is designed to evaluate the electricity consumption on day d, as shown in Equation (25), where the input is a 24-hour price vector:

[0174]

[0175] Where K represents the number of ECP categories to which a particular consumer belongs; D k Indicates category C k The number of ECPs in F; k It is category C k The set of polynomials.

[0176] Step 5.2: Recalculation and Update. If the assessment date changes, it is necessary to reclassify residential user electricity consumption behavior, assess the electricity demand price elasticity for each category, and model distributed flexible loads. Since the trained deep learning model does not require retraining, it does not incur additional computational burden.

[0177] like Figures 6-8 The image shows the clustering results of the PC&M classification method, where the vertical axis represents the user's power consumption at 24 hours, which is the user's power consumption pattern / behavior.

[0178] like Figure 9 The results of the distributed flexible load assessment model over five consecutive working days are shown. The vertical axis represents the actual power consumption of users at 24 hours. GRU, LSTM, and U-net represent comparative models, and Proposed represents the model proposed in this embodiment.

[0179] Example 3

[0180] This embodiment provides a distributed flexible load modeling system for distribution network reliability assessment, including:

[0181] The electricity consumption behavior classification module is configured to classify the electricity consumption behavior of residential users based on parallel computing and merging classification methods, and convert the classification results into grayscale images;

[0182] The training sample generation module is configured to construct a set of demand curve functions based on multi-information fusion and generate image training samples;

[0183] The evaluation model building module is configured to take image training samples as input to train a nonlinear demand price elasticity evaluation model based on deep learning methods; input grayscale images into the trained model, extract the polynomial representation of the demand curve, and output a set of polynomials of the demand price elasticity curve.

[0184] The load model building module is configured to construct a distributed flexible load model based on historical data and a set of polynomials of demand price elasticity curves to evaluate the electricity consumption of a single user at a given electricity price, serving as input parameters for distribution network reliability assessment.

[0185] Preferably, the electricity consumption behavior classification module includes:

[0186] The data preprocessing unit is configured to normalize the collected residential electricity consumption behavior data and use principal component analysis algorithm to reduce dimensionality, and distribute the processed data evenly to each data block;

[0187] Parallel computing units are configured to perform electricity consumption behavior sample calculations in each data block, using three indicators—cumulative proximity, eccentricity, and standardized eccentricity—for recursive calculations, and determining the category of new electricity consumption behavior samples using Chebyshev's inequality.

[0188] The category merging unit is configured to merge categories collected from various data blocks by finding the nearest neighbor category and merging when the merging conditions are met, and updating the parameters of the merged category.

[0189] The data integration unit is configured to pair residential electricity consumption in each category with the corresponding 24-hour electricity price data to form data points;

[0190] The image conversion unit is configured to convert hourly data points into grayscale images, generating 24 images for each category.

[0191] Preferably, the training sample generation module includes:

[0192] The demand curve function set building unit is configured to select logarithmic demand function and exponential demand function as the basis, and adjust the parameters through experiments to balance stability and diversity.

[0193] The data point generation unit is configured to randomly select the demand curve function to generate data points, which constitute image training samples.

[0194] The image training sample generation unit is configured to synthesize a large number of data points based on the demand curve function set and the data generation function for model training.

[0195] Preferably, the evaluation model building module includes:

[0196] The model training unit is configured to train a nonlinear demand price elasticity assessment model based on an attention U-Net network architecture using generated image training samples, and to learn the intrinsic relationship between electricity price and electricity consumption.

[0197] The pixel mapping and polynomial fitting unit is configured to extract a polynomial representation of the demand curve from the input grayscale image through the trained model.

[0198] The price elasticity assessment unit is configured to obtain a set of polynomials for the demand price elasticity curves through pixel mapping and polynomial fitting.

[0199] Preferably, the load model construction module includes:

[0200] The data segmentation unit is configured to divide the test data into weekdays and rest days, and to reflect weather factors in residents' electricity consumption behavior, collecting historical data for preset weekdays, including electricity prices and corresponding user electricity consumption.

[0201] The model building unit is configured to use the obtained electricity consumption behavior classification results and the nonlinear demand price elasticity assessment model to establish a distributed flexible load model.

[0202] The recalculation and update unit is configured to reclassify residential user electricity consumption behavior, assess electricity demand price elasticity, and model distributed flexible loads when the assessment date changes.

[0203] Example 4

[0204] This embodiment is based on embodiment 1:

[0205] This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the distributed flexible load modeling method for distribution network reliability assessment described in Embodiment 1. The computer program can be in the form of source code, object code, executable file, or some intermediate form.

[0206] Example 5

[0207] This embodiment is based on embodiment 1:

[0208] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the distributed flexible load modeling method for distribution network reliability assessment described in Embodiment 1. The computer program can be in the form of source code, object code, executable file, or some intermediate form. The storage medium includes any entity or device capable of carrying computer program code, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the storage medium can be appropriately added to or subtracted according to the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the storage medium does not include electrical carrier signals and telecommunication signals.

[0209] It should be noted that, for the sake of simplicity, the foregoing method embodiments are described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

Claims

1. A distributed flexible load modeling method for distribution network reliability assessment, characterized in that, include: The electricity consumption behavior of residential users is classified based on parallel computing and merging classification methods, and the classification results are converted into grayscale images. A set of demand curve functions is constructed based on multi-information fusion, and image training samples are generated. The image training samples are used as input to train a nonlinear demand price elasticity assessment model based on deep learning methods; the grayscale image is input into the trained model to extract the polynomial representation of the demand curve and output a set of polynomials of the demand price elasticity curve. Based on historical data and the set of polynomials of demand price elasticity curves, a distributed flexible load model is constructed to evaluate the electricity consumption of a single user under a given electricity price, serving as an input parameter for distribution network reliability assessment. The method of classifying residential users' electricity consumption behavior based on parallel computing and merging classification, and converting the classification results into grayscale images, includes: Data preprocessing: The collected residential electricity consumption data is normalized and dimensionality is reduced using principal component analysis algorithm. The processed data is then evenly distributed to each data block. Parallel computing: Electricity consumption behavior samples are calculated separately in each data block. The cumulative proximity, eccentricity and standardized eccentricity are used for recursive calculation. The category of new electricity consumption behavior samples is determined by Chebyshev's inequality. Category merging: Merge the categories collected from each data block by finding the nearest neighbor category and merging when the merging conditions are met, and update the parameters of the merged category; Data integration: Pair residential electricity consumption in each category with the corresponding 24-hour electricity price data to form data points; Image conversion: Convert hourly data points into grayscale images, generating 24 images for each category; Using the image training samples as input, a nonlinear demand price elasticity assessment model based on deep learning methods is trained; the grayscale image is input into the trained model, the polynomial representation of the demand curve is extracted, and a set of polynomials for the demand price elasticity curve is output, including: Model training: The nonlinear demand price elasticity assessment model based on the attention U-Net network architecture is trained using generated image training samples to learn the intrinsic relationship between electricity price and electricity consumption; Pixel mapping and polynomial fitting: Extracting a polynomial representation of the demand curve from the input grayscale image using a trained model; Price elasticity assessment: The set of polynomial curves for demand price elasticity is obtained by pixel mapping and polynomial fitting.

2. The distributed flexible load modeling method for distribution network reliability assessment according to claim 1, characterized in that, The process of constructing a set of demand curve functions based on multi-information fusion and generating image training samples includes: Construction of the demand curve function set: Logarithmic demand function and exponential demand function are selected as the basis, and parameters are adjusted through experiments to balance stability and diversity; Data point generation: Randomly select the demand curve function to generate data points to form image training samples; Image training sample generation: Based on the set of demand curve functions and data generation functions, a large number of data points are synthesized for model training.

3. The distributed flexible load modeling method for distribution network reliability assessment according to claim 1, characterized in that, The distributed flexible load model, constructed based on historical data and the set of polynomials for demand price elasticity curves, includes: Data segmentation: The test data is segmented into weekdays and rest days, and weather factors are reflected in residents' electricity consumption behavior. Historical data for preset weekdays is collected, including electricity prices and corresponding user electricity consumption. Model Construction: A distributed flexible load model is established using the obtained electricity consumption behavior classification results and the nonlinear demand price elasticity assessment model; Recalculation and Update: If the assessment date changes, the classification of residential user electricity consumption behavior, the assessment of electricity demand price elasticity, and the modeling of distributed flexible loads will be re-performed.

4. A distributed flexible load modeling system for distribution network reliability assessment, characterized in that, include: The electricity consumption behavior classification module is configured to classify the electricity consumption behavior of residential users based on parallel computing and merging classification methods, and convert the classification results into grayscale images; The training sample generation module is configured to construct a set of demand curve functions based on multi-information fusion and generate image training samples; The evaluation model building module is configured to take the image training samples as input to train a nonlinear demand price elasticity evaluation model based on deep learning methods; input the grayscale image into the trained model, extract the polynomial representation of the demand curve, and output a set of polynomials of the demand price elasticity curve. The load model building module is configured to construct a distributed flexible load model based on historical data and the set of polynomials of the demand price elasticity curves to evaluate the electricity consumption of a single user at a given electricity price, as an input parameter for the reliability assessment of the distribution network. The electricity consumption behavior classification module includes: The data preprocessing unit is configured to normalize the collected residential electricity consumption behavior data and use principal component analysis algorithm to reduce dimensionality, and distribute the processed data evenly to each data block; Parallel computing units are configured to perform electricity consumption behavior sample calculations in each data block, using three indicators—cumulative proximity, eccentricity, and standardized eccentricity—for recursive calculations, and determining the category of new electricity consumption behavior samples using Chebyshev's inequality. The category merging unit is configured to merge categories collected from various data blocks by finding the nearest neighbor category and merging when the merging conditions are met, and updating the parameters of the merged category. The data integration unit is configured to pair residential electricity consumption in each category with the corresponding 24-hour electricity price data to form data points; The image conversion unit is configured to convert hourly data points into grayscale images, generating 24 images for each category; The evaluation model construction module includes: The model training unit is configured to train a nonlinear demand price elasticity assessment model based on an attention U-Net network architecture using generated image training samples, and to learn the intrinsic relationship between electricity price and electricity consumption. The pixel mapping and polynomial fitting unit is configured to extract a polynomial representation of the demand curve from the input grayscale image through the trained model. The price elasticity assessment unit is configured to obtain a set of polynomials for the demand price elasticity curves through pixel mapping and polynomial fitting.

5. A distributed flexible load modeling system for distribution network reliability assessment according to claim 4, characterized in that, The training sample generation module includes: The demand curve function set building unit is configured to select logarithmic demand function and exponential demand function as the basis, and adjust the parameters through experiments to balance stability and diversity. The data point generation unit is configured to randomly select the demand curve function to generate data points, which constitute image training samples. The image training sample generation unit is configured to synthesize a large number of data points based on the demand curve function set and the data generation function for model training.

6. A distributed flexible load modeling system for distribution network reliability assessment according to claim 4, characterized in that, The load model construction module includes: The data segmentation unit is configured to divide the test data into weekdays and rest days, and to reflect weather factors in residents' electricity consumption behavior, and to collect historical data for preset weekdays, including electricity prices and corresponding user electricity consumption. The model building unit is configured to use the obtained electricity consumption behavior classification results and the nonlinear demand price elasticity assessment model to establish a distributed flexible load model. The recalculation and update unit is configured to reclassify residential user electricity consumption behavior, assess electricity demand price elasticity, and model distributed flexible loads when the assessment date changes.

Images