New user load forecasting model construction method, device, equipment and storage medium

Abstract

The present disclosure relates to the field of machine learning technology and provides a method, apparatus, device, and storage medium for constructing a new user load forecasting model. The method comprises: obtaining sample data of typical users among non-new users, wherein the typical users and the new users belong to the same industry category; using the sample data to train and test a preset non-new user load forecasting model until the load forecasting model meets the preset convergence conditions, saving the model parameters of the load forecasting model; and using the model parameters to construct a new user load forecasting model. The present disclosure can effectively solve the problem that due to the lack of historical data on new users, it is difficult to apply existing prediction technologies based on deep neural networks for load forecasting, resulting in low accuracy of load forecasting results.

Description

Technical Field

[0001] The present disclosure relates to the field of machine learning technology, and in particular to a method, apparatus, device, and storage medium for constructing a new user load prediction model. Background Art

[0002] User load forecasting (hereinafter referred to as "user load forecasting") involves studying existing user electricity usage data to predict future user electricity usage behavior. With the deepening reform of my country's power market and the continued opening up of electricity sales services in recent years, user load forecasting has become a crucial step for power sales companies to reduce performance deviations and enhance profitability.

[0003] With the rapid development of artificial intelligence technology in recent years, load forecasting technology based on deep neural networks has been widely researched and applied. Compared with traditional linear regression and trend extrapolation methods, although deep neural network-based forecasting technology can consider the impact of a wider range of external factors on user electricity consumption behavior, thereby achieving better forecasting results, it requires a large amount of user historical data for model training, otherwise it will not achieve the desired results. However, for some new users of power sales companies, due to the limited amount of data accumulated on new users in the early stages of access, it is impossible to use deep neural network-based forecasting technology for load forecasting. As a result, the load forecast accuracy for these new users remains low, posing a significant risk to the power sales companies in formulating trading strategies.

[0004] It can be seen that there is little historical data for new users, and it is difficult to apply the existing prediction technology based on deep neural networks for load forecasting, so the accuracy of the load forecast results is low. Summary of the Invention

[0005] In view of this, the embodiments of the present disclosure provide a method, device, electronic device and computer-readable storage medium for constructing a new user load forecasting model to solve the problem in the prior art that due to the lack of historical data on new users, it is difficult to apply existing deep neural network-based forecasting technology for load forecasting, resulting in low accuracy of load forecasting results.

[0006] A first aspect of an embodiment of the present disclosure provides a method for constructing a new user load forecasting model, comprising:

[0007] Obtain sample data of typical users among non-new users, where the typical users and new users belong to the same industry category;

[0008] Using sample data to train and test a preset non-new user load forecasting model until the load forecasting model meets a preset convergence condition, and then saving the model parameters of the load forecasting model;

[0009] Use the model parameters to build a new user load forecasting model.

[0010] A second aspect of the embodiments of the present disclosure provides a device for constructing a new user load prediction model, comprising:

[0011] A data acquisition module is configured to acquire sample data of typical users among non-new users, wherein the typical users and the new users belong to the same industry category;

[0012] a parameter saving module configured to train and test a preset non-new user load forecasting model using sample data until the load forecasting model meets a preset convergence condition, and then save the model parameters of the load forecasting model;

[0013] The model building module is configured to build a new user load forecasting model using the model parameters.

[0014] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0015] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above method are implemented.

[0016] Compared with the prior art, the beneficial effects of the embodiments of the present disclosure include at least: obtaining sample data of typical users among non-new users, wherein the typical users and the new users belong to the same industry category; using the sample data to train and test a preset non-new user load forecasting model until the load forecasting model meets the preset convergence conditions, and saving the model parameters of the load forecasting model; using the model parameters to construct a new user load forecasting model, which can well solve the problem that due to the lack of historical data on new users, it is difficult to apply the existing deep neural network-based forecasting technology for load forecasting, resulting in low accuracy of the load forecasting results. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 is a schematic diagram of an application scenario of an embodiment of the present disclosure;

[0019] Figure 2This is a flow chart of a method for constructing a new user load forecasting model provided by an embodiment of the present disclosure;

[0020] Figure 3 This is a schematic diagram of the network structure of a deep learning model in a new user load forecasting model construction method provided by an embodiment of the present disclosure;

[0021] Figure 4 This is a schematic structural diagram of a new user load forecasting model construction device provided by an embodiment of the present disclosure;

[0022] Figure 5 It is a structural diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0023] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present disclosure. However, it will be apparent to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present disclosure with unnecessary detail.

[0024] A method and apparatus for constructing a new user load prediction model according to an embodiment of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0025] Figure 1 10 is a schematic diagram of an application scenario of an embodiment of the present disclosure. The application scenario may include a terminal device 101, a server 102, and a network 103.

[0026] The terminal device 101 can be hardware or software. When the terminal device 101 is hardware, it can be various electronic devices with a display screen and supporting communication with the server 102, including but not limited to smartphones, tablet computers, laptop computers and desktop computers; when the terminal device 101 is software, it can be installed in the above electronic devices. The terminal device 101 can be implemented as multiple software or software modules, or as a single software or software module, and the embodiments of the present disclosure are not limited to this. Furthermore, various applications can be installed on the terminal device 101, such as data processing applications, instant messaging tools, social platform software, search applications, shopping applications, etc.

[0027] Server 102 may be a server that provides various services, such as a backend server that receives requests sent by terminal devices that establish communication connections with it. The backend server may receive and analyze the requests sent by the terminal devices, and generate processing results. Server 102 may be a single server, a server cluster consisting of multiple servers, or a cloud computing service center, all of which are not limited in the present disclosure.

[0028] It should be noted that the server 102 can be either hardware or software. When the server 102 is hardware, it can be various electronic devices that provide various services to the terminal device 101. When the server 102 is software, it can be multiple software or software modules that provide various services to the terminal device 101, or it can be a single software or software module that provides various services to the terminal device 101, and this is not limited in the present embodiment.

[0029] The network 103 may be a wired network connected by coaxial cables, twisted pairs, and optical fibers, or a wireless network that can interconnect various communication devices without wiring, such as Bluetooth, NFC, infrared, etc. The embodiments of the present disclosure do not limit this.

[0030] The user can establish a communication connection with the server 102 via the terminal device 101 via the network 103 to receive or send information, etc. Specifically, the server 102 obtains sample data of typical users among non-new users belonging to the same industry category as the new user, and then uses the sample data of the typical users to train and test the preset non-new user load forecasting model until the load forecasting model meets the preset convergence conditions. The model parameters of the load forecasting model are then saved; finally, the model parameters are used to construct a new user load forecasting model. When there is less historical data accumulated for new users, it is applicable to use a prediction model based on a deep neural network for load forecasting, thereby improving the accuracy of load forecasting for new users.

[0031] It should be noted that the specific types, quantities and combinations of the terminal devices 101, the server 102 and the network 103 can be adjusted according to the actual needs of the application scenario, and the embodiments of the present disclosure do not limit this.

[0032] Accurate user load forecasting is crucial for power retail companies. With the development of the electricity spot market, market price volatility will increase. In both day-ahead and mid-day quotes, power retail companies must adjust their declared power consumption based on user load forecasts to avoid biased assessments and maximize profit margins.

[0033] For new power users, after signing the agency contract, the power sales company typically collects load data by installing new electricity meters. Historical data is often unavailable. For these users, power sales companies typically use methods such as linear regression and recent averages for initial forecasting. Once a sufficient amount of user data has accumulated, they employ deep neural network-based forecasting.

[0034] However, numerous factors influence user power load, including the user's own production plans, local weather conditions, holidays, and national policies. Forecasting methods such as linear regression and recent averages simply extrapolate the power load data series and fail to account for the combined impact of multiple factors on user load, resulting in suboptimal forecasting accuracy. Forecasting techniques based on deep neural networks can comprehensively consider the combined impact of multiple factors on user load, significantly improving forecasting accuracy over extrapolation methods. However, these methods require extensive historical data to train the models, making them inappropriate for newly connected users.

[0035] Based on the above, the present disclosure proposes a method for constructing a new user load prediction model. The model constructed by this method can be used for electricity load prediction of new users with insufficient historical data, and the prediction accuracy is high. It well solves the problem in the existing technology that due to insufficient historical data of new users, it is impossible to use deep neural networks for electricity load prediction, and the prediction results are low in accuracy.

[0036] Figure 2 This is a flow chart of a method for constructing a new user load forecasting model provided by an embodiment of the present disclosure. Figure 2 The new user load forecasting model construction method can be obtained by Figure 1 The server 102 executes. Figure 2 As shown, the new user load forecasting model construction method includes:

[0037] Step S201 : obtaining sample data of typical users among non-new users, wherein the typical users and the new users belong to the same industry category.

[0038] As an example, the customers represented by a power sales company include users who have just been connected to the power sales company for a short time (i.e., new users) and users who have been using the power sales company for a long time (i.e., non-new users). For example, users who just connected to the power sales company last week can be considered new users, and users who have been connected to the power sales company for 2 years can be considered non-new users. In other words, the distinction between new and non-new users mainly depends on how long they have been connected to the power sales company. Generally, users who have been connected to the power sales company for a longer time (non-new users) have a relatively large amount of historical data accumulation on their electricity load, while users who have been connected to the power sales company for a shorter time (new users) have a relatively small amount of historical data accumulation on their electricity load. In actual applications, users with more than 300 historical data can generally be determined as non-new users, and users with less than 300 historical data can be determined as new users. Usually, a piece of data can refer to the electricity consumption for one day, or it can be the electricity consumption for a week, etc.

[0039] Typical users usually refer to users with a large amount of historical data among the above non-new users, for example, users with more than 500 historical data items; or users with the most outstanding electricity consumption or electricity consumption time among the above non-new users.

[0040] Sample data includes the user's energy-consuming equipment, the amount of electricity drawn from the power system or power plant by the equipment, the time of draw or the time period of draw, etc. For example, user A's energy meter draws Y kilowatt-hours of electricity from the power system or power plant from 0:00 on January 1, 20XX to 24:00 on January 1, 20XX.

[0041] Industry categories include but are not limited to petrochemicals, chemicals, building materials, steel, nonferrous metals, papermaking, electricity, aviation, etc.

[0042] As an example, sample data of typical users among non-new users is obtained. Specifically, the user can upload sample data such as energy consumption (such as electricity consumption, water consumption, natural gas consumption, etc.), energy consumption time (such as electricity consumption time, water consumption time, natural gas consumption time, etc.) to the server 102 through the terminal device 101 (for example, a smart phone, a personal computer, etc.); or the user's energy consumption data (including energy consumption and energy consumption time, etc.) can be automatically uploaded to the server 102 in real time through the electricity meters installed in each household.

[0043] Step S202: Using sample data to train and test a preset non-new user load forecasting model until the load forecasting model meets a preset convergence condition, and then saving the model parameters of the load forecasting model.

[0044] The preset non-new user load forecasting model is a deep neural network-based load forecasting model. Specifically, the loads can be electricity, water, or gas. In practical applications, corresponding non-new user load forecasting models can be constructed based on the needs of the specific scenario. For example, for electricity load forecasting, a non-new user electricity load forecasting model can be constructed, while for water load forecasting, a non-new user water load forecasting model can be constructed.

[0045] Satisfying the preset convergence conditions, usually the loss value that satisfies the preset loss function.

[0046] Model parameters usually refer to the weights and biases of the load forecasting model.

[0047] Step S203: construct a new user load prediction model using the model parameters.

[0048] The technical solution provided by the embodiments of the present disclosure obtains sample data of typical users among non-new users, wherein the typical users and the new users belong to the same industry category; uses the sample data to train and test a preset non-new user load forecasting model until the load forecasting model meets the preset convergence conditions, and saves the model parameters of the load forecasting model; uses the model parameters to construct a new user load forecasting model, which can effectively solve the problem that due to the lack of historical data on new users, it is difficult to apply existing deep neural network-based forecasting technology for load forecasting, resulting in low accuracy of load forecasting results.

[0049] In some embodiments, the above step S201 includes:

[0050] M non-new users whose sample data volume meets a preset data volume range are selected from the proxy users, and the M non-new users are classified to obtain N non-new user sets, where each non-new user set includes at least one non-new user, N≤M, and both N and M are positive integers;

[0051] Obtain sample data of typical users from the set of non-new users who belong to the same industry category as the new user.

[0052] The preset data volume range can be flexibly set according to actual conditions. For example, it can be greater than 300 items or greater than 400 items, etc., and is not specifically limited here.

[0053] As an example, the following detailed description uses users (including non-new and new users) represented by a power sales company. First, the users represented by the power sales company are divided into new and non-new users. Then, a sample data volume of non-new users is obtained, and M non-new users with a sample data volume greater than a preset data volume (e.g., greater than 300) are selected. Cluster analysis is then performed on these M non-new users using the K-Means clustering algorithm (k-means clustering algorithm), resulting in N non-new user sets, one of which represents an industry category. Assume that six non-new users are screened out through the above steps, respectively denoted as non-new users 01, 02, 03, 04, 05, and 06. After cluster analysis, these six non-new users are divided into three non-new user sets, denoted as set a (chemical industry), set b (petrochemical industry), and set c (building materials industry). Set a includes non-new users 01 and 02, set b includes non-new users 03, 05, and 06, and set c includes non-new user 04. The typical user in set a is non-new user 02, the typical user in set b is non-new user 06, and the typical user in set c is non-new user 04. Assuming that the new user belongs to the chemical industry, sample data of non-new user 02 is obtained.

[0054] Among them, the industry category classification of M non-new users can also be performed using methods such as the mean shift clustering algorithm, the density-based clustering method (DBSCAN), and the Gaussian mixture model (GMMs).

[0055] In some embodiments, obtaining sample data of typical users from a set of non-new users belonging to the same industry category as the new user specifically includes:

[0056] Filter out a set of typical non-new users of the same type as the new user and belonging to the same industry category from N non-new user sets;

[0057] Calculate the distance between each non-new user in the typical non-new set of the same type and the cluster center point of the typical non-new set of the same type;

[0058] The non-new user with the smallest distance to the cluster center is determined as the typical user of the same type of typical non-new set, and sample data of the typical user is obtained.

[0059] Based on the above example, after classifying the six non-new users, we can obtain three non-new user sets with industry category labels and three cluster centers. The new user belongs to the chemical industry, and set a also belongs to the chemical industry. That is, the new user and set a belong to the same industry category. Set a is a set of typical non-new users of the same industry category as the new user. Calculate the Euclidean distance d1 between non-new user 01 in set a and its cluster center, and the Euclidean distance d2 between non-new user 02 and its cluster center. Where d1>d2, non-new user 02 with the smallest Euclidean distance value is determined as the typical user of the typical non-new set of the same type.

[0060] The calculation of the above-mentioned Euclidean distance is a mature existing technology and will not be described in detail here.

[0061] In some embodiments, a typical non-new user set of the same type as the new user and belonging to the same industry category is screened out from the N non-new user sets, specifically including:

[0062] Determine the non-new typical users in each non-new user set to obtain N non-new typical users;

[0063] Acquire first sample data of N non-new typical users, and generate a first load curve for each non-new typical user based on the first sample data;

[0064] Acquire second sample data of the new user, and generate a second load curve of the new user based on the second sample data;

[0065] According to the correlation between the second load curve and each first load curve, a set of similar typical non-new users belonging to the same industry category as the new user is screened out.

[0066] In combination with the above example, according to the above steps, the Euclidean distance between the non-new users in each non-new user set and the cluster center in the set can be calculated, and the non-new user with the smallest Euclidean distance is determined as the atypical user of the set.

[0067] Exemplarily, according to the above steps, when it is determined that the typical user in set a is non-new user 02, the typical user in set b is non-new user 06, and the typical user in set c is non-new user 04, the first sample data of non-new user 02, non-new user 06, and non-new user 04 are obtained respectively. The first sample data includes the energy-consuming devices, energy consumption time, and energy consumption of non-new users. For example, the first sample data of non-new user 02 includes the electrical devices (such as household appliances, etc.) of non-new user 02, the electricity consumption time (such as February 1, 20XX to March 5, 20XX, etc.), and the electricity consumption (such as XX kWh). Then, based on the above first sample data, a first load curve is generated, that is, a change curve graph with time as the horizontal axis and electricity consumption as the vertical axis is generated.

[0068] Similarly, the first load curves of non-new user 06 and non-new user 04, and the second load curve of the new user can be obtained by referring to the generation steps of the first load curve of non-new user 02, which will not be repeated here.

[0069] The second sample data includes the new user's energy-consuming equipment, energy-consuming time, and energy consumption.

[0070] As another example, the industry category of the new user can also be determined manually, for example, by the staff of the power sales company who are responsible for distinguishing the industry category of the new user, or the new user can fill in and report the industry category himself, and the industry category of the new user will be marked when importing data.

[0071] In some embodiments, based on the correlation between the second load curve and each first load curve, a set of typical non-new users of the same type as the new user and belonging to the same industry category is screened out, specifically including:

[0072] calculating the Pearson coefficient of the second load curve and each first load curve respectively;

[0073] Based on the Pearson coefficient, a set of typical non-new users of the same type as new users who belong to the same industry category is screened out.

[0074] Among them, the Pearson coefficient is widely used to measure the degree of correlation between two variables, and its value is between -1 and 1. Specifically, the Pearson correlation coefficient between two variables is defined as the quotient of the covariance and standard deviation between the two variables.

[0075] Based on the above example, calculate the Pearson coefficient P1 between the first load curve of non-new user 02 and the second load curve of new users; the Pearson coefficient P2 between the first load curve of non-new user 04 and the second load curve of new users; and the Pearson coefficient P3 between the first load curve of non-new user 06 and the second load curve of new users. Compare P1, P2, and P3, and determine the set of non-new users with the maximum value as the set of typical non-new users belonging to the same industry category as the new user.

[0076] Assuming that the above comparison result is P1>P2>P3, then the set a where the non-new user 02 is located is determined to be a set of typical non-new users of the same type as the new user who belong to the same industry category.

[0077] In some embodiments, the above step S202 includes:

[0078] Divide the sample data into training data set and test data set;

[0079] Use the training data set to train the preset deep learning model to obtain the load forecasting model;

[0080] Use the test data set to evaluate the load forecasting model and obtain the evaluation test results;

[0081] When the evaluation test results meet the preset convergence conditions, the model parameters of the load forecasting model are saved.

[0082] As an example, the network structure of a preset deep learning model usually consists of an input layer, multiple hidden layers, and an output layer. Its network structure can be abstracted as Figure 3 The structure shown. Neurons are connected together in a fully connected manner. The data transfer formula between layers is shown in the following formula (1):

[0083] a l =f l (W l a l-1 +b l ) (1).

[0084] Among them, a l represents the output of the neurons in the lth layer; f l Represents the activation function of the lth layer of neurons, l = 1, 2, ···, n, and the input x is represented by a 0 , W l is the weight matrix connecting the neurons in the l-1 layer and the l layer, b l is the bias term of the lth layer of neurons. Among them, the output of the last layer of neurons in the nth layer is the final output expression of the neural network, as shown in the following formula (2):

[0085] Y t =f n (f n-1 (...f l (x)) (2).

[0086] Among them, the function f l (x) is the activation function of the lth layer. In the hidden layer, the embodiment of the present disclosure selects LReLU (Leaky Rectified Linear Unit) as the activation function, which is shown in the following formula (3):

[0087] f(x)=max(0.01x,x) (3).

[0088] In the output layer, the linear function is selected as the activation function, which can greatly reduce the amount of calculation. Its expression is shown in the following formula (4):

[0089] f(x)=x (4).

[0090] Typically, deep neural network training uses labeled samples to adjust parameters θ = {W, b}, where W is the weight and b is the bias, to minimize the loss function L. The loss function used in the embodiment of the present disclosure is the mean square error loss function shown in the following formula (5).

[0091]

[0092] Where m is the number of samples, is the output value of the deep neural network, and y is the load of the sample on that day.

[0093] As an example, the root mean square propagation (RMSProp) algorithm can be used to train a deep neural network (a preset deep learning model) to obtain the optimal parameter θ, i.e., the model parameter.

[0094] As an example, combined with the above example, the historical sample data of non-new user 02 can be obtained first, and then the historical sample data can be divided into a training data set and a test data set. Then, the training data set is used to train the above-mentioned preset deep learning model. After the training is completed, the load forecasting model is output, and then the test data set is used to evaluate the load forecasting model to obtain the evaluation test results. When the evaluation test results reach the minimum value of the above-mentioned mean square error loss function, it meets the preset convergence conditions, and the model parameters θ of the load forecasting model, that is, the weight W and bias b, are saved.

[0095] If the evaluation test result does not reach the minimum value of the above-mentioned mean square error loss function, the model parameters of the load forecasting model can be adjusted manually or automatically until the preset convergence conditions are met.

[0096] In some embodiments, the above step S203 includes:

[0097] Construct a pre-trained model with the same network structure as the non-new user load forecasting model;

[0098] The model parameters are transferred to the pre-trained model to obtain a new user load forecasting model.

[0099] As an example, first, a pre-trained model with the same network structure as the non-new user load prediction model is constructed to facilitate the "error-free" migration of model parameters, reduce the probability of deviation in model parameters during the migration process due to differences in network structure, and ensure the prediction accuracy of the constructed new user load prediction model.

[0100] The technology provided by the embodiments of the present disclosure can quickly build a new user load prediction model (an electric power load prediction model based on a deep neural network) by migrating the model parameters of a non-new user load prediction model of a typical user among non-new users who belong to the same industry as the new user, thereby obtaining a higher load prediction accuracy in the initial stage of new user access.

[0101] All of the above optional technical solutions can be combined in any way to form optional embodiments of the present application, and will not be described in detail here.

[0102] The following are embodiments of the apparatus disclosed herein, which can be used to implement the method embodiments disclosed herein. For details not disclosed in the apparatus embodiments disclosed herein, please refer to the method embodiments disclosed herein.

[0103] Figure 4 Schematic diagram of a new user load forecasting model device provided by the embodiment of the present disclosure. Figure 4 As shown, the new user load prediction model device includes a data acquisition module 401 , a parameter storage module 402 and a model construction module 403 .

[0104] The data acquisition module 401 is configured to acquire sample data of typical users among non-new users, wherein the typical users and the new users belong to the same industry category;

[0105] The parameter saving module 402 is configured to train and test the preset non-new user load forecasting model using sample data until the load forecasting model meets the preset convergence conditions, and then save the model parameters of the load forecasting model;

[0106] The model construction module 403 is configured to use the model parameters to construct a new user load forecasting model. The technical solution provided by the embodiment of the present disclosure obtains sample data of typical users among non-new users through the data acquisition module 401, wherein the typical users and the new users belong to the same industry category; the parameter preservation module 402 uses the sample data to train and test the preset non-new user load forecasting model until the load forecasting model meets the preset convergence conditions, and then saves the model parameters of the load forecasting model; the model construction module 403 uses the model parameters to construct a new user load forecasting model, which can effectively solve the problem that due to the lack of historical data on new users, it is difficult to apply existing prediction technologies based on deep neural networks for load forecasting, resulting in low accuracy of load forecasting results.

[0107] In some embodiments, the data acquisition module 401 includes a classification unit and a sample data acquisition unit.

[0108] The classification unit is configured to screen out M non-new users whose sample data volume meets a preset data volume range from the proxy users, classify the M non-new users, and obtain N non-new user sets, wherein each non-new user set includes at least one non-new user, N≤M, and N and M are both positive integers;

[0109] The sample data acquisition unit is configured to acquire sample data of typical users from a set of non-new users who belong to the same industry category as the new user.

[0110] In some embodiments, the sample data acquisition unit may be specifically configured as follows:

[0111] Filter out a set of typical non-new users of the same type as the new user and belonging to the same industry category from N non-new user sets;

[0112] Calculate the distance between each non-new user in the typical non-new set of the same type and the cluster center point of the typical non-new set of the same type;

[0113] The non-new user with the smallest distance to the cluster center is determined as the typical user of the same type of typical non-new set, and sample data of the typical user is obtained.

[0114] In some embodiments, a typical non-new user set of the same type as the new user and belonging to the same industry category is screened out from the N non-new user sets, specifically including:

[0115] Determine the non-new typical users in each non-new user set to obtain N non-new typical users;

[0116] Acquire first sample data of N non-new typical users, and generate a first load curve for each non-new typical user based on the first sample data;

[0117] Acquire second sample data of the new user, and generate a second load curve of the new user based on the second sample data;

[0118] According to the correlation between the second load curve and each first load curve, a set of similar typical non-new users belonging to the same industry category as the new user is screened out.

[0119] In some embodiments, the above method of screening out a set of typical non-new users belonging to the same industry category as the new user based on the correlation between the second load curve and each first load curve specifically includes:

[0120] calculating the Pearson coefficient of the second load curve and each first load curve respectively;

[0121] Based on the Pearson coefficient, a set of typical non-new users of the same type as new users who belong to the same industry category is screened out.

[0122] In some embodiments, the parameter storage module 402 includes a division unit, a training unit, an evaluation and testing unit, and a storage unit.

[0123] The partitioning unit is configured to partition the sample data into a training data set and a test data set;

[0124] The training unit is configured to train a preset deep learning model using a training data set to obtain a load forecasting model;

[0125] The evaluation test unit is configured to perform an evaluation test on the load forecasting model using a test data set to obtain an evaluation test result;

[0126] The saving unit is configured to save the model parameters of the load forecasting model when the evaluation test result meets the preset convergence condition.

[0127] In some embodiments, the model building module 403 includes a building unit and a migration unit.

[0128] The building unit is configured to build a pre-trained model having the same network structure as the non-new user load prediction model;

[0129] The migration unit is configured to migrate the model parameters to the pre-trained model to obtain a new user load prediction model.

[0130] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure.

[0131] Figure 5 Schematic diagram of the electronic device 5 provided by the embodiment of the present disclosure. Figure 5 As shown, the electronic device 5 of this embodiment includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable by the processor 501. When the processor 501 executes the computer program 503, the steps of the above-mentioned method embodiments are implemented. Alternatively, when the processor 501 executes the computer program 503, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0132] For example, the computer program 503 may be divided into one or more modules / units, which are stored in the memory 502 and executed by the processor 501 to implement the present disclosure. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program 503 in the electronic device 5.

[0133] The electronic device 5 may be a desktop computer, a notebook, a PDA, a cloud server, or other electronic device. The electronic device 5 may include but is not limited to a processor 501 and a memory 502. Those skilled in the art will appreciate that Figure 5 It is only an example of the electronic device 5 and does not constitute a limitation of the electronic device 5. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0134] The processor 501 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0135] The memory 502 can be an internal storage unit of the electronic device 5, such as a hard drive or memory of the electronic device 5. The memory 502 can also be an external storage device of the electronic device 5, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the electronic device 5. Furthermore, the memory 502 can include both an internal storage unit of the electronic device 5 and an external storage device. The memory 502 is used to store computer programs and other programs and data required by the electronic device. The memory 502 can also be used to temporarily store data that has been output or is about to be output.

[0136] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0137] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0138] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this disclosure.

[0139] In the embodiments provided in the present disclosure, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely schematic. For example, the division of modules or units is merely a logical function division. In actual implementation, there may be other division methods. Multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.

[0140] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0141] In addition, the functional units in the various embodiments of the present disclosure may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0142] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present disclosure implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. The computer program may include computer program code, which may be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0143] The above embodiments are only used to illustrate the technical solutions of the present disclosure, rather than to limit them. Although the present disclosure 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 of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present disclosure, and should all be included in the scope of protection of the present disclosure.

Claims

1. A method for constructing a new user load forecasting model, characterized in that: include: Obtaining sample data of typical users among non-new users, wherein the typical users and the new users belong to the same industry category; Using the sample data to train and test a preset non-new user load forecasting model until the load forecasting model meets a preset convergence condition, and saving model parameters of the load forecasting model; Using the model parameters to construct a new user load forecasting model; Obtain sample data for typical non-new users, including: Screening out M non-new users whose sample data volume meets a preset data volume range from the proxy users, classifying the M non-new users to obtain N non-new user sets, wherein each non-new user set includes at least one non-new user, N≤M, and N and M are both positive integers; Obtaining sample data of typical users from a set of non-new users belonging to the same industry category as the new user; Obtain sample data of typical users from a set of non-new users belonging to the same industry category as the new user, including: Screening out a set of typical non-new users of the same type as the new user and belonging to the same industry category from the N non-new user sets; Calculating the distance between each non-new user in the similar typical non-new set and the cluster center point of the similar typical non-new set; The non-new user with the smallest distance from the cluster center is determined as a typical user of the same type typical non-new set, and sample data of the typical user is obtained.

2. The method according to claim 1, characterized in that The step of screening out a set of typical non-new users of the same type as the new user and belonging to the same industry category as the new user from the N sets of non-new users includes: Determine a non-new typical user in each of the non-new user sets to obtain N non-new typical users; Acquire first sample data of the N non-new typical users, and generate a first load curve for each of the non-new typical users based on the first sample data; Acquire second sample data of the new user, and generate a second load curve of the new user based on the second sample data; According to the correlation between the second load curve and each of the first load curves, a set of typical non-new users of the same type as the new user and belonging to the same industry category is screened out.

3. The method according to claim 2, characterized in that The step of screening out a set of typical non-new users of the same type as the new user and belonging to the same industry category according to the correlation between the second load curve and each of the first load curves includes: respectively calculating the Pearson coefficient between the second load curve and each of the first load curves; Based on the Pearson coefficient, a set of typical non-new users of the same type as the new user who belong to the same industry category is screened out.

4. The method according to claim 1, wherein The method of training and testing a preset non-new user load forecasting model using the sample data until the load forecasting model satisfies a preset convergence condition and saving model parameters of the load forecasting model includes: Dividing the sample data into a training data set and a test data set; Using the training data set to train a preset deep learning model to obtain a load forecasting model; Using the test data set to perform an evaluation test on the load forecasting model to obtain an evaluation test result; When the evaluation test result meets the preset convergence condition, the model parameters of the load forecasting model are saved.

5. The method according to claim 1, wherein The constructing a new user load forecasting model using the model parameters includes: Construct a pre-trained model with the same network structure as the non-new user load forecasting model; The model parameters are transferred to the pre-trained model to obtain a new user load prediction model.

6. A new user load forecasting model construction device, characterized in that: include: a data acquisition module configured to acquire sample data of typical users among non-new users, wherein the typical users and the new users belong to the same industry category; a parameter saving module configured to train and test a preset non-new user load forecasting model using the sample data until the load forecasting model meets a preset convergence condition, and then save the model parameters of the load forecasting model; A model building module is configured to build a new user load forecasting model using the model parameters; Obtain sample data for typical non-new users, including: Screening out M non-new users whose sample data volume meets a preset data volume range from the proxy users, classifying the M non-new users to obtain N non-new user sets, wherein each non-new user set includes at least one non-new user, N≤M, and N and M are both positive integers; Obtaining sample data of typical users from a set of non-new users belonging to the same industry category as the new user; Obtain sample data of typical users from a set of non-new users belonging to the same industry category as the new user, including: Screening out a set of typical non-new users of the same type as the new user and belonging to the same industry category from the N non-new user sets; Calculating the distance between each non-new user in the similar typical non-new set and the cluster center point of the similar typical non-new set; The non-new user with the smallest distance from the cluster center is determined as a typical user of the same type typical non-new set, and sample data of the typical user is obtained.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images