A real-time load resource scheduling method, system, device, and storage medium

Abstract

This invention discloses a real-time load resource scheduling method, system, device, and storage medium. The real-time load resource scheduling method acquires historical load data and demand response information of each user, and uses the K-means clustering algorithm to efficiently and accurately obtain electricity consumption behavior data. Simultaneously, it calculates the participation value of each user, improving the reliability of subsequent real-time load resource scheduling schemes by considering user participation values. A combined neural network of convolutional neural networks and long short-term memory networks is used for load prediction during the scheduling period. The convolutional neural network has strong feature extraction capabilities from historical load data, and the long short-term memory network can learn long-term dependencies, making the prediction process more efficient and improving the accuracy of load prediction results. Subsequently, a real-time load resource scheduling scheme is generated based on electricity consumption behavior data, participation values, scheduling data, and load prediction results, achieving precise reduction and timely scheduling of peak loads on the power grid.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a real-time load resource scheduling method, system, device and storage medium. Background Technology

[0002] With the rapid popularization of digital smart meters, power grid information acquisition systems have collected a large amount of user load data, but this load data is still not being efficiently mined and utilized. The rapid growth in energy consumption and the increasing uncertainties facing the power grid have made real-time scheduling of power grid load resources increasingly important. However, existing load resource scheduling methods fail to utilize user load data, resulting in poor timeliness and reliability of scheduling, and an inability to accurately reduce peak loads and ensure power grid security. Summary of the Invention

[0003] The purpose of this invention is to at least partially solve one of the technical problems existing in the prior art.

[0004] To this end, embodiments of the present invention provide a real-time load resource scheduling method, system, device, and storage medium, which generates a real-time load resource scheduling scheme, realizes precise reduction and timely scheduling of peak loads of the power grid, and protects the safety of the power grid.

[0005] To achieve the above-mentioned technical objectives, the technical solutions adopted in the embodiments of the present invention include:

[0006] On one hand, embodiments of the present invention provide a real-time load resource scheduling method, including the following steps:

[0007] Acquire historical load data for each user and demand response status for each user, including the number of demand responses received and the number of responses received;

[0008] Based on the preprocessed historical load data, the K-means clustering algorithm is used to classify the users' electricity consumption behavior to obtain electricity consumption behavior data. The preprocessing of the historical load data includes missing value completion, outlier correction, and normalization.

[0009] The participation value of each user is calculated based on the demand response status, and the participation value is the ratio of the number of responses to the number of demand responses received.

[0010] In response to receiving a load resource scheduling task, a scheduling period and scheduling data are determined according to the load resource scheduling task, wherein the scheduling period is the period during which load resource scheduling is required, and the scheduling data is the data that needs to be scheduled within the scheduling period;

[0011] Based on the preprocessed historical load data, a load prediction result for the scheduling period is obtained by combining a convolutional neural network and a long short-term memory network.

[0012] A real-time load resource scheduling scheme is generated based on the electricity consumption behavior data, the participation value, the scheduling data, and the load forecast results.

[0013] In addition, the real-time load resource scheduling method according to the above embodiments of the present invention may also have the following additional technical features:

[0014] Furthermore, in a real-time load resource scheduling method according to an embodiment of the present invention, the step of classifying users' electricity consumption behavior using a K-means clustering algorithm based on the preprocessed historical load data to obtain electricity consumption behavior data includes:

[0015] Select a first cluster center with a first number of elements from the historical load data of each user;

[0016] The first cluster center is used as the current cluster center;

[0017] Calculate the distance between the historical load data of each user and each current cluster center, and then assign the historical load data of each user to the nearest current cluster center in sequence to form a cluster of the first number;

[0018] The number of second cluster centers is equal to the first number, calculated based on the clustering.

[0019] Based on the historical load data, the current cluster center, and the second cluster center, determine whether the preset termination condition is met;

[0020] If not, the second cluster center is used as the current cluster center, and the distance between the historical load data of each user and each current cluster center is calculated. The historical load data of each user is then assigned to the nearest current cluster center in sequence to form a number of clusters equal to the first number.

[0021] If so, the electricity consumption behavior data is obtained based on the clustering.

[0022] Furthermore, in one embodiment of the present invention, the electricity consumption behavior data includes electricity consumption behavior types and corresponding electricity consumption behavior percentages. The electricity consumption behavior types include low electricity consumption with no fluctuation, low electricity consumption with fluctuation, high electricity consumption with no fluctuation, and high electricity consumption with fluctuation, and the first quantity is 4.

[0023] The process of obtaining the electricity consumption behavior data based on the clustering includes:

[0024] The number of electricity consumption behavior types is the first number obtained based on the clustering;

[0025] The ratio of the number of historical load data points in each cluster to the total number of historical load data points is calculated to obtain the proportion of electricity consumption behavior.

[0026] Furthermore, in one embodiment of the present invention, the real-time load resource scheduling scheme includes target demand response transmission and load resource surplus configuration, and the load prediction result includes load prediction value and response load prediction value;

[0027] The step of generating a real-time load resource scheduling scheme based on the electricity consumption behavior data, the participation value, the scheduling data, and the load forecast results includes:

[0028] The target demand response is sent based on the participation value, the scheduling data, and the response load prediction value.

[0029] Load resource reserves are allocated based on the electricity consumption behavior data and the load forecast values.

[0030] Furthermore, in one embodiment of the present invention, the step of sending a target demand response based on the participation value, the scheduling data, and the response load prediction value includes:

[0031] Generate a target demand response based on the scheduling data and the predicted response load;

[0032] The target demand response is sent to the users corresponding to the participation values ​​in descending order of participation value.

[0033] Furthermore, in one embodiment of the present invention, the step of configuring load resource margin based on the electricity consumption behavior data and the load forecast value includes:

[0034] Target users are identified based on the electricity consumption behavior data; these target users are those who have experienced fluctuations in electricity consumption.

[0035] The load resource margin is calculated based on the load forecast value;

[0036] Allocate the remaining load resources to the target user.

[0037] Furthermore, in one embodiment of the present invention, the historical load data and the demand response status are obtained from a historical database that is updated in real time;

[0038] The real-time load resource scheduling method further includes:

[0039] The load data during the scheduling period is stored in the historical database.

[0040] On the other hand, embodiments of the present invention propose a real-time load resource scheduling system, comprising:

[0041] The first module is used to obtain historical load data of each user and demand response status of each user, wherein the demand response status includes the number of demand responses received and the number of responses received.

[0042] The second module is used to classify users' electricity consumption behavior using the K-means clustering algorithm based on the preprocessed historical load data to obtain electricity consumption behavior data. The preprocessing of the historical load data includes missing value completion, outlier correction, and normalization.

[0043] The third module is used to calculate the participation value of each user based on the demand response status, wherein the participation value is the ratio of the number of responses to the number of received demand responses;

[0044] The fourth module is used to respond to receiving a load resource scheduling task, and determine the scheduling period and scheduling data according to the load resource scheduling task, wherein the scheduling period is the period during which load resource scheduling is required, and the scheduling data is the data that needs to be scheduled within the scheduling period;

[0045] The fifth module is used to make predictions based on the preprocessed historical load data, combined with convolutional neural networks and long short-term memory networks, to obtain the load prediction results for the scheduling period.

[0046] The sixth module is used to generate a real-time load resource scheduling scheme based on the electricity consumption behavior data, the participation value, the scheduling data, and the load forecast results.

[0047] On the other hand, embodiments of the present invention provide a real-time load resource scheduling device, comprising:

[0048] At least one processor;

[0049] At least one memory for storing at least one program;

[0050] When the at least one program is executed by the at least one processor, the at least one processor implements the real-time load resource scheduling method.

[0051] On the other hand, embodiments of the present invention provide a storage medium storing a processor-executable program, which, when executed by a processor, is used to implement the aforementioned real-time load resource scheduling method.

[0052] Advantages and beneficial effects of the present invention:

[0053] This invention acquires historical load data and demand response information for each user, and uses K-means clustering algorithm to efficiently and accurately obtain electricity consumption behavior data based on the preprocessed historical load data. Simultaneously, it calculates the participation value of each user based on the demand response information, improving the reliability of subsequent real-time load resource scheduling schemes by considering user participation values. A combined neural network of convolutional neural networks and long short-term memory networks is used for load forecasting within the scheduling period. Convolutional neural networks have strong feature extraction capabilities from historical load data, while long short-term memory networks can learn long-term dependencies, making them suitable for modeling historical load data. This makes the forecasting process more efficient and improves the accuracy of load forecast results. Subsequently, a real-time load resource scheduling scheme is generated based on electricity consumption behavior data, participation values, scheduling data, and load forecast results, achieving precise reduction and timely scheduling of peak loads on the power grid, thus protecting power grid security. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following description is provided with accompanying drawings of the relevant technical solutions in the embodiments of this application or the prior art. It should be understood that the accompanying drawings described below are only for the purpose of clearly illustrating some embodiments of the technical solutions in this application. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0055] Figure 1 This is a flowchart illustrating a specific embodiment of a real-time load resource scheduling method according to the present invention;

[0056] Figure 2 This is the basic unit structure of a combination of convolutional neural network and long short-term memory network in a specific embodiment of a real-time load resource scheduling method of the present invention;

[0057] Figure 3 This is a schematic diagram of a specific embodiment of the real-time load resource scheduling system of the present invention;

[0058] Figure 4 This is a schematic diagram of a specific embodiment of a real-time load resource scheduling device according to the present invention. Detailed Implementation

[0059] The embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application. The step numbers in the following embodiments are set only for ease of explanation, and there is no limitation on the order between the steps. The execution order of each step in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0060] The terms "first," "second," "third," and "fourth," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0061] In this invention, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0062] With the rapid popularization of digital smart meters, power grid information acquisition systems have collected a large amount of user load data, but this load data is still not being efficiently mined and utilized. The rapid growth in energy consumption and the increasing uncertainties facing the power grid have made real-time scheduling of power grid load resources increasingly important. However, existing load resource scheduling methods fail to utilize user load data, resulting in poor timeliness and reliability of scheduling, and an inability to accurately reduce peak loads and ensure power grid security. To address this, this invention proposes a real-time load resource scheduling method, system, device, and storage medium. It acquires historical load data and demand response information from each user, and uses a K-means clustering algorithm to efficiently and accurately obtain electricity consumption behavior data based on the preprocessed historical load data. Simultaneously, it calculates the participation value of each user based on the demand response information, improving the reliability of subsequent real-time load resource scheduling schemes by considering user participation values. A combined neural network of convolutional neural networks and long short-term memory networks is used for load prediction during the scheduling period. The convolutional neural network has strong feature extraction capabilities from historical load data, while the long short-term memory network can learn long-term dependencies, making it suitable for modeling historical load data. This makes the prediction process more efficient and improves the accuracy of load prediction results. Subsequently, a real-time load resource scheduling scheme is generated based on electricity consumption behavior data, participation values, scheduling data, and load prediction results, achieving precise reduction and timely scheduling of peak loads on the power grid, thus protecting power grid security.

[0063] The following describes in detail, with reference to the accompanying drawings, a real-time load resource scheduling method, system, apparatus, and storage medium according to an embodiment of the present invention. First, a real-time load resource scheduling method according to an embodiment of the present invention will be described with reference to the accompanying drawings.

[0064] Reference Figure 1 This invention provides a real-time load resource scheduling method, which can be applied to a terminal, a server, or software running on a terminal or server. The terminal can be a tablet computer, laptop computer, desktop computer, etc., but is not limited to these. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. The real-time load resource scheduling method in this invention mainly includes the following steps S101-S106:

[0065] S101. Obtain historical load data and demand response status for each user;

[0066] The demand response status includes the number of demand responses received and the number of responses received.

[0067] Understandably, a user's historical load data includes load data for each time period of each day in each past cycle.

[0068] Optionally, in some embodiments, smart meters can be used to obtain historical load data for each user.

[0069] Optionally, in some embodiments, each day in the past cycle included 24 time periods (one hour per time period), and the historical load data of a single user was recorded as a set:

[0070] U i =(L i W i g i p i P i )

[0071]

[0072] In the formula, U i L represents the i-th user; i W represents the load of the i-th user in each time period; i The temperature represents the electricity consumption of the i-th user at various time periods; g i This represents the identifier of the i-th user receiving the request response in each time period; p i P represents the identifier indicating the response of the i-th user to the request in each time period; i This represents the response load of the i-th user in each time period.

[0073] Understandably, according to g i The number of times the i-th user's request to be responded to can be obtained, based on p. i We can obtain the number of responses for the i-th user, that is, we can obtain the demand response status of each user based on the historical load data of each user.

[0074] S102. Based on the preprocessed historical load data, the K-means clustering algorithm is used to classify the users' electricity consumption behavior to obtain electricity consumption behavior data.

[0075] In some embodiments, the preprocessing of historical load data includes missing value completion, outlier correction, and normalization.

[0076] It is understandable that power grid systems are widely distributed and operate in complex environments, resulting in a large volume of historical load data. This data is susceptible to loss and anomalies due to performance fluctuations or faults within the power grid system. Therefore, in this embodiment of the invention, missing values ​​are filled in and outliers are corrected in the historical load data to improve its completeness and accuracy, thereby making subsequent load forecasts more accurate and reliable.

[0077] Based on prior knowledge, missing value completion can handle both single missing data values ​​and multiple missing data values ​​in historical load data.

[0078] In an embodiment of the present invention, for a single missing data value in historical load data, based on the regularity of load data and time series correlation, the average value of the two load data before and after the missing value is selected, and the missing value is filled in using the average value.

[0079] Optionally, in some embodiments, if load data is missing in the third time period of a certain day in the historical load data, the load data of the second and fourth time periods are selected, and the missing load data of the third time period is supplemented by the average of the load data of the second and fourth time periods.

[0080] For historical load data with multiple missing values, a similarity-based imputation method is used to complete the missing values. This involves selecting the three dates with the highest similarity to the dates containing the missing values, calculating the average load data for the same time period as the missing values ​​on these three dates, and using this average to impute the missing values. Here, similarity refers to the similarity of load data at corresponding times between two different dates.

[0081] Optionally, in some embodiments, if the third, fourth, and fifth time periods of day d are missing load data in the historical load data, the three dates d1, d2, and d3 with the highest similarity to day d are selected, and the average load data of the third time period of d1, d2, and d3 is calculated. The average load data of the third time period of d1, d2, and d3 is used to fill in the missing values ​​of the third time period of day d. The average load data of the fourth time period of d1, d2, and d3 is calculated, and the average load data of the fourth time period of d1, d2, and d3 is used to fill in the missing values ​​of the fourth time period of day d. The average load data of the fifth time period of d1, d2, and d3 is calculated, and the average load data of the fifth time period of d1, d2, and d3 is used to fill in the missing values ​​of the fifth time period of day d.

[0082] That is, the missing values ​​at time t on day d:

[0083] x(d,t)=[x(d1,t)+x(d2,t)+x(d3,t)] / 3

[0084] Where x(d1,t) represents the load data at time t on day d1, x(d2,t) represents the load data at time t on day d2, and x(d3,t) represents the load data at time t on day d3.

[0085] It is understandable that before correcting outliers in historical load data, it is necessary to first identify outliers (the time points where outliers occur) in the historical load data. In some embodiments of this invention, a combination of horizontal and vertical judgment methods is used to identify outliers. Based on prior knowledge and the aforementioned regularity of load data and time series correlation, the horizontal judgment method determines whether the change in load data at a certain time point exceeds a set first threshold ε compared to the load data at adjacent time points. p To determine whether a given time point is a horizontal outlier, the following formula is used:

[0086] max[|x(d,t)-x(d,t-1)|, |x(d,t)-x(d,t+1)|]>ε p

[0087] The vertical judgment method determines whether there are significant abrupt changes (exceeding a set second threshold η) in the load data at various times on a given day compared to the average load data at corresponding times on the previous and following days. p This is used to determine whether there are outliers (vertical outliers) on a given day. The formula is as follows:

[0088] |x(d,t)-a(t)|>η p

[0089] In the formula, a(t) is the average load data at time t for 5 days before and after day d.

[0090] Based on prior knowledge, different categories of historical load data, such as load volume and the identifier of received demand response, exhibit different orders of magnitude and units. If raw historical load data is directly input into a machine learning model, data with smaller fluctuations will be lost, leading to reduced prediction accuracy due to incomplete information mining. Simultaneously, excessively large data volumes will slow down model training, degrading the timeliness of the prediction model. Therefore, in the embodiments of this invention, after imputing missing values ​​and correcting outliers in the historical load data, normalization processing is required to ensure uniformity in the scale (order of magnitude) of various data points within the historical load data.

[0091] Optionally, in some embodiments, the historical load data is normalized using the extreme value normalization method, as shown in the following formula:

[0092]

[0093] In the formula, x represents the sample data, x max x min These represent the maximum and minimum values ​​of the sample data, x. * For standardized sample data, [b g a g ] represents the normalization interval, which is generally [-1, 1] or [0, 1].

[0094] Specifically, in the embodiments of the present invention, after the historical load data is preprocessed, the user's electricity consumption behavior is classified using the K-means clustering algorithm based on the preprocessed historical load data to obtain electricity consumption behavior data.

[0095] Based on prior knowledge, K-means clustering is a distance-based, iterative clustering analysis algorithm that uses Euclidean distance as the similarity metric to evaluate samples. Furthermore, K-means clustering is efficient, converges quickly, and produces good clustering results when processing large amounts of data, thus enabling efficient and accurate acquisition of electricity consumption behavior data. In some embodiments of this invention, K-means clustering is used to select K historical load data points from all historical load data as initial cluster centers, and these initial cluster centers are considered as the current cluster centers. The distance between each historical load data point and each current cluster center is then calculated, and each historical load data point is assigned to the nearest current cluster center, thereby forming multiple clusters. It is understood that during the allocation of historical load data, the current cluster center of each cluster is recalculated based on the existing objects (historical load data) in the cluster, and the distance between each historical load data point and each current cluster center is recalculated. Then, each historical load data point is again assigned to the nearest current cluster center, until a preset termination condition is met.

[0096] Therefore, S102 can be further divided into the following steps S1021-S1027:

[0097] Step S1021: Select a first cluster center from the historical load data of each user, with the number of cluster centers being the first quantity.

[0098] Optionally, the electricity consumption behavior data includes electricity consumption behavior types and corresponding electricity consumption behavior percentages. In some embodiments, the electricity consumption behavior types include four categories: low electricity consumption with no fluctuation (T1), low electricity consumption with fluctuation (T2), high electricity consumption with no fluctuation (T3), and high electricity consumption with fluctuation (T4). Therefore, in some embodiments, four first cluster centers C1 are selected from the historical load data of each user, i.e., the first number is 4.

[0099] Based on prior knowledge, the ultimate goal of the K-means clustering algorithm is to find the minimum value of the Sum of Squared Error (SSE) function:

[0100]

[0101] In the formula, x represents the data object, i.e., the historical load data of each user; C i Let i be the i-th cluster center (the first cluster center when i=1). It can be understood that i+1 is obtained by incrementing the number of iterations of the K-means clustering algorithm by one; k is the number of clusters, i.e., the first number in this embodiment. The size of SSE represents the reliability of the clustering results; the smaller the SSE, the higher the reliability of the clustering results.

[0102] Step S1022: Set the first cluster center as the current cluster center;

[0103] Step S1023: Calculate the distance between the historical load data of each user and each current cluster center, and assign the historical load data of each user to the nearest current cluster center in sequence to form a cluster of the first number;

[0104] Based on prior knowledge, we know that the historical load data of each user is sequentially assigned to the nearest current cluster center, which means that the historical load data of each user is sequentially assigned to the current cluster center with the highest similarity.

[0105] Step S1024: Calculate the number of second cluster centers, which is the first number, based on clustering calculations;

[0106] Understandably, when historical load data is distributed to the nearest current cluster center to form clusters, the current cluster center of each cluster is recalculated based on the existing objects (historical load data) in the cluster, thus obtaining a second cluster center with the same number as the first cluster center.

[0107] Step S1025: Determine whether the preset termination conditions are met based on historical load data, the current cluster center, and the second cluster center;

[0108] Optionally, in some embodiments of the present invention, the preset termination condition may be one of the following:

[0109] a) No (or a preset minimum number) historical load data were reassigned to different clusters;

[0110] b) No (or the preset minimum number) cluster centers change again;

[0111] c) The sum of squared errors of clustering is locally minimized.

[0112] Step S1026: If not, take the second cluster center as the current cluster center and return to step S1023;

[0113] Step S1027: If so, obtain electricity consumption behavior data based on clustering.

[0114] Optionally, as described in step S1021, in some embodiments, the electricity consumption behavior data includes electricity consumption behavior types and corresponding electricity consumption behavior proportions. The electricity consumption behavior types include low electricity consumption with no fluctuation (T1), low electricity consumption with fluctuation (T2), high electricity consumption with no fluctuation (T3), and high electricity consumption with fluctuation (T4). It is understood that obtaining electricity consumption behavior data based on clustering includes determining each electricity consumption behavior type and calculating the electricity consumption behavior proportion for each type.

[0115] It is understood that, in the embodiments of the present invention, the K-means clustering algorithm ultimately obtains each cluster as a type of electricity consumption behavior, and the ratio of the number of historical load data in each cluster to the total number of historical load data can be calculated to obtain the proportion of electricity consumption behavior.

[0116] Optionally, in some embodiments, the number of historical load data (number of users) in the low power consumption fluctuation (T1) category of area A is N. A1 The number of historical load data points (number of users) in the low-electricity-consumption-fluctuation (T2) category is N. A2 The number of historical load data points (number of users) in the high-consumption-no-fluctuation (T3) category is N. A3 The number of historical load data points (number of users) in the high-consumption fluctuation (T4) category is N. A4 Then, the percentage of electricity consumption behavior with low and stable electricity consumption (T1) in area A is:

[0117]

[0118] S103. Calculate the participation value of each user based on the demand response;

[0119] The participation value is the ratio of the number of responses to the number of requests received.

[0120] Specifically, in embodiments of the present invention, the participating values ​​are:

[0121]

[0122] Where i represents the i-th user; m represents the m-th time; g im p represents the number of times the i-th user receives a request response at time m; im This represents the number of responses from the i-th user at time m.

[0123] It is understandable that a higher ratio of the number of responses to the number of received demand responses indicates a higher user response rate; that is, the participation value reflects the enthusiasm of each user in responding. This embodiment of the invention improves the reliability of subsequent real-time load resource scheduling schemes by considering the participation values ​​of each user.

[0124] S104. In response to receiving a load resource scheduling task, determine the scheduling period and scheduling data according to the load resource scheduling task;

[0125] The scheduling period refers to the time period during which load resource scheduling is required, and the scheduling data refers to the data during the scheduling period that requires load resource scheduling.

[0126] Optionally, in some embodiments, the load resource scheduling task can be a manually issued load resource scheduling task, or it can be a load resource scheduling task automatically issued by the power grid system based on real-time load conditions and preset load plans.

[0127] S105. Based on the preprocessed historical load data, a prediction is made using a convolutional neural network and a long short-term memory network to obtain the load prediction results for the scheduling period.

[0128] Based on prior knowledge, Convolutional Neural Networks (CNNs) excel at extracting features from continuous historical time-series data, meaning they can efficiently and accurately extract features from the historical load data in this embodiment of the invention. Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network capable of learning long-term dependencies, making them suitable for modeling time-series data, and therefore also applicable to modeling historical load data based on this embodiment of the invention. It is understood that by combining CNNs and LSTM networks, this embodiment of the invention, based on a CNN-LSTM network, can not only obtain the temporal relationships of historical load data but also extract its spatial features, resulting in more accurate load prediction results and a more efficient prediction process.

[0129] Optionally, refer to Figure 2 In some embodiments, the CNN-LSTM network directly uses convolutions as part of the LSTM input. The core essence of the CNN-LSTM network is the same as LSTM, consisting of memory cells, input gates, output gates, and forget gates. The difference lies in the fact that the connections between the input and each gate are changed from feedforward to convolution, and convolution operations are also used between states. Figure 2 The calculation formula for the CNN-LSTM network is as follows:

[0130]

[0131]

[0132]

[0133]

[0134]

[0135] In the formula, i t Indicates the input gate; f t Indicates the forgetting gate; C t Indicates a transitional state; o t Indicates the input gate; H t The current state is represented by ; W is the weight coefficient matrix of the corresponding convolution kernel; b is the bias term of the corresponding gate; * represents convolution; ο is the Hadamard product; σ and tanh are activation functions, where σ is the Sigmoid function.

[0136] Optionally, in some embodiments, the accuracy of the model prediction is evaluated based on the load forecast results using the Mean Absolute Percentage Error (MAPE), as shown in the following formula:

[0137]

[0138] Optionally, in some embodiments, historical load data and corresponding historical weather, temperature and wind direction are input into the CNN-LSTM network for training and learning to obtain load prediction results for the scheduling period.

[0139] Optionally, in some embodiments, the load forecast result includes a load forecast value and a response load forecast value. It is understood that, within the same distribution area, the total load of the distribution area during the scheduling period is obtained by adding the load forecast values ​​of each user, and the total response load of the distribution area during the scheduling period is obtained by adding the response load forecast values ​​of each user.

[0140] S106. Generate a real-time load resource scheduling scheme based on electricity consumption behavior data, participation values, scheduling data, and load forecast results.

[0141] Optionally, in some embodiments, the real-time load resource scheduling scheme includes target demand response transmission and load resource surplus configuration. As can be seen from step S105, in some embodiments, the load forecasting result includes load forecast values ​​and response load forecast values.

[0142] Specifically, S106 can be further divided into the following steps S1061-S1062:

[0143] Step S1061: Send the target demand response based on the participation value, scheduling data, and response load forecast value;

[0144] As shown in step S103, the participation value represents the user's response rate to the demand response. It is understandable that sending the target demand response to users with higher participation values ​​increases the probability of receiving a response.

[0145] In some embodiments, step S1061 specifically includes:

[0146] 1) Generate the target demand response based on scheduling data and predicted response load;

[0147] Specifically, in some embodiments, the target demand response to be sent to the user is determined based on the scheduling data within the scheduling period determined by the load resource scheduling task and the response load forecast value of the scheduling period.

[0148] 2) Send the target demand response to the users corresponding to the participation values ​​in descending order of participation value.

[0149] Step S1062: Configure load resource reserves based on electricity consumption behavior data and load forecast values.

[0150] Specifically, in embodiments of the present invention, load resource reserves are reserved for users within the distribution area who have experienced electricity consumption fluctuations, using load forecast values ​​during the scheduling period, in order to suppress electricity consumption fluctuations. It is understood that the amount of load resource reserves configured is determined based on the load forecast values.

[0151] In some embodiments, step S1062 specifically includes:

[0152] 1) Identify target users based on electricity consumption behavior data;

[0153] The target users are those who have experienced fluctuations in electricity consumption.

[0154] Optionally, in some embodiments, the electricity consumption behavior types in the electricity consumption behavior data include low electricity consumption with no fluctuation (T1), low electricity consumption with fluctuation (T2), high electricity consumption with no fluctuation (T3), and high electricity consumption with fluctuation (T4), wherein users of T2 and T4 are target users.

[0155] 2) Calculate the load resource margin based on the load forecast values;

[0156] 3) Allocate the remaining load resources to the target users.

[0157] It is understood that the real-time load resource scheduling scheme generated in the embodiments of the present invention is a real-time load resource scheduling scheme within a transformer area. When applied to multiple transformer areas, it is only necessary to use the real-time load resource scheduling method described in the embodiments of the present invention to generate real-time load resource scheduling schemes for each transformer area, thereby realizing real-time load resource scheduling for multiple transformer areas.

[0158] Optionally, in some embodiments, historical load data and demand response status for each user are obtained from a real-time updated historical database. In embodiments of the present invention, after load resource scheduling is performed using a real-time load resource scheduling scheme, the load data within the scheduling period is stored in a historical database for real-time updates.

[0159] As described in steps S101-S106, this invention acquires historical load data and demand response information for each user. Based on the preprocessed historical load data, it uses a K-means clustering algorithm to efficiently and accurately obtain electricity consumption behavior data. Simultaneously, it calculates the participation value of each user based on the demand response information, improving the reliability of subsequent real-time load resource scheduling schemes by considering user participation values. Furthermore, it employs a combined neural network of convolutional neural networks and long short-term memory networks for load prediction during the scheduling period. The convolutional neural network has strong feature extraction capabilities for historical load data, while the long short-term memory network can learn long-term dependencies, making it suitable for modeling historical load data. This makes the prediction process more efficient and improves the accuracy of load prediction results. Subsequently, a real-time load resource scheduling scheme is generated based on electricity consumption behavior data, participation values, scheduling data, and load prediction results, achieving precise reduction and timely scheduling of peak loads on the power grid, thus protecting power grid safety.

[0160] Secondly, a real-time load resource scheduling system according to an embodiment of this application is described with reference to the accompanying drawings.

[0161] Figure 3 This is a schematic diagram of a real-time load resource scheduling system according to an embodiment of this application.

[0162] The system specifically includes:

[0163] The first module 301 is used to acquire historical load data of each user and demand response status of each user, wherein the demand response status includes the number of demand responses received and the number of responses received.

[0164] The second module 302 is used to classify the user's electricity consumption behavior using the K-means clustering algorithm based on the preprocessed historical load data to obtain electricity consumption behavior data. The preprocessing of the historical load data includes missing value completion, outlier correction and normalization.

[0165] The third module 303 is used to calculate the participation value of each user based on the demand response status, wherein the participation value is the ratio of the number of responses to the number of received demand responses;

[0166] The fourth module 304 is used to respond to receiving a load resource scheduling task, and determine a scheduling period and scheduling data according to the load resource scheduling task, wherein the scheduling period is the period during which load resource scheduling needs to be performed, and the scheduling data is the data that needs to be scheduled during the scheduling period;

[0167] The fifth module 305 is used to make predictions based on the preprocessed historical load data, combined with convolutional neural networks and long short-term memory networks, to obtain the load prediction results for the scheduling period.

[0168] The sixth module 306 is used to generate a real-time load resource scheduling scheme based on the electricity consumption behavior data, the participation value, the scheduling data, and the load forecast results.

[0169] It is evident that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0170] Reference Figure 4 This application provides a real-time load resource scheduling device, including:

[0171] At least one processor 401;

[0172] At least one memory 402 is used to store at least one program;

[0173] When the at least one program is executed by the at least one processor 401, the at least one processor 401 implements the real-time load resource scheduling method described in steps S101-S106.

[0174] Similarly, the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0175] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this application are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and sub-operations described as part of a larger operation are executed independently.

[0176] Furthermore, although this application is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding this application. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional technology for an engineer. Therefore, those skilled in the art can implement the application set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of this application, which is determined by the full scope of the appended claims and their equivalents.

[0177] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several programs to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0178] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequential list of executable programs for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, a program execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can retrieve and execute a program from or in conjunction with such a program execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit a program for use by or in conjunction with a program execution system, apparatus, or device.

[0179] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0180] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable program execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0181] In the foregoing description of this specification, the references to terms such as "one embodiment," "another embodiment," or "some embodiments," etc., indicate that a specific feature, structure, material, or characteristic described in connection with an embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0182] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0183] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A real-time load resource scheduling method, characterized in that, Includes the following steps: Acquire historical load data for each user and demand response status for each user, including the number of demand responses received and the number of responses received; Based on the preprocessed historical load data, the K-means clustering algorithm is used to classify the users' electricity consumption behavior to obtain electricity consumption behavior data. The preprocessing of the historical load data includes missing value completion, outlier correction, and normalization. The participation value of each user is calculated based on the demand response status, and the participation value is the ratio of the number of responses to the number of received demand responses; In response to receiving a load resource scheduling task, a scheduling period and scheduling data are determined according to the load resource scheduling task, wherein the scheduling period is the period during which load resource scheduling is required, and the scheduling data is the data that needs to be scheduled within the scheduling period; Based on the preprocessed historical load data, a load prediction result for the scheduling period is obtained by combining a convolutional neural network and a long short-term memory network. A real-time load resource scheduling scheme is generated based on the electricity consumption behavior data, the participation value, the scheduling data, and the load forecast results. The real-time load resource scheduling scheme includes target demand response transmission and load resource surplus configuration, and the load prediction result includes load prediction value and response load prediction value; The step of generating a real-time load resource scheduling scheme based on the electricity consumption behavior data, the participation value, the scheduling data, and the load forecast results includes: The target demand response is sent based on the participation value, the scheduling data, and the response load prediction value. Load resource reserves are allocated based on the electricity consumption behavior data and the load forecast values. The step of sending a target demand response based on the participation value, the scheduling data, and the response load prediction value includes: Generate a target demand response based on the scheduling data and the predicted response load; The target demand response is sent to the users corresponding to the participation values ​​in descending order of participation value; The electricity consumption behavior data includes electricity consumption behavior types and corresponding electricity consumption behavior percentages. The electricity consumption behavior types include low electricity consumption with no fluctuation, low electricity consumption with fluctuation, high electricity consumption with no fluctuation, and high electricity consumption with fluctuation. The process of allocating load resource reserves based on the electricity consumption behavior data and the load forecast value includes: Target users are identified based on the electricity consumption behavior data, and these target users are those who have experienced fluctuations in electricity consumption. The load resource margin is calculated based on the load forecast value; Allocate the remaining load resources to the target user.

2. The real-time load resource scheduling method according to claim 1, characterized in that, The step involves classifying user electricity consumption behavior using the K-means clustering algorithm based on the preprocessed historical load data to obtain electricity consumption behavior data, including: Select a first cluster center with a first number of elements from the historical load data of each user; The first cluster center is used as the current cluster center; Calculate the distance between the historical load data of each user and each current cluster center, and then assign the historical load data of each user to the nearest current cluster center in sequence to form a cluster of the first number; The number of second cluster centers is equal to the first number, calculated based on the clustering. Based on the historical load data, the current cluster center, and the second cluster center, determine whether the preset termination condition is met; If not, the second cluster center is used as the current cluster center, and the distance between the historical load data of each user and each current cluster center is calculated. The historical load data of each user is then assigned to the nearest current cluster center in sequence to form a number of clusters equal to the first number. If so, the electricity consumption behavior data is obtained based on the clustering.

3. The real-time load resource scheduling method according to claim 2, characterized in that, The first quantity is 4; The process of obtaining the electricity consumption behavior data based on the clustering includes: The number of electricity consumption behavior types is the first number obtained based on the clustering; The ratio of the number of historical load data points in each cluster to the total number of historical load data points is calculated to obtain the proportion of electricity consumption behavior.

4. The real-time load resource scheduling method according to claim 1, characterized in that, The historical load data and the demand response status are obtained from a real-time updated historical database. The real-time load resource scheduling method further includes: The load data during the scheduling period is stored in the historical database.

5. A real-time load resource scheduling system, characterized in that, include: The first module is used to obtain historical load data of each user and demand response status of each user, wherein the demand response status includes the number of demand responses received and the number of responses received. The second module is used to classify users' electricity consumption behavior using the K-means clustering algorithm based on the preprocessed historical load data to obtain electricity consumption behavior data. The preprocessing of the historical load data includes missing value completion, outlier correction, and normalization. The third module is used to calculate the participation value of each user based on the demand response status, wherein the participation value is the ratio of the number of responses to the number of received demand responses; The fourth module is used to respond to receiving a load resource scheduling task, and determine the scheduling period and scheduling data according to the load resource scheduling task, wherein the scheduling period is the period during which load resource scheduling is required, and the scheduling data is the data that needs to be scheduled within the scheduling period; The fifth module is used to make predictions based on the preprocessed historical load data, combined with convolutional neural networks and long short-term memory networks, to obtain the load prediction results for the scheduling period. The sixth module is used to generate a real-time load resource scheduling scheme based on the electricity consumption behavior data, the participation value, the scheduling data, and the load forecast results; The real-time load resource scheduling scheme includes target demand response transmission and load resource surplus configuration, and the load prediction result includes load prediction value and response load prediction value; The step of generating a real-time load resource scheduling scheme based on the electricity consumption behavior data, the participation value, the scheduling data, and the load forecast results includes: The target demand response is sent based on the participation value, the scheduling data, and the response load prediction value. Load resource reserves are allocated based on the electricity consumption behavior data and the load forecast values. The step of sending a target demand response based on the participation value, the scheduling data, and the response load prediction value includes: Generate a target demand response based on the scheduling data and the predicted response load; The target demand response is sent to the users corresponding to the participation values ​​in descending order of participation value; The electricity consumption behavior data includes electricity consumption behavior types and corresponding electricity consumption behavior percentages. The electricity consumption behavior types include low electricity consumption with no fluctuation, low electricity consumption with fluctuation, high electricity consumption with no fluctuation, and high electricity consumption with fluctuation. The process of allocating load resource reserves based on the electricity consumption behavior data and the load forecast value includes: Target users are identified based on the electricity consumption behavior data, and these target users are those who have experienced fluctuations in electricity consumption. The load resource margin is calculated based on the load forecast value; Allocate the remaining load resources to the target user.

6. A real-time load resource scheduling device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements a real-time load resource scheduling method as described in any one of claims 1-4.

7. A storage medium storing a processor-executable program, characterized in that, The processor-executable program, when executed by the processor, is used to implement a real-time load resource scheduling method as described in any one of claims 1-4.

Images