Demand response user determination method and device, equipment and storage medium

Abstract

The invention discloses a demand response user determination method and apparatus, a device and a storage medium. The method comprises the steps of obtaining a plurality of target historical demand events matched with a target demand event; determining a candidate user set based on historical response information of the user to each target historical demand event; obtaining user basic information and power consumption behavior information of each candidate user in the candidate user set, event characteristics of the target demand event and historical response information of each candidate user to a plurality of historical demand events; determining the response probability of each candidate user to the target demand event based on the user basic information and the power consumption behavior information of each candidate user, the event characteristics of the target demand event and the historical response information of each candidate user to a plurality of historical demand events; and determining a plurality of target users from the candidate user set based on the response probability of each candidate user to the target demand event. The demand response pushing method and device are beneficial to improving the success rate of demand response pushing.

Description

Technical Field

[0001] The present application relates to the field of power regulation technology, and in particular to a method, device, equipment and storage medium for determining a demand response user. Background Art

[0002] In power planning, demand response is a mechanism that allows users to dynamically adjust their electricity consumption patterns based on demand events provided by grid operators, such as price changes, direct load control instructions, etc., to help balance power supply and demand, relieve grid pressure, reduce operating costs, and promote the absorption of renewable energy. When the grid faces peak loads, difficulties in absorbing renewable energy, or sudden failures, grid operators can push demand events to users on the demand side so that users respond to the demand events and adjust their electricity consumption behaviors and patterns to reduce grid load, avoid power outages, or help stabilize frequency and voltage.

[0003] However, after the grid operator pushes the demand event to each user on the demand side, the proportion of users who actually respond to the demand event is relatively low among all users, and the success rate of the push is relatively low. Summary of the invention

[0004] In order to solve the above-mentioned problems existing in the prior art, the embodiments of the present application provide a demand response user determination method, apparatus, device and storage medium, by determining a candidate user set corresponding to a target demand event, and inputting the user basic information, electricity usage behavior information, event characteristics of the target demand event and historical response information of each candidate user in the candidate user set to multiple historical demand events into a target user response probability prediction model to predict the response probability of each candidate user to the target demand event, thereby determining multiple target users corresponding to the target demand event, and the determined target users have a higher probability of responding to the target demand event, and push messages to the target users, which is conducive to improving the success rate of the push messages.

[0005] In a first aspect, an embodiment of the present application provides a method for determining a demand response user, comprising:

[0006] Acquire multiple target historical demand events that match the target demand event;

[0007] Based on the historical response information of the user to each target historical demand event, determining a candidate user set corresponding to the target demand event;

[0008] Obtaining basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events for each candidate user in the candidate user set;

[0009] Inputting the basic information of each candidate user, the electricity consumption behavior information, the event characteristics of the target demand event, and the historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event;

[0010] Based on the response probability of each candidate user to the target demand event, a plurality of target users corresponding to the target demand event are determined from the candidate user set.

[0011] In a second aspect, an embodiment of the present application provides a demand response user determination device, the demand response user determination device comprising an acquisition unit and a processing unit;

[0012] The acquisition unit is used to acquire multiple target historical demand events matching the target demand event;

[0013] Based on the historical response information of the user to each target historical demand event, determining a candidate user set corresponding to the target demand event;

[0014] Obtaining basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events for each candidate user in the candidate user set;

[0015] The processing unit is used to input the basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event;

[0016] Based on the response probability of each candidate user to the target demand event, a plurality of target users corresponding to the target demand event are determined from the candidate user set.

[0017] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor and a memory, the processor being connected to the memory, the memory being used to store a computer program, and the processor being used to execute the computer program stored in the memory, so that the electronic device performs the method described in the first aspect.

[0018] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program enables a computer to execute the method described in the first aspect.

[0019] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the method described in the first aspect.

[0020] Implementing the embodiments of the present application has the following beneficial effects:

[0021] In an embodiment of the present application, first, by obtaining multiple target historical demand events matching the target demand event, based on the historical response information of the user to each target historical demand event, a candidate user set corresponding to the target demand event is determined. Then, the user basic information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events of each candidate user in the candidate user set are obtained. Next, the user basic information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events are input into a pre-trained target user response probability prediction model to obtain the response probability of each candidate user to the target demand event. Finally, based on the response probability of each candidate user to the target demand event, multiple target users corresponding to the target demand event are determined from the candidate user set. Thus, the candidate user set can be determined by the historical response information of the user to each target historical demand event, and the response probability of each candidate user to the target demand event is predicted based on the user basic information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events, so as to determine multiple target users with a higher probability of responding to the target demand event. The target user response probability prediction model can effectively predict the response probability of each candidate user to the target demand event by learning data such as the user's basic information, electricity consumption behavior information, event characteristics of the target demand event, historical response information, etc., thereby improving the accuracy of determining the demand response user, so as to accurately push the target demand event to the target user and improve the success rate of the push. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 A schematic diagram of a demand response system provided in an embodiment of the present application;

[0024] Figure 2A flowchart of a method for determining a demand response user provided in an embodiment of the present application;

[0025] Figure 3 A training diagram of a target user response probability prediction model provided in an embodiment of the present application;

[0026] Figure 4 A schematic diagram of the structure of a linear model provided in an embodiment of the present application;

[0027] Figure 5 A schematic diagram of the structure of a deep neural network model provided in an embodiment of the present application;

[0028] Figure 6 A schematic diagram of an application of a target user response probability prediction model provided in an embodiment of the present application;

[0029] Figure 7 A schematic diagram of a demand response user determination device provided in an embodiment of the present application;

[0030] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0032] The terms "first", "second", "third" and "fourth" etc. in the specification and claims of the present application and the drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices.

[0033] Reference to "embodiments" herein means that a particular feature, result, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0034] First, see Figure 1 , Figure 1 A schematic diagram of a demand response system provided in an embodiment of the present application. The demand response system includes: a power grid, a demand response user determination device, a demand side, and users corresponding to the demand side.

[0035] In an embodiment of the present application, there may be multiple demand sides in the demand response system, for example, household electricity demand side, enterprise electricity demand side, and so on. Figure 1 Only two demand sides are shown, and the present application does not limit the number of demand sides. Each demand side may also correspond to multiple users. The present application embodiment only takes one user corresponding to each demand side as an example for explanation, and does not limit the number of users. The power grid is a power supply network in the power system, which is mainly used to deliver the reserved electric energy to each demand side to meet the power demand of the electrical equipment on the demand side. For example, the power grid delivers the reserved electric energy to the household power demand side to meet the power demand of the user's electrical equipment such as air conditioners, water heaters, etc.

[0036] In the embodiment of the present application, the demand response user determination device may include: a central processing unit (CPU), a microcontroller unit (MCU), a single chip microcomputer (Single Chip Microcomputer), a digital signal processing (DSP) chip, a microprocessor (MPU), and other electronic devices with data storage, data processing, and communication functions. The demand response user determination device may also be a power grid operator.

[0037] Among them, the demand response user determination device can be used to monitor the power supply load of the power grid and the power load of each demand side. When the power grid is in a high load state, the demand response user determination device can push demand events to the users corresponding to each demand side. For example, the demand event can be a power compensation for saving electricity. After receiving the demand event pushed by the demand response user determination device, the user responds to the demand event and adjusts the power consumption behavior and power consumption mode. For example, in response to the demand event, the user turns off the air conditioner, water heater, and other electrical equipment or adjusts the output power of the electrical equipment, thereby reducing the power load on the demand side corresponding to the user, so as to reduce the high load state of the power grid and avoid the failure of the power grid under a long-term high load state. When the power grid is in a low load state, the demand response user determination device can also push demand events to the users corresponding to each demand side. At this time, the demand event can be a power preferential rule, and the user can adjust the power consumption behavior and power consumption mode in response to the demand event. For example, the user can increase the power of the electrical equipment to increase the power load on the demand side corresponding to the user, thereby balancing the load state of the power grid.

[0038] It should be noted that after receiving the demand event pushed by the demand response user determination device, the user can respond to the demand event or not. Therefore, when the load state of the power grid is in a high load state or a low load state, the same demand event is pushed to all users, and the number of users who actually respond to the demand event is small, and the power regulation effect is not ideal. In addition, the probability of each user on the demand side responding to different demand events is different. Pushing demand events to all users will cause a waste of resources and a low success rate of pushing.

[0039] To this end, applied to the above-mentioned demand response system, in an embodiment of the present application, the demand response user determination device obtains a plurality of target historical demand events that match the target demand event;

[0040] The demand response user determination device determines a candidate user set corresponding to the target demand event based on the historical response information of the user to each target historical demand event;

[0041] The demand response user determination device obtains basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events in the candidate user set;

[0042] The demand response user determination device inputs the basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event;

[0043] The demand response user determination device determines a plurality of target users corresponding to the target demand event from a set of candidate users based on the response probability of each candidate user to the target demand event.

[0044] It can be seen that in the embodiment of the present application, applied to the above-mentioned demand response system, the demand response user determination device can monitor the load state of the power grid. When the load state of the power grid is a high load state or a low load state, the target demand event is pushed to the target user so that the target user responds to the target demand event and adjusts the power consumption mode of the target user, thereby adjusting the load state of the power grid and improving the stability of the power grid. Among them, the demand response user determination device can predict the response probability of each candidate user to the target demand event through the user basic information, power consumption behavior information, event characteristics of the target demand event and the historical response information of each candidate user, so as to determine multiple target users, thereby accurately pushing the target demand event to the target user and improving the success rate of the push.

[0045] See also Figure 2 , Figure 2 A flowchart of a method for determining a demand response user provided in an embodiment of the present application. The method is applied to the above-mentioned demand response user determination device. The method includes but is not limited to the following steps:

[0046] 201: Acquire multiple target historical demand events that match the target demand event.

[0047] In the embodiment of the present application, the target demand event is a demand event that the demand response user determination device plans to push to the target user. The multiple target historical demand events are multiple historical demand events with the highest similarity to the target demand event among the multiple historical demand events. Optionally, the target demand event can be, for example, electricity compensation, and the target historical demand event can include, for example: electricity discount, price adjustment, recommendation information, etc.

[0048] Among them, the demand response user determination device can establish an event similarity matrix and obtain multiple target historical demand events with the highest similarity to the target demand event from the event similarity matrix, that is, multiple target historical demand events that match the target demand event.

[0049] Exemplarily, obtaining multiple target historical demand events that match the target demand event may include, for example:

[0050] If there is a target demand event among multiple historical demand events, then based on the event similarity matrix, multiple target historical demand events matching the target demand event are obtained;

[0051] If the target demand event does not exist in the multiple historical demand events, then based on the event characteristics of the target demand event and the event characteristics of each historical demand event, determine the similarity between the target demand event and each historical demand event;

[0052] Based on the similarity between the target demand event and each historical demand event, determine the similar historical demand event with the highest similarity to the target demand event;

[0053] Based on the event similarity matrix, multiple target similar historical demand events matching similar historical demand events are obtained; and the multiple target similar historical demand events are used as multiple target historical demand events.

[0054] In an embodiment of the present application, each element in the event similarity matrix represents the similarity between two historical demand events corresponding to the element, and the similarity can be determined, for example, by the number of responses of each user to the two historical demand events.

[0055] Specifically, the similarity between any two historical demand events can be expressed by formula (1):

[0056]

[0057] Among them, S ab Represents the similarity between historical demand event a and historical demand event b. ai represents the number of responses of user i to historical demand event a, R bi Represents the number of responses of user i to historical demand event b. n represents the number of users. The demand response user determination device can determine the similarity between any two historical demand events by obtaining the number of responses of each user to each historical demand event and combining it with formula (1). Then, all the determined similarities are spliced ​​to obtain the event similarity matrix shown below:

[0058]

[0059] It should be noted that the embodiment of the present application is only described by taking the example of determining the similarity of any two historical demand events based on the number of responses of each user to each historical demand event. Those skilled in the art can also determine the similarity of any two historical demand events based on other event features of each historical demand event, such as event type, response time, etc. The method of determining the similarity based on other event features is similar to that of the embodiment of the present application and will not be repeated here.

[0060] Based on this, when obtaining multiple target demand events that match the target demand event, if there is a target demand event among multiple historical demand events, the demand response user determination device obtains multiple target historical demand events with the highest similarity to the target demand event from the event similarity matrix. For example, the similarity between each historical demand event and the target demand event is obtained from the event similarity matrix, and the similarity between each historical demand event and the target demand event is sorted in descending order to obtain a sorting result. The historical demand events corresponding to the first 10 similarities in the sorting result are used as multiple target historical demand events.

[0061] It should be noted that the event similarity matrix includes the similarity between each historical demand event and itself, so multiple target historical demand events may include the target demand event.

[0062] If the target demand event does not exist among the multiple historical demand events, the demand response user determination device may determine the similarity between the target demand event and each historical demand event based on the event characteristics of the target demand event and the event characteristics of each historical demand event. Optionally, the event characteristics may include: the number of responses of each user to the demand event corresponding to the event characteristics, the response duration, the response time period, the total response load capacity, and the response event type, event number, response compensation, etc. Among them, the method for determining the similarity between the target demand event and each historical demand event based on the event characteristics of the target demand event and the event characteristics of each historical demand event is similar to the above-mentioned method for determining the similarity between any two historical demand events, and will not be repeated here.

[0063] Then, the demand response user determination device determines the similar historical demand event with the highest similarity to the target demand event from the multiple historical demand events. Further, the similarity between each historical demand event and the similar historical demand event is obtained from the event similarity matrix. The similarity between each historical demand event and the similar historical demand event is sorted in descending order, and the multiple target similar historical demand events with the highest sorting results are used as multiple target historical demand events.

[0064] It can be seen that in the embodiment of the present application, when there is a target demand event among multiple historical demand events, the demand response user determination device can obtain multiple target historical demand events that match the target demand event from the event similarity matrix. When there is no target demand event among multiple historical demand events, the demand response user determination device can determine the similar historical demand event with the highest similarity to the target demand event from multiple historical demand events. The multiple target similar historical demand events with the highest similarity to the similar historical demand events in the event similarity matrix are taken as multiple target historical demand events. In this way, multiple target historical demand events that match the target demand event can be obtained, and the candidate user set can be screened through the user's historical response information to the multiple target historical demand events, so that the determined target users are more accurate, thereby improving the success rate of the push.

[0065] 202: Determine a candidate user set corresponding to the target demand event based on the user's historical response information to each target historical demand event.

[0066] In the embodiment of the present application, the historical response information includes: the number of responses, the average duration of the response, the load of the response, etc. Among them, the demand response user determination device can determine the candidate user set corresponding to the target demand event based on the number of responses of the user to multiple target historical demand events.

[0067] It should be noted that the embodiment of the present application only uses the historical response information as the number of responses as an example for explanation. The processing method of historical response information as other information is similar to the embodiment of the present application and will not be repeated here.

[0068] Exemplarily, based on the historical response information of the user to each target historical demand event, determining the candidate user set corresponding to the target demand event may include, for example:

[0069] Taking users whose number of responses to each target historical demand event is greater than a first threshold as candidate users for each target historical demand event;

[0070] The candidate users of each target historical demand event are merged and deduplicated to obtain a candidate user set corresponding to the target demand event.

[0071] Specifically, the demand response user determination device determines the users whose number of responses to each target historical demand event is greater than a first threshold from all users on the demand side based on the number of responses of the users to each target historical demand event. The first threshold can be pre-set based on the screening effect of the candidate user set. For example, based on the number of responses of the users to each of the 10 target historical demand events, the users whose number of responses to each target historical demand event is greater than 2 times are used as candidate users corresponding to each target historical demand event, and all candidate users corresponding to the 10 target historical demand events are obtained.

[0072] Then, the demand response user determination device merges and removes duplicates of the candidate users of each target historical demand event, that is, when the number of responses of a candidate user to any two target historical demand events is greater than the first threshold, only one basic user information of the candidate user is retained in the candidate user set. Based on this, the candidate user set corresponding to the target demand event can be determined based on the number of responses of the user to each target historical demand event.

[0073] Optionally, the demand response user determination device can also select users whose total response times to multiple target historical demand events are greater than a second threshold, or users whose average response times to multiple target historical demand events are greater than a third threshold, as candidate users in the candidate user set based on the total response times or average response times of users to multiple target historical demand events. The second threshold and the third threshold can be pre-set based on the screening effect of the candidate user set. Thus, based on the total response times or average response times of users to multiple target historical demand events, the historical response information of users to demand events similar to the target demand events can be more accurately reflected to determine a candidate user set with higher accuracy.

[0074] It can be seen that in the embodiment of the present application, the demand response user determination device can use users whose number of responses to each target historical demand event is greater than the first threshold as candidate users for each target historical demand event. Then, the candidate users of each target historical demand event are merged and deduplicated to obtain a candidate user set corresponding to the target demand event. Thus, through the number of responses of the user to each target historical demand event, each candidate user in the candidate user set is determined to be a user with a high probability of responding to the target demand event, so as to determine more accurate target users and improve the success rate of push.

[0075] 203: Obtain basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events in the candidate user set.

[0076] In an embodiment of the present application, basic user information may include: user number, industry, electricity type, electricity capacity, demand response equipment, self-provided energy storage information, self-provided power plant information, number of user responses, average response time, etc. Electricity behavior information may include an electricity load curve, which includes the electricity load at each moment. Event features of target demand events may include: event number, total response load capacity, response time period, response time, response event type, response compensation, etc. Historical response information includes: average response time, response load, number of responses, etc.

[0077] 204: Inputting each candidate user's basic information, electricity consumption behavior information, event characteristics of the target demand event, and each candidate user's historical response information to multiple historical demand events into the target user response probability prediction model to obtain each candidate user's response probability to the target demand event.

[0078] In an embodiment of the present application, before inputting each candidate user's basic information, electricity usage behavior information, event characteristics of the target demand event, and each candidate user's historical response information to multiple historical demand events into the target user response probability prediction model, data preprocessing and feature crossing are required to obtain a feature vector that can be input into the target user response probability prediction model.

[0079] Exemplarily, before inputting the basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model, the following further includes:

[0080] Performing data preprocessing on basic user information, electricity consumption behavior information, event characteristics of target demand events, and historical response information of each candidate user to multiple historical demand events to obtain multiple numerical response characteristics and multiple discrete response characteristics corresponding to each candidate user;

[0081] Feature crossover is performed on multiple numerical response features and multiple discrete response features to obtain multiple cross response features.

[0082] In the embodiment of the present application, the numerical response characteristics may include, for example, the power consumption capacity, the number of responses, the average response time, the total response load capacity, the response time period, the response time, the response load amount, etc. The discrete response characteristics may include, for example, the user number, the industry, the type of power consumption, the self-provided energy storage information, the self-provided power plant information, the event number, the response event type, the response compensation, etc. The cross-response characteristics may include, for example, the cross-characteristics of the user number and the industry, the cross-characteristics of the power consumption type and the power consumption capacity, etc.

[0083] Specifically, the demand response user determination device first performs data preprocessing on each candidate user's basic user information, electricity consumption behavior information, event characteristics of the target demand event, and each candidate user's historical response information to multiple historical demand events. Among them, in the data preprocessing, the demand response user determination device first performs data cleaning on the above-mentioned feature data to remove abnormal values ​​in the above-mentioned feature data and fill in the missing values ​​in the above-mentioned feature data. For example, when the electricity consumption capacity jumps at a specific moment, the abnormal value of the jump needs to be removed. When filling in missing values, the numerical response characteristics can be filled by the average value or the median. For example, when the response time of a user in a specific time period is missing, the missing response time can be filled by the average response time of each time period or the median response time of each time period.

[0084] Next, the demand response user determination device will standardize or normalize the numerical response features so that the numerical response features can be input into the target user response probability prediction model. The discrete response features are encoded, for example, using unique hot encoding or label encoding, so that the discrete response features can be input into the target user response probability prediction model. Thus, the demand response user determination device can obtain multiple pre-processed numerical response features and multiple discrete response features.

[0085] Furthermore, the demand response user determination device may perform feature crossover on multiple numerical response features and multiple discrete response features to obtain multiple crossover features. For example, crossover features between the industry and the electricity consumption category, crossover features between the electricity consumption type and the electricity consumption capacity, etc. The crossover features between the industry and the electricity consumption category may include, for example, crossover features between agricultural and sideline food processing industry and residents' life, crossover features between textile industry and large industrial electricity consumption, crossover features between wholesale and retail industry and commerce, etc.

[0086] It should be noted that the cross-response feature may be a cross-feature between a numerical response feature and a numerical response feature, a cross-feature between a numerical response feature and a discrete response feature, or a cross-feature between a discrete response feature and a discrete response feature.

[0087] It can be seen that the demand response user determination device can obtain multiple numerical response features and multiple discrete response features by performing data preprocessing on the user basic information, electricity consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events. Then, multiple numerical response features and multiple discrete response features are cross-featured to obtain multiple cross-response features. Thus, multiple numerical response features and multiple discrete response features can be input into the target user response probability prediction model to predict the candidate user's response probability to the target demand event, thereby determining the target user, and pushing the target demand event to the target user can improve the success rate of the push.

[0088] Then, the multiple numerical response features, the multiple discrete response features, and the multiple cross-response features are input into the target user response probability prediction model to predict the response probability of each candidate user to the target demand event. The target user response probability prediction model is obtained by pre-training. Exemplarily, the method further includes:

[0089] Acquire multiple training samples, each sample pair including basic user information, power consumption behavior information, event features of sample demand events, historical response information of the sample user to multiple sample historical demand events, and sample labels;

[0090] The model is trained based on multiple training samples to obtain a target user response probability prediction model.

[0091] In an embodiment of the present application, the sample label is used to indicate the actual response of the sample user to the sample demand event. Specifically, the demand response user determination device will first obtain multiple training samples from the database. Based on the basic user information, electricity consumption behavior information, event characteristics of the sample demand event, sample discrete response characteristics and sample cross-response characteristics in the historical response information of the sample user to multiple sample historical demand events of each training sample, the third response probability is determined. Based on the numerical response characteristics and discrete response characteristics corresponding to the sample user of each training sample, the fourth response probability is determined, and based on the third response probability and the fourth response probability, the response probability of each sample user to the sample demand event is obtained. Finally, model training is performed based on the response probability of each sample user to the sample demand event and the sample label of each training sample to obtain a target user response probability prediction model.

[0092] Exemplarily, model training is performed based on multiple training samples to obtain a target user response probability prediction model, which may include:

[0093] Obtaining basic user information, electricity consumption behavior information, event characteristics of sample demand events, feature vectors corresponding to each sample numerical response feature in historical response information of the sample user to multiple sample historical demand events, feature vectors corresponding to each sample discrete response feature, and feature vectors corresponding to each sample cross-response feature of each sample user;

[0094] The feature vector corresponding to each sample discrete response feature and the feature vector corresponding to each sample cross-response feature are concatenated to obtain a sample response feature matrix corresponding to each sample user; based on the response feature matrix corresponding to each sample user, the third response probability of each sample user to the sample demand event is determined;

[0095] Perform feature conversion on each sample discrete response feature to obtain a sample embedding vector corresponding to each sample discrete response feature; forward propagate the feature vector corresponding to each sample numerical response feature and each sample embedding vector to obtain the fourth response probability of each sample user responding to the sample demand event;

[0096] Based on the third response probability and the fourth response probability of each sample user responding to the sample demand event, obtaining the response probability of each sample user to the sample demand event;

[0097] Based on the response probability of each sample user to the sample demand event and the sample label of each training sample, determine the model loss corresponding to each training sample;

[0098] Based on the model loss corresponding to each training sample, the model parameters are updated to perform model training and obtain the target user response probability prediction model.

[0099] In an embodiment of the present application, the demand response user determination device first needs to perform feature conversion on the sample numerical response features, sample discrete response features, and sample cross-response features in multiple training samples to obtain a feature vector corresponding to each sample numerical response feature, a feature vector corresponding to each sample discrete response feature, and a feature vector corresponding to each sample cross-response feature. Among them, the sample cross-response feature is obtained by cross-feature cross-reaction of the sample numerical response feature and the sample discrete response feature. The method for determining the sample cross-response feature is similar to the method for determining the cross-response feature in the embodiment of the present application, and will not be repeated here.

[0100] See also Figure 3 , Figure 3 A training diagram of a target user response probability prediction model provided in an embodiment of the present application. The demand response user determination device will splice the feature vector corresponding to each sample discrete response feature and the feature vector corresponding to each sample cross response feature into a sample response feature matrix. Figure 3 The “+” operation shown here concatenates the feature vector corresponding to each sample discrete response feature and the feature vector corresponding to each sample cross response feature, and inputs the sample response feature matrix into the linear model. The linear model can be expressed by formula (2):

[0101] y1=w T x+b Formula (2)

[0102] In formula (2), y1 represents the third response probability of each sample user to the sample demand event. w represents the parameter matrix [w1, w2, ..., wn] formed by splicing the n model parameters of the linear model, where w1, w2, ..., wn are the n model parameters of the linear model. x represents the sample response feature matrix [x1, x2, ..., xn] corresponding to each sample user, where x1, x2, ..., xn are the eigenvectors in the sample response feature matrix. b represents the offset of the linear model. The structure of the linear model is as follows: Figure 4 As shown, the characteristic vector in the sample response characteristic matrix corresponding to each sample user is subjected to linear operation of formula (2) to obtain the third response probability of each sample user to the sample demand event.

[0103] It should be noted that the parameter matrix and offset of the linear model are both model parameters of the linear model. During the model training process, by continuously updating the model parameters of the linear model, the linear model can better determine the probability of the third response of each sample user to the sample demand event.

[0104] Furthermore, if Figure 3 As shown, the demand response user determination device performs feature conversion on each sample discrete response feature to obtain a sample embedding vector corresponding to each sample discrete response feature. The feature conversion of each sample discrete response feature can be performed, for example, by an embedding layer to convert each sample discrete response feature into a dense sample embedding vector. The feature vector corresponding to each sample numerical response feature and each sample embedding vector can be input into a deep neural network model for forward propagation to obtain the fourth response probability of each sample user responding to the sample demand event. Optionally, the deep neural network model can be a deep feedforward neural network model.

[0105] Among them, the structure of the deep neural network model is as follows Figure 5 The input layer is used to input the sample numerical response features and sample discrete response features corresponding to each training sample. Figure 5 The x1, ..., xm shown in the figure are the sample numerical response features corresponding to each training sample, and xm+1, ..., xn are the sample discrete response features corresponding to each training sample. The embedding layer converts the features of each sample discrete response feature to obtain the sample embedding vector corresponding to each sample discrete response feature. Among them, each sample discrete feature can correspond to two sample embedding vectors. Figure 5As shown, the two sample embedding vectors corresponding to xm+1 are pm+1 and qm+1, and the two sample embedding vectors corresponding to xn are pn and qn. The demand response user determination device inputs the feature vector corresponding to each sample numerical response feature and each sample embedding vector into the hidden layer for forward propagation, so as to obtain the response probability of each sample user to the sample demand event in the output layer. The number of hidden layers k and the number of activation values ​​c of each hidden layer can be preset based on the prediction effect of the response probability. Figure 5 As shown, the elements of the first hidden layer are a1(1),…,ac(1), and the elements of the kth hidden layer are a1(k),…,ac(k).

[0106] Among them, the activation value of the kth hidden layer can be expressed by formula (3):

[0107] a(k)=f[W(k-1 ) a(k-1 )+b(k-1)] Formula (3)

[0108] In formula (3), a(k) represents the activation value matrix of the k-th hidden layer [a1(k), ..., ac(k)], and a(k-1) is the activation value of the k-1th hidden layer. f[] represents the activation function, wherein the activation function can be, for example, the ReLUs function. W(k-1) is the weight parameter matrix of the k-1th hidden layer [W1, W2, ..., Wc]. b(k-1) represents the offset matrix of the k-1th hidden layer [b1, b2..., bn]. Among them, the weight parameter matrix and offset matrix of each hidden layer are the model parameters of the deep neural network model. During model training, by continuously updating the model parameters of the deep neural network model, the accuracy of the prediction of the fourth response probability of each sample user in response to the sample demand event can be improved. After forward propagation through all hidden layers, the demand response user determination device can obtain the fourth response probability y2 of each sample user in response to the sample demand event at the output layer.

[0109] Furthermore, the demand response user determination device performs a weighted summation of the third response probability and the fourth response probability of each sample user responding to the sample demand event, such as Figure 3 The "%+" operation shown in the figure obtains the response probability of each sample user to the sample demand event. The response probability of each sample user to the sample demand event can be expressed by formula (4):

[0110]

[0111] Wherein, P(Y=1|x) represents the response probability of each sample user to the sample demand event. σ represents the activation function, for example, the activation function can be a Sigmoid function. W1 T Represents the weight of the linear model, W2T represents the weight of the deep neural network model. y1 represents the third response probability of each sample user to the sample demand event, and y2 represents the fourth response probability of each sample user to the sample demand event. bp represents the offset of the weighted sum. Among them, W1 T 、W2 T and bp as model parameters of the target user response probability prediction model can be pre-set. During the model training process, the demand response user determination device can continuously update the model parameters based on the model loss to obtain a user response probability prediction model with better training effect.

[0112] After obtaining the response probability of each sample user to the sample demand event, the demand response user determination device can determine the model loss corresponding to each training sample based on the response probability of each sample user to the sample demand event and the sample label of each training sample. Optionally, the model loss can be determined by a cross entropy loss function, where the model loss can be expressed by formula (5):

[0113] loss i =r i logt i +(1-r i )log(1-t i ) Formula (5)

[0114] Among them, loss i Represents the model loss corresponding to training sample i. i Represents the sample label of training sample i. i represents the response probability of the sample user of training sample i to the sample demand event. Based on the above formula (5), the model loss corresponding to each training sample can be determined.

[0115] Thus, the demand response user determination device updates the model parameters based on the model loss corresponding to each training sample. The model parameters include the model parameters of the linear model, the model parameters of the deep neural network model, and the model parameters of the target user response probability prediction model. When the model loss corresponding to each training sample is less than the fourth threshold, or the number of iterations of the model training is greater than the fifth threshold, it means that the model has good performance, and the demand response user determination device will save the model parameters at this time to obtain the target user response probability prediction model.

[0116] It can be seen that in the embodiment of the present application, the demand response user determination device can obtain the sample response feature matrix corresponding to each sample user by splicing the feature vector of the sample discrete response feature and the feature vector of the sample cross response feature in each training sample, and input the sample response feature matrix corresponding to each sample user into the linear model to obtain the third response probability of each sample user to the sample demand event. Then, the discrete response feature in each training sample is converted into an embedding vector, and forward propagated with the feature vector of the numerical response feature to obtain the fourth response probability of each sample user to the sample demand event. Further, the third response probability and the fourth response probability of each sample user to the sample demand event are weighted and summed to obtain the response probability of each sample user to the sample demand event. Then, the model loss is calculated for the response probability of each sample user to the sample demand event and the sample label of each training sample, and based on the model loss, the model parameters are updated to train the target user response probability prediction model. Thus, the target user response probability prediction model obtained has good performance, and can effectively determine the target user with a high probability of response to the target demand event to improve the success rate of push.

[0117] Thus, the demand response user determination device can determine the response probability of each candidate in the candidate user set to the target demand event through the above-mentioned pre-trained target user response probability prediction model. Exemplarily, the user basic information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events are input into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event, including:

[0118] Obtaining basic user information, power consumption behavior information, event features of the target demand event, and feature vectors corresponding to each numerical response feature, each discrete response feature, and each cross-response feature in historical response information of each candidate user to multiple historical demand events;

[0119] The feature vector corresponding to each discrete response feature and the feature vector corresponding to each cross-response feature are concatenated to obtain a response feature matrix corresponding to each candidate user; based on the response feature matrix corresponding to each candidate user, the first response probability of each candidate user to the target demand event is determined;

[0120] Perform feature conversion on each discrete response feature to obtain the embedding vector corresponding to each discrete response feature; forward propagate the feature vector corresponding to each numerical response feature and each embedding vector to obtain the second response probability of each candidate user responding to the target demand event;

[0121] Based on the first response probability and the second response probability of each candidate user in responding to the target demand event, the response probability of each candidate user to the target demand event is obtained.

[0122] Specifically, Figure 6 As shown, Figure 6 The user basic information, electricity consumption behavior information, event characteristics, and historical response information in the above are the user basic information, electricity consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events. The demand response user determination device can splice the feature vector corresponding to each discrete response feature and the feature vector corresponding to each cross response feature in the above features, such as Figure 6 The "+" operation shown in the figure obtains the response feature matrix corresponding to each candidate user. The response feature matrix corresponding to each candidate user is input into the linear model, wherein the processing method of the response feature matrix in the linear model is similar to any embodiment of the present application, and will not be repeated here.

[0123] Therefore, the demand response user determination device can process the response feature matrix corresponding to each candidate user through a linear model to obtain the probability of each candidate user's first response to the target demand event.

[0124] Then, the demand response user determination device converts each discrete response feature into an embedding vector corresponding to each discrete response feature. It should be noted that the method of converting each discrete response feature into an embedding vector corresponding to each discrete response feature is similar to the above-mentioned method of converting each sample discrete response feature into a sample embedding vector corresponding to each sample discrete response feature, and will not be repeated here. The feature vector corresponding to each numerical response feature and each embedding vector are forward propagated and input into the deep neural network model for forward propagation to obtain the second response probability of each candidate user responding to the target demand event. It should be noted that the deep neural network model processes discrete response features and embedding vectors in a similar way to the above-mentioned processing of sample discrete response features and sample embedding vectors, and will not be repeated here.

[0125] Therefore, the demand response user determination device can process the feature vector corresponding to each numerical response feature and each embedded vector through a deep neural network model to obtain the second response probability of each candidate user to the target demand event.

[0126] Furthermore, a weighted sum is performed on the first response probability and the second response probability of each candidate user to the target demand event, such as Figure 6The "%+" operation shown in the figure obtains the response probability of each candidate user to the target demand event. Among them, the method of weighted summing the first response probability and the second response probability of each candidate user to the target demand event is similar to the method of weighted summing the third response probability and the fourth response probability of each sample user to the sample demand event in the above embodiment, which will not be repeated here. In this way, the response probability of each candidate user to the target demand event can be determined.

[0127] It can be seen that the demand response user determination device applied to the above-mentioned pre-trained target user response probability prediction model can predict the response probability of each candidate user to the target demand event, so as to determine the target user with a high response probability and push the target demand event to the target user, which is conducive to improving the success rate of the push.

[0128] 205: Based on the response probability of each candidate user to the target demand event, determine a plurality of target users corresponding to the target demand event from the candidate user set.

[0129] In the embodiment of the present application, the demand response user determination device can sort the response probability of each candidate user in the candidate user set to the target demand event in descending order, and use the first N candidate users in the sorting result as multiple target users corresponding to the target demand event. The target users can be pre-set based on the actual push demand.

[0130] Optionally, when determining the target user, it is necessary to obtain the total load response capacity corresponding to each candidate user. The N candidate users whose total load response capacity is greater than the sixth threshold and whose ranking results are high are taken as multiple target users corresponding to the target demand event.

[0131] Finally, the demand response user determination device pushes the target demand event to multiple target users to adjust the load state of the power grid through the responses of the multiple target users to the target demand event.

[0132] In summary, in an embodiment of the present application, first, by obtaining multiple target historical demand events that match the target demand event, based on the user's historical response information to each target historical demand event, a candidate user set corresponding to the target demand event is determined. Then, the user basic information, electricity usage behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events of each candidate user in the candidate user set are obtained. Next, the user basic information, electricity usage behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events are input into a pre-trained target user response probability prediction model to obtain the response probability of each candidate user to the target demand event. Finally, based on the response probability of each candidate user to the target demand event, multiple target users corresponding to the target demand event are determined from the candidate user set. Thus, the candidate user set can be determined through the user's historical response information to each target historical demand event, and the response probability of each candidate user to the target demand event can be predicted based on each candidate user's basic information, electricity consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events, so as to determine multiple target users with a higher probability of responding to the target demand event. The target user response probability prediction model can effectively predict the response probability of each candidate user to the target demand event by learning data such as the user's basic information, electricity consumption behavior information, event characteristics of the target demand event, and historical response information, thereby improving the accuracy of determining the demand response user, so as to accurately push the target demand event to the target user and improve the success rate of the push.

[0133] See also Figure 7 , Figure 7 A schematic diagram of a demand response user determination device provided in an embodiment of the present application. The demand response user determination device 700 includes an acquisition unit 701 and a processing unit 702 .

[0134] An acquisition unit 701 is used to acquire multiple target historical demand events that match the target demand event;

[0135] Determine a candidate user set corresponding to the target demand event based on the user's historical response information to each target historical demand event;

[0136] Obtaining basic user information, power consumption behavior information, event characteristics of a target demand event, and historical response information of each candidate user to multiple historical demand events for each candidate user in the candidate user set;

[0137] The processing unit 702 is used to input the basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event;

[0138] Based on the response probability of each candidate user to the target demand event, a plurality of target users corresponding to the target demand event are determined from the candidate user set.

[0139] In a feasible embodiment, in inputting the basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event, the processing unit 702 is specifically used to:

[0140] Obtaining basic user information, power consumption behavior information, event features of the target demand event, and feature vectors corresponding to each numerical response feature, each discrete response feature, and each cross-response feature in historical response information of each candidate user to multiple historical demand events;

[0141] The feature vector corresponding to each discrete response feature and the feature vector corresponding to each cross-response feature are concatenated to obtain a response feature matrix corresponding to each candidate user; based on the response feature matrix corresponding to each candidate user, the first response probability of each candidate user to the target demand event is determined;

[0142] Perform feature conversion on each discrete response feature to obtain the embedding vector corresponding to each discrete response feature; forward propagate the feature vector corresponding to each numerical response feature and each embedding vector to obtain the second response probability of each candidate user responding to the target demand event;

[0143] Based on the first response probability and the second response probability of each candidate user in responding to the target demand event, the response probability of each candidate user to the target demand event is obtained.

[0144] In a feasible embodiment, in terms of acquiring multiple target historical demand events matching the target demand event, the acquisition unit 701 is specifically configured to:

[0145] If there is a target demand event among multiple historical demand events, then based on the event similarity matrix, multiple target historical demand events matching the target demand event are obtained; wherein the event similarity matrix includes the similarity between any two historical demand events, and the similarity is determined by the number of responses of users to the two historical demand events corresponding to the similarity;

[0146] If the target demand event does not exist in the multiple historical demand events, then based on the event characteristics of the target demand event and the event characteristics of each historical demand event, determine the similarity between the target demand event and each historical demand event;

[0147] Based on the similarity between the target demand event and each historical demand event, determine the similar historical demand event with the highest similarity to the target demand event;

[0148] Based on the event similarity matrix, multiple target similar historical demand events matching similar historical demand events are obtained; and the multiple target similar historical demand events are used as multiple target historical demand events.

[0149] In a feasible embodiment, the historical response information includes the number of responses;

[0150] In terms of determining a candidate user set corresponding to a target demand event based on historical response information of the user to each target historical demand event, the acquisition unit 701 is specifically configured to:

[0151] Taking users whose number of responses to each target historical demand event is greater than a first threshold as candidate users for each target historical demand event;

[0152] The candidate users of each target historical demand event are merged and deduplicated to obtain a candidate user set corresponding to the target demand event.

[0153] In a feasible embodiment, before inputting the basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model, the processing unit 702 is further used to:

[0154] Performing data preprocessing on basic user information, electricity consumption behavior information, event characteristics of target demand events, and historical response information of each candidate user to multiple historical demand events to obtain multiple numerical response characteristics and multiple discrete response characteristics corresponding to each candidate user;

[0155] Feature crossover is performed on multiple numerical response features and multiple discrete response features to obtain multiple cross response features.

[0156] In a feasible embodiment, the processing unit 702 is further configured to:

[0157] Acquire multiple training samples, each sample pair includes basic user information, power consumption behavior information, event characteristics of a sample demand event, historical response information of the sample user to multiple sample historical demand events, and a sample label, wherein the sample label is used to indicate the actual response of the sample user to the sample demand event;

[0158] The model is trained based on multiple training samples to obtain a target user response probability prediction model.

[0159] In a feasible embodiment, in performing model training based on multiple training samples to obtain a target user response probability prediction model, the processing unit 702 is specifically configured to:

[0160] Obtaining basic user information, electricity consumption behavior information, event characteristics of sample demand events, feature vectors corresponding to each sample numerical response feature in historical response information of the sample user to multiple sample historical demand events, feature vectors corresponding to each sample discrete response feature, and feature vectors corresponding to each sample cross-response feature of each sample user;

[0161] The feature vector corresponding to each sample discrete response feature and the feature vector corresponding to each sample cross-response feature are concatenated to obtain a sample response feature matrix corresponding to each sample user; based on the response feature matrix corresponding to each sample user, the third response probability of each sample user to the sample demand event is determined;

[0162] Perform feature conversion on each sample discrete response feature to obtain a sample embedding vector corresponding to each sample discrete response feature; forward propagate the feature vector corresponding to each sample numerical response feature and each sample embedding vector to obtain the fourth response probability of each sample user responding to the sample demand event;

[0163] Based on the third response probability and the fourth response probability of each sample user responding to the sample demand event, obtaining the response probability of each sample user to the sample demand event;

[0164] Based on the response probability of each sample user to the sample demand event and the sample label of each training sample, determine the model loss corresponding to each training sample;

[0165] Based on the model loss corresponding to each training sample, the model parameters are updated to perform model training and obtain the target user response probability prediction model.

[0166] See also Figure 8 , Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 8As shown, the electronic device 800 includes a transceiver 801, a processor 802, and a memory 803. They are connected via a bus 804. The memory 803 is used to store computer programs and data, and can transmit the data stored in the memory 803 to the processor 802. The electronic device 800 can be a demand response user determination device 700. The transceiver 801 can be an acquisition unit 701, and the processor 802 can be a processing unit 702.

[0167] The processor 802 is used to read the computer program in the memory 803 and perform the following operations:

[0168] Acquire multiple target historical demand events that match the target demand event;

[0169] Determine a candidate user set corresponding to the target demand event based on the user's historical response information to each target historical demand event;

[0170] Obtaining basic user information, power consumption behavior information, event characteristics of a target demand event, and historical response information of each candidate user to multiple historical demand events for each candidate user in the candidate user set;

[0171] Input the basic information of each candidate user, electricity consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event;

[0172] Based on the response probability of each candidate user to the target demand event, a plurality of target users corresponding to the target demand event are determined from the candidate user set.

[0173] The above mainly introduces the scheme of the embodiment of the present application from the perspective of the execution process of the method side. It is understandable that the electronic device 800 includes a hardware structure and / or software module corresponding to each function in order to realize the above functions. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiment provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present application.

[0174] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. The computer program is executed by a processor to implement part or all of the steps of any demand response user determination method recorded in the above method embodiments.

[0175] An embodiment of the present application also provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute part or all of the steps of any one of the demand response user determination methods recorded in the above method embodiments.

[0176] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0177] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0178] In the several embodiments provided in the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the device or unit can be electrical or other forms.

[0179] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0180] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of a software program module.

[0181] If the integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory, including a number of instructions to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or CD-ROM and other media that can store program codes.

[0182] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable memory, and the memory can include: a flash drive, a read-only memory (English: Read-Only Memory, abbreviated as: ROM), a random access memory (English: Random Access Memory, abbreviated as: RAM), a magnetic disk or an optical disk, etc.

[0183] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for determining a demand response user, It is characterized in that include: Acquire multiple target historical demand events that match the target demand event; Based on the historical response information of the user to each target historical demand event, determining a candidate user set corresponding to the target demand event; Obtaining basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events for each candidate user in the candidate user set; Inputting the basic information of each candidate user, the electricity consumption behavior information, the event characteristics of the target demand event, and the historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event; Based on the response probability of each candidate user to the target demand event, a plurality of target users corresponding to the target demand event are determined from the candidate user set.

2. The method according to claim 1, It is characterized in that The basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events are input into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event, including: Obtaining basic user information, power consumption behavior information, event features of the target demand event, and feature vectors corresponding to each numerical response feature, each discrete response feature, and each cross-response feature in historical response information of each candidate user to multiple historical demand events; The feature vector corresponding to each discrete response feature and the feature vector corresponding to each cross-response feature are concatenated to obtain a response feature matrix corresponding to each candidate user; based on the response feature matrix corresponding to each candidate user, the first response probability of each candidate user to the target demand event is determined; Perform feature conversion on each discrete response feature to obtain an embedding vector corresponding to each discrete response feature; forward propagate the feature vector corresponding to each numerical response feature and each embedding vector to obtain a second response probability of each candidate user responding to the target demand event; Based on the first response probability and the second response probability of each candidate user in responding to the target demand event, the response probability of each candidate user in responding to the target demand event is obtained.

3. The method according to claim 1 or 2, It is characterized in that The obtaining of multiple target historical demand events matching the target demand event includes: If the target demand event exists in the multiple historical demand events, then based on the event similarity matrix, multiple target historical demand events matching the target demand event are obtained; wherein the event similarity matrix includes the similarity between any two historical demand events, and the similarity is determined by the number of responses of users to the two historical demand events corresponding to the similarity; If the target demand event does not exist in the multiple historical demand events, determining the similarity between the target demand event and each historical demand event based on the event feature of the target demand event and the event feature of each historical demand event; Based on the similarity between the target demand event and each historical demand event, determining a similar historical demand event with the highest similarity to the target demand event; Based on the event similarity matrix, a plurality of target similar historical demand events matching the similar historical demand events are obtained; and the plurality of target similar historical demand events are used as the plurality of target historical demand events.

4. The method according to claim 3, It is characterized in that The historical response information includes the number of responses; The determining of a candidate user set corresponding to each target historical demand event based on the historical response information of the user to the target historical demand event includes: Taking users whose number of responses to each target historical demand event is greater than a first threshold as candidate users for each target historical demand event; The candidate users of each target historical demand event are merged and deduplicated to obtain a candidate user set corresponding to the target demand event.

5. The method according to claim 2, It is characterized in that Before inputting the basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model, the method further includes: Performing data preprocessing on basic user information, electricity consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events to obtain multiple numerical response characteristics and multiple discrete response characteristics corresponding to each candidate user; Perform feature crossover on the multiple numerical response features and the multiple discrete response features to obtain multiple cross response features.

6. The method according to any one of claims 1 to 5, It is characterized in that The method further comprises: Acquire multiple training samples, each sample pair including basic user information, power consumption behavior information, event characteristics of a sample demand event, historical response information of the sample user to multiple sample historical demand events, and a sample label, wherein the sample label is used to indicate a real response of the sample user to the sample demand event; Model training is performed based on the multiple training samples to obtain the target user response probability prediction model.

7. The method according to claim 5, It is characterized in that The performing model training based on the multiple training samples to obtain the target user response probability prediction model includes: Obtaining basic user information, electricity consumption behavior information, event characteristics of the sample demand event, feature vectors corresponding to each sample numerical response feature in historical response information of the sample user to multiple sample historical demand events, feature vectors corresponding to each sample discrete response feature, and feature vectors corresponding to each sample cross-response feature of each sample user; Concatenate the feature vector corresponding to each sample discrete response feature and the feature vector corresponding to each sample cross-response feature to obtain a sample response feature matrix corresponding to each sample user; determine the third response probability of each sample user to the sample demand event based on the response feature matrix corresponding to each sample user; Perform feature conversion on each sample discrete response feature to obtain a sample embedding vector corresponding to each sample discrete response feature; forward propagate the feature vector corresponding to each sample numerical response feature and each sample embedding vector to obtain a fourth response probability of each sample user responding to the sample demand event; Based on the third response probability and the fourth response probability of each sample user responding to the sample demand event, obtaining the response probability of each sample user to the sample demand event; Determine the model loss corresponding to each training sample based on the response probability of each sample user to the sample demand event and the sample label of each training sample; Based on the model loss corresponding to each training sample, the model parameters are updated to perform model training to obtain the target user response probability prediction model.

8. A demand response user determination device, It is characterized in that The demand response user determination device includes an acquisition unit and a processing unit; The acquisition unit is used to acquire multiple target historical demand events matching the target demand event; Based on the historical response information of the user to each target historical demand event, determining a candidate user set corresponding to the target demand event; Obtaining basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events for each candidate user in the candidate user set; The processing unit is used to input the basic user information, power consumption behavior information, event characteristics of the target demand event, and historical response information of each candidate user to multiple historical demand events into the target user response probability prediction model to obtain the response probability of each candidate user to the target demand event; Based on the response probability of each candidate user to the target demand event, a plurality of target users corresponding to the target demand event are determined from the candidate user set.

9. An electronic device, It is characterized in that include: A processor and a memory, wherein the processor is connected to the memory, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device executes the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method according to any one of claims 1 to 7.

Images