Method and system for online recommendation of demand response users

Abstract

The application provides a power demand response user online recommendation method and system, and relates to the technical field of power industry demand response.The application realizes online learning of user demand response behavior patterns in the case of lacking user demand response behavior data, accurately infers user demand response behavior, can effectively improve the accuracy of power demand response user online recommendation, and quickly determines reliable users participating in demand response for a power seller.

Description

Technical Field

[0001] This invention relates to the field of demand response technology in the power industry, specifically to an online recommendation method and system for power demand response users. Background Technology

[0002] Demand response is an effective way for electricity retailers to cope with soaring spot electricity prices, eliminate discrepancies between forward contracts and short-term load forecasts, and mitigate the risk of losses.

[0003] Currently, most modeling of user demand response behavior relies on user optimization problems and demand elasticity to establish the relationship between price or incentives and demand response. In this process, all parameters except price or incentives are set to fixed values, meaning that existing modeling methods assume demand response behavior depends solely on price or incentives.

[0004] However, in reality, when other contextual factors such as user personality type, education level, temperature, and income change, users' demand response behavior may also change. Existing modeling methods ignore this, resulting in poor online recommendation performance for electricity demand response users. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method and system for online recommendation of electricity demand response users, solving the technical problem that existing modeling methods lead to poor online recommendation results for electricity demand response users.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a method for online recommendation of electricity demand response users, comprising:

[0008] Obtain forward contract electricity and short-term load forecast values. When the short-term load forecast value is detected to be greater than the forward contract electricity, send a demand response invitation to the user.

[0009] Obtain user feedback on the demand response invitation, determine the set of potential users based on the user feedback, the set of potential users includes the ID of each potential user, the set of timestamps of participation in demand response projects, historical response status, and context data; set auxiliary matrices and auxiliary vectors based on the context data;

[0010] Based on the context data of each potential user in the potential user set, the set of timestamps of participation in the demand response project, the auxiliary matrix, the auxiliary vector, and the historical response status, the recommendation value of the potential users in the potential user set is calculated using the user recommendation value calculation rules. Based on the recommendation value of the potential users and the target value of the demand response project, the recommended user set is determined, and a demand response contract is signed with the recommended users in the recommended user set.

[0011] Monitor whether the recommended users who have signed the demand response contract are fulfilling the contract, and update the auxiliary matrix, auxiliary vector, and historical response status of each recommended user in the recommended user set.

[0012] Preferably, the method for obtaining the short-term load forecast value includes:

[0013] Collect historical load data, historical weather data, and historical time characteristic data, and use time series forecasting methods or machine learning methods to perform short-term load forecasting to obtain short-term load forecast values.

[0014] Preferably, the user recommendation value calculation rules include:

[0015] Calculate the variance σ of the expected response volume estimate for the current intended user ID (user_id). 2 :

[0016]

[0017] if Calculate the estimated expected response volume for the current intended user ID. First recommendation value

[0018] if Calculate the expected response volume estimate for the current user ID. Second Recommendation Value

[0019] Where A is an auxiliary matrix, x t For contextual data, x t =(x t,f x t,e ), where x t,f The characteristic parameter x represents the intended user. t,e b represents environmental parameters; b is an auxiliary vector. This is a set of timestamps for participating in a requirements response project; t represents the timestamp. This represents a bound for estimating the variance, where α represents a weighting parameter; t total This indicates the total number of demand response projects organized to date.

[0020] Updated auxiliary matrix Updated auxiliary vector b new =b+c t x t Updated historical response status C new =[C,c t ]; where C represents historical response data, c t This indicates the actual response of the recommended user, when the recommended user responds according to the contract c. t =1, when the recommended user does not respond to c according to the contract. t =0.

[0021] Preferably, determining the recommended user set based on the recommendation value of potential users and the target value of the demand response project specifically includes:

[0022] The first referral value V of potential users a Add the data to the first recommendation value set V1, perform Min-Max normalization on all data in V1, and obtain the first updated user recommendation value:

[0023]

[0024] The second referral value V of potential users b Input the data into the second recommendation value set V2, perform Min-Max normalization on all data in V2, and obtain the second updated user recommendation value:

[0025]

[0026] according to Arrange all the potential user IDs in the non-ascending order of the potential user list to form a new potential user sequence. Calculate the expected cumulative value sequentially according to the order of the new potential user sequence. Determine the number K of recommended users at timestamp t. t This makes the first K t -1 potential user's estimated response time The sum is less than or equal to D t -1 / 2, and the first K t Estimated response volume per user The sum is greater than D t -1 / 2, extract the top K from the sequence of new interested users. t Each user ID is added to the recommended user set;

[0027] Among them, D t D represents the target value for the demand response project. t = Short-term load forecast - forward contract electricity volume.

[0028] Secondly, the present invention provides an online recommendation system for electricity demand response users, comprising: a user terminal module and a recommendation system module;

[0029] The user-end module is used by user nodes to receive and respond to information sent by the recommendation system module during the execution of the demand response project.

[0030] The recommendation system module includes: a deviation detection unit, an information interaction unit, a data acquisition unit, a user recommendation unit, and an online learning unit;

[0031] The deviation detection unit is used to obtain the forward contract electricity volume and the short-term load forecast value. When the short-term load forecast value is detected to be greater than the forward contract electricity volume, the unit sends a demand response invitation to the user through the information interaction unit.

[0032] The information interaction unit is used to send demand response invitations to users, obtain feedback intention information from users regarding demand response invitations, and determine the set of potential users based on user feedback intention information; it is also used to send demand response contracts to the user-end module and receive demand response contracts signed by users.

[0033] The data acquisition unit is used to collect the timestamp set of each participating demand response item in the set of intended users. Historical response data (C), context data (x) t =(x t,f x t,e ), where x t,f The characteristic parameter x represents the intended user. t,e Indicates environmental parameters;

[0034] The user recommendation unit is used to recommend users based on their contextual data and timestamps of participation in demand response projects, from the set of intended users. Auxiliary matrix A, auxiliary vector b, and historical response status C are used to calculate the recommendation value of potential users in the potential user set using the user recommendation value calculation rule. The recommended user set is determined based on the recommendation value of potential users and the target value of the demand response project, where the target value of the demand response project = short-term load forecast value - long-term contract electricity.

[0035] The online learning unit is used to monitor whether the recommended users who signed the demand response contract are fulfilling their obligations under the contract, and to update the auxiliary matrix of each recommended user in the recommended user set. Auxiliary vector b new =b+c t x t Historical response situation C new =[C,c t ]; where c t This indicates the actual response of the recommended user, when the recommended user responds according to the contract c. t =1, when the recommended user does not respond to c according to the contract. t =0.

[0036] Preferably, the recommendation system module further includes a data processing unit, which is used to clean and transform the data collected in the data acquisition unit.

[0037] Preferably, the recommendation system module further includes a data storage unit, which is used to store the current total number of organizational demand response projects, t. total Store the set of timestamps of participants in the demand response project according to the intended user ID (user_id). Context preprocessed by the data processing unit Auxiliary matrix updated by online learning unit Auxiliary vector Historical response situation Where d = d1 + d2, d represents the data dimension in the context data, d1 represents the data dimension of user parameters, and d2 represents the data dimension of environment parameters. When the user ID (user_id) has never participated in a requirement response project, then the timestamp set... Auxiliary matrix A = I d The auxiliary vector b = 0; it is also used to store the set of recommended users sent from the user recommendation unit and the demand response contract signed from the information interaction unit.

[0038] Preferably, the user recommendation value calculation rules include:

[0039] Calculate the variance σ of the expected response volume estimate for the current intended user ID (user_id). 2 :

[0040]

[0041] if Calculate the estimated expected response volume for the current intended user ID. First recommendation value

[0042] if Calculate the expected response volume estimate for the current user ID. Second Recommendation Value

[0043] Where t represents the timestamp. This represents a bound for estimating the variance, where α represents a weighting parameter; t total This indicates the total number of demand response projects organized so far.

[0044] Thirdly, the present invention provides a computer-readable storage medium storing a computer program for online user recommendation in electricity demand response, wherein the computer program causes a computer to execute the online user recommendation method for electricity demand response as described above.

[0045] Fourthly, an electronic device comprising:

[0046] One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including methods for performing online user recommendation methods for power demand response as described above.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] This invention provides a method and system for online user recommendation in electricity demand response. Compared with existing technologies, it has the following advantages:

[0049] This invention enables online learning of user demand response behavior patterns and accurate inference of user demand response behavior in the absence of user demand response behavior data. This effectively improves the accuracy of online user recommendations for electricity demand response and quickly identifies reliable users for electricity retailers to participate in demand response. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the steps of an online user recommendation method for electricity demand response in Example 1;

[0052] Figure 2 This is a system flowchart of an online user recommendation system for electricity demand response in Example 2. Detailed Implementation

[0053] To make the above-mentioned objectives, features, and advantages of this application more apparent and understandable, the specific embodiments of this application are described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0054] This application provides a method and system for online recommendation of electricity demand response users, which solves the technical problem that existing modeling methods lead to poor online recommendation results for electricity demand response users. In the absence of user demand response behavior data, it learns user demand response behavior patterns online, accurately infers user demand response behavior, and quickly identifies reliable users participating in demand response for electricity retailers, thereby improving recommendation results.

[0055] The technical solution in this application is to solve the above-mentioned technical problems, and the general idea is as follows:

[0056] Demand response is an effective way for electricity retailers to mitigate loss risks. However, influenced by time-varying user and environmental parameters (referred to as context), uncertain user demand response behavior affects its effectiveness. Therefore, selecting reliable users is crucial for electricity retailers to ensure the success of demand response. In other words, improving the accuracy of online user recommendations for electricity demand response is key to guaranteeing its effectiveness. Existing online user recommendation practices for electricity demand response suffer from three main problems:

[0057] First, it neglects the influence of context on demand response behavior. Modeling user demand response behavior often involves modeling the relationship between price or incentives and demand response through user optimization problems and demand elasticity. In this modeling approach, all parameters except price or incentives are fixed, meaning demand response behavior is only related to price or incentives. This modeling method only considers the impact of price or incentives on demand response behavior, ignoring the reality that user demand response behavior can also change when other contextual factors (such as user personality type, education level, temperature, and income) change.

[0058] Secondly, the uncertainty of user behavior is either not considered or oversimplified. In real life, users' demand response behavior may deviate from the planned demand response amount. For example, if a contract stipulates that a user needs to reduce the load by 10kW at a certain time, the user may default at that time, reducing the load by only 5kW or even not reducing the load at all. Current modeling of user demand response behavior mostly does not consider this uncertainty.

[0059] Third, it fails to consider the lack of user demand response behavior data for e-commerce businesses. Data-driven prediction methods, such as piecewise linear regression models, hybrid density recurrent neural networks (MD-RNN), and long short-term memory (LSTM), excel in predicting demand response potential based on context. However, the accuracy of these methods relies on large amounts of historical data with long update intervals. The lack of user demand response behavior data is a common problem for e-commerce businesses, such as newly established businesses lacking such data or those lacking data on new customers' demand response behavior. Therefore, the aforementioned data-driven prediction methods are ineffective in inferring user behavior for e-commerce businesses lacking such data.

[0060] To address the aforementioned issues, this invention provides an online recommendation method and system for electricity demand response users. The method involves detecting the deviation between long-term contracts (e.g., annual or monthly contracts) and short-term load forecasts; sending demand response invitations to all electricity retailers' service users; determining a set of potential users based on user feedback; calculating recommendation values ​​for potential users in the set using a user demand response behavior prediction model based on user characteristic parameters and environmental parameters; determining a recommended user set based on these recommendation values; signing demand response contracts; and updating the parameters of the user demand response behavior prediction model based on whether users are fulfilling their contractual obligations. This approach enables online learning of user demand response behavior patterns and accurate inference of user demand response behavior even in the absence of user demand response behavior data. It effectively improves the accuracy of online recommendation for electricity demand response users and quickly identifies reliable users for electricity retailers to participate in demand response.

[0061] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0062] Example 1:

[0063] This invention provides an online recommendation method for electricity demand response users, such as... Figure 1 As shown, it includes:

[0064] S1. Obtain the forward contract electricity volume and short-term load forecast value. When the short-term load forecast value is detected to be greater than the forward contract electricity volume, send a demand response invitation to the user.

[0065] S2. Obtain user feedback on the demand response invitation, and determine the set of potential users based on the user feedback. The set of potential users includes the ID of each potential user and a set of timestamps of participation in the demand response project. Historical response data (C), context data (x) t =(x t,f xt,e ), where x t,f User's characteristic parameters, x t,e Indicates environmental parameters;

[0066] S3. Based on the context data of each user in the set of intended users and the set of timestamps of participation in the demand response project. Auxiliary matrix A, auxiliary vector b, and historical response data C are used to calculate the recommendation value of potential users in the potential user set using user recommendation value calculation rules. Based on the recommendation value of potential users and the target value of demand response projects, a recommended user set is determined, and demand response contracts are signed with the recommended users in the recommended user set. The target value of the demand response project = short-term load forecast value - long-term contract electricity volume.

[0067] S4. Monitor whether the recommended users who signed the demand response contract are fulfilling their obligations under the contract, and update the auxiliary matrix of each recommended user in the recommended user set. Auxiliary vector b new =b+c t x t Historical response situation C new =[C,c t ]; where c t This indicates the actual response of the recommended user, when the recommended user responds according to the contract c. t =1, when the recommended user does not respond to c according to the contract. t =0.

[0068] This invention enables online learning of user demand response behavior patterns and accurate inference of user demand response behavior in the absence of user demand response behavior data. This can effectively improve the accuracy of online user recommendations for electricity demand response and quickly identify reliable users for electricity retailers to participate in demand response.

[0069] The following is a detailed description of each step:

[0070] In step S1, the forward contract electricity volume and short-term load forecast value are obtained. When the short-term load forecast value is detected to be greater than the forward contract electricity volume, a demand response invitation is sent to the user.

[0071] Collect historical load data, weather data (temperature, humidity, precipitation, etc.), and time characteristic data (hours, days of the week, months, seasons, weekdays, weekends, holidays, etc.). Use time series forecasting methods or machine learning methods to perform short-term load forecasting and obtain short-term load forecast values. When it is detected that the long-term contract electricity is less than the short-term load forecast value, send a demand response invitation to the user.

[0072] In step S2, user feedback on the demand response invitation is obtained. A set of potential users is determined based on this feedback, including the user IDs and a set of timestamps from each user's participation in the demand response project. Historical response data (C), context data (x) t =(x t,f x t,e ), where x t,f User's characteristic parameters, x t,e This represents environmental parameters. The specific implementation process is as follows:

[0073] When a prospective user receives an invitation to participate in a requirements response project, they can choose to participate or not. Once the system receives a user's expression of interest, it creates a set of prospective users. This set includes various information, such as each prospective user's ID (user_id) and a set of timestamps indicating participation in the requirements response project. Historical response data (C), context data (x) t =(x t,f x t,e )wait.

[0074] In step S3, based on the context data of each potential user in the potential user set and the set of timestamps of participation in the demand response project... Auxiliary matrix A, auxiliary vector b, and historical response data C are used to calculate the recommendation value of potential users in the potential user set using user recommendation value calculation rules. Based on the recommendation value of potential users and the target value of the demand response project, a recommended user set is determined, and demand response contracts are signed with users in the recommended user set. The target value of the demand response project = short-term load forecast value - long-term contract electricity volume. The specific implementation process is as follows:

[0075] The rules for calculating user recommendation value are as follows:

[0076] Calculate the variance σ of the expected response volume estimate for the current intended user ID (user_id). 2 :

[0077]

[0078] if Calculate the estimated expected response volume for the current intended user ID. First recommendation value

[0079] if Calculate the expected response volume estimate for the current user ID. Second Recommendation Value

[0080] Where t represents the timestamp. This represents a bound for estimating the variance, where α represents a weighting parameter; t total This indicates the total number of demand response projects organized so far.

[0081] The recommended user set is determined based on the referral value of potential users and the target value of the demand response project, specifically including:

[0082] The first referral value V of potential users a Add the data to the first recommendation value set V1, perform Min-Max normalization on all data in V1, and obtain the first updated user recommendation value:

[0083]

[0084] The second referral value V of potential users b Input the data into the second recommendation value set V2, perform Min-Max normalization on all data in V2, and obtain the second updated user recommendation value:

[0085]

[0086] according to All user IDs (user_id) in the list of prospective users are sorted in non-ascending order to form a new sequence of prospective users. The expected cumulative value is then calculated sequentially according to the order of the user sequence. Determine the number K of recommended users at timestamp t. t This makes the first K t -1 user estimated response time The sum is less than or equal to D t -1 / 2, and the first K t Estimated response volume per user The sum is greater than D t -1 / 2. Extract the top K from the sequence of new interested users. t Each user ID (user_id) is added to the recommended user set.

[0087] Sign demand response contracts with users in the recommended user set.

[0088] In step S4, monitor whether the recommended users who signed the demand response contract are performing according to the contract, and update the auxiliary matrix of each user in the recommended user set. Auxiliary vector b new =b+c t x t Historical response situation C new =[C,c t ]; where c t This indicates the actual response of the recommended user, when the recommended user responds according to the contract c.t =1, when the recommended user does not respond to c according to the contract. t =0. The specific implementation process is as follows:

[0089] The monitoring of the recommended user set (recommend_user) includes all recommended user IDs (user_id) and their actual response status (c). t When the recommended user responds to c according to the contract t =1, when the recommended user does not respond to c according to the contract. t =0. Simultaneously, obtain the current total number of organizational demand response projects, t. total , timestamp set T, context x t Auxiliary matrix A, auxiliary vector b, historical response data C, and update t. total =t total +1, update the auxiliary matrix of user IDs (user_id) contained in the recommended user set. Auxiliary vector b new =b+c t x t Historical response situation C new =[C,c t ].

[0090] Example 2:

[0091] This invention provides an online recommendation system for electricity demand response users, such as... Figure 2 As shown, the system includes: a user-side module and a recommendation system module;

[0092] The user-end module is used by user nodes to receive and respond to information sent by the recommendation system module during the execution of the demand response project.

[0093] The recommendation system module includes: a deviation detection unit, an information interaction unit, a data acquisition unit, a user recommendation unit, and an online learning unit;

[0094] The deviation detection unit is used to obtain the forward contract electricity volume and the short-term load forecast value. When the short-term load forecast value is detected to be greater than the forward contract electricity volume, the unit sends a demand response invitation to the user through the information interaction unit.

[0095] The information interaction unit is used to send demand response invitations to users, obtain feedback intention information from users regarding demand response invitations, and determine the set of potential users based on user feedback intention information; it is also used to send demand response contracts to the user-end module and receive demand response contracts signed by users.

[0096] The data acquisition unit is used to collect the timestamp set of each participating demand response item in the set of intended users. Historical response data (C), context data (x) t =(x t,f x t,e ), where x t,f The characteristic parameter x represents the intended user. t,e Indicates environmental parameters;

[0097] The user recommendation unit is used to recommend users based on their contextual data and timestamps of participation in demand response projects, from the set of intended users. Auxiliary matrix A, auxiliary vector b, and historical response status C are used to calculate the recommendation value of potential users in the potential user set using the user recommendation value calculation rule. The recommended user set is determined based on the recommendation value of potential users and the target value of the demand response project, where the target value of the demand response project = short-term load forecast value - long-term contract electricity.

[0098] The online learning unit is used to monitor whether the recommended users who signed the demand response contract are fulfilling their obligations under the contract, and to update the auxiliary matrix of each recommended user in the recommended user set. Auxiliary vector b new =b+c t x t Historical response situation C new =[C,c t ]; where c t This indicates the user's actual response, when the user is recommended to respond according to the contract. t =1, when the recommended user does not respond to c according to the contract. t =0.

[0099] In the specific implementation process, the recommendation system module also includes a data processing unit and a data storage unit. The data processing unit is used to clean and transform the data collected by the data acquisition unit; the data storage unit is used to store various types of data in the online recommendation system for electricity demand response users.

[0100] The following is combined Figure 2 Detailed explanation of the user-side module and the recommendation system module:

[0101] The user-side module refers to the processor that receives and responds to information sent by the recommendation system module during the execution of the demand response project. Specifically, it assigns user IDs to new users, accepts invitations from the information interaction module within the recommendation system module, and proactively sends participation intention information to the information interaction module within the invitation deadline. This intention information includes the user ID and participation details, i.e., "participate" or "not participate." If the user does not proactively send intention information within the invitation deadline, it automatically sends a "not participate" message to the information interaction module. It also receives demand response contracts from the information interaction module, or information indicating that participation in the current demand response project has failed. Finally, it performs an electronic signature operation on the demand response contract and sends it to the information interaction module.

[0102] The recommender system module refers to the processor that monitors demand response objectives, collects and processes data, and completes user recommendations for demand response. It includes a deviation detection unit, an information interaction unit, a data acquisition unit, a data processing unit, a data storage unit, a user recommendation unit, and an online learning unit.

[0103] The deviation detection unit is used for short-term load forecasting based on time series forecasting or machine learning methods. It calculates the deviation between long-term contracts (such as annual or monthly contracts) and short-term load forecasts to determine the target value D of the demand response project. t (Demand response project target value = short-term load forecast value - long-term contract electricity volume) and the corresponding timestamp t, and send them to the information exchange module.

[0104] Information Interaction Unit: After receiving the target value of the response project sent by the deviation detection module, it broadcasts the demand response invitation information to all service users of the e-commerce platform. The invitation information includes the target value D of the demand response project. t Response timestamp t, invitation deadline After sending the invitation message, accept the intention message sent by the user-side module, and within the invitation deadline... After completion, the list of intended users for this demand response project, `user_memu`, is compiled and formed, containing the user IDs (`user_id`) of all intended users, and sent to the user recommendation unit. The unit receives the recommended user set `recommend_user` from the user recommendation unit, sends demand response contracts to the user-side modules corresponding to all user IDs (`user_id`) in `recommend_user`, and sends a failure message to the user-side modules corresponding to the remaining user IDs (`user_id`) in the `user_memu`. The unit receives signed demand response contracts from user-side modules and sends them to the data storage unit.

[0105] Data acquisition unit: Collects user parameters after sending invitation information. and environmental parameters The user ID (user_id) is sent to the data processing module along with the user's information. User parameters include residential characteristic parameters and user characteristic parameters. Residential characteristic parameters include the number of rooms and floor level, while user characteristic parameters include the number of residents, total household income, and education level. Environmental parameters include time parameters and weather parameters. Time parameters include month, day, hour, and week, while weather parameters include temperature, humidity, wind speed, and air pressure.

[0106] Data processing unit: processes the context x acquired by the data acquisition module. t =(x t,f x t,e The data is cleaned and transformed before being sent to the data storage module. Data cleaning refers to checking for errors, duplicates, or incomplete parts in the data and correcting or deleting them. Data transformation refers to normalizing, standardizing, discretizing, etc., the raw data into a form that is easier to process and analyze.

[0107] Data storage unit: Stores the current total number of organizational demand response projects (t). total Store the set of timestamps of participation in the demand response project according to the user ID (user_id). Context preprocessed by the data processing unit Auxiliary matrix updated by online learning unit Auxiliary vector Historical response situation Where d = d1 + d2, d represents the data dimension in the context data, d1 represents the data dimension of user parameters, and d2 represents the data dimension of environment parameters. When the user ID (user_id) has never participated in a requirement response project, then the timestamp set... Auxiliary matrix A = I d The auxiliary vector b = 0. It stores the set of recommended users (recommend_user) sent by the user recommendation unit and the signed demand response contracts from the information interaction unit.

[0108] User recommendation unit: Receives a list of potential users (user_memu), and extracts the set of timestamps corresponding to the user IDs (user_id) contained in the user list (user_memu) from the data storage module. Context Auxiliary matrix A, auxiliary vector b, historical response data C. Calculate the variance of the expected response estimate for the current system to user ID (user_id). Calculate the referral value of the user corresponding to the user ID user_id according to the following rules:

[0109] if Calculate the estimated expected response time for the current user ID (user_id). First recommendation value The data is placed into a recommendation value set V1, and all data in V1 is normalized using Min-Max to obtain the recommendation value for the first updated user. if Calculate the estimated expected response time for the current user ID (user_id). Second Recommendation Value The data is placed into the recommendation value set V2, and all data in V2 is normalized using Min-Max to obtain the second updated user recommendation value: In an embodiment of the present invention, α = 2.5. The standardized user recommendation value is denoted as... according to All user IDs (user_id) in the list of intended users (user_memu) are sorted in non-ascending order to form a new sequence of intended users (user_sequance). The expected cumulative value is then calculated sequentially according to the user sequence (user_sequance). Determine the number K of recommended users at timestamp t. t This makes the first K t -1 user estimated response time The sum is less than or equal to D t -1 / 2, and the first K t Estimated response volume per user The sum is greater than D t -1 / 2. Extract the top K from the new intention user sequence user_sequance. t Each user ID (user_id) is added to the recommended user set (recommend_user), and the recommended user set (recommend_user) is sent to the data storage unit and the online learning unit. This represents a bound on the estimated variance. α is a weighting parameter that represents the relative weight of the second term in the expression for calculating V (where the first and second terms in the expression for calculating V represent utilization value and exploration value, respectively. Utilization value is the current estimate of the response to the user's expected needs, and exploration value is the quantification of the user's future potential). The smaller the parameter α, the smaller the proportion of exploration value, and users with high utilization value are more likely to be recommended; conversely, the higher the proportion of exploration value, the more likely users with high exploration value are to be recommended.

[0110] It should be noted that, in the embodiments of the present invention, to make the first K... t -1 user estimated response time The sum is less than or equal to D t -1 / 2, and the first Kt Estimated response volume per user The sum is greater than D t The reason for -1 / 2 is as follows:

[0111] At each time point timestamp t, the system needs to determine a set of recommended users called recommend_user (denoted as S). t This can achieve the total expected loss function. Minimum, where the first term is the squared deviation from the expected target, which quantifies the loss caused by the deviation from the demand response target, and the second term is the sum of variances, which quantifies the uncertainty in achieving the demand response target.

[0112] So, how do we ensure that we can find this optimal user combination? We calculate the expected cumulative value of the demand response of the new intention user sequence user_sequance in sequence. Then determine the number of recommended users K. t This makes the first K t -1 user estimated response time The sum is less than or equal to D t -1 / 2, and the first K t Estimated response volume per user The sum is greater than D S,t -1 / 2. Finally, the determined set of recommended users, recommend_user, is the top K of the new intended user sequence, user_sequance. t One user. That is to say, This is a necessary and sufficient condition for minimizing the objective function.

[0113] Online learning module: Receives the recommended user set `recommend_user` from the user recommendation module, and monitors the actual response status of all user IDs (`user_id`) in the `recommend_user` set. t When the user responds to c according to the contract t =1, when the user does not respond to c according to the contract. t =0. Retrieve the current total number of organizational demand response projects t from the data storage unit. total Timestamp set Context x t Auxiliary matrix A, auxiliary vector b, historical response data C. Update t total =t total +1, add the timestamp t to the recommended user set recommend_user, which contains all user IDs and response timestamps. Send to the data storage unit. Update the auxiliary matrix of user IDs (user_id) contained in the recommended user set (recommend_user). Auxiliary vector b new =b+c t x t Historical response situation C new =[C,c t The data is sent to the data storage unit.

[0114] This invention also provides a computer-readable storage medium storing a computer program for online user recommendation in electricity demand response, wherein the computer program causes a computer to execute the online user recommendation method for electricity demand response as described above.

[0115] This invention also provides an electronic device, including: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs including methods for performing the online user recommendation method for power demand response as described above.

[0116] In summary, compared with existing technologies, it has the following beneficial effects:

[0117] 1. The embodiments of the present invention enable online learning of user demand response behavior patterns in the absence of user demand response behavior data, accurately inferring user demand response behavior, effectively improving the accuracy of online user recommendations for electricity demand response, and quickly identifying reliable users participating in demand response for electricity retailers.

[0118] 2. The embodiments of the present invention can optimize the parameters of the user's expected response quantity estimation model based on the actual response situation of the user's participation in the demand response project, improve the estimation accuracy, and thus further improve the online recommendation effect of the power demand response user.

[0119] The above description is merely a preferred embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

[0120] It should be understood that any parts not described in detail in this specification belong to the prior art.

[0121] It should be understood that the above description of the preferred embodiments is quite detailed, but it should not be considered as a limitation on the scope of protection of this invention. Those skilled in the art, under the guidance of this invention, can make substitutions or modifications without departing from the scope of protection of the claims of this invention, and all such substitutions or modifications fall within the scope of protection of this invention. The scope of protection of this invention should be determined by the appended claims.

Claims

1. A method for online recommendation of demand response users, characterized in that, include: Obtain forward contract electricity and short-term load forecast values. When the short-term load forecast value is detected to be greater than the forward contract electricity, send a demand response invitation to the user. Obtain user feedback on the demand response invitation, determine the set of potential users based on the user feedback, the set of potential users includes the ID of each potential user, the set of timestamps of participation in demand response projects, historical response status, and context data; set auxiliary matrices and auxiliary vectors based on the context data; Based on the context data of each potential user in the potential user set, the set of timestamps of participation in the demand response project, the auxiliary matrix, the auxiliary vector, and the historical response status, the recommendation value of the potential users in the potential user set is calculated using the user recommendation value calculation rules. Based on the recommendation value of the potential users and the target value of the demand response project, the recommended user set is determined, and a demand response contract is signed with the recommended users in the recommended user set. The rules for calculating user recommendation value include: Compute current intended user number user_id Estimate variance of desired response amount : If , calculate the expected response amount estimate value of the current intended user number , the first recommended value ; If , calculate the expected response amount estimate value of the current intended user number , the second recommended value ; in, For auxiliary matrix, For contextual data, =( , ),in, Characteristic parameters representing potential users, Indicates environmental parameters; For auxiliary vectors; A set of timestamps for participating in the requirements response project; t Represents a timestamp. This represents a bound for estimating the variance. This represents a weight parameter; t total This indicates the total number of demand response projects organized to date. Updated auxiliary matrix Updated auxiliary vectors Updated historical response information ;in For historical response situations, This indicates the actual response of the recommended user, when the recommended user responds according to the contract. c t =1, when the recommended user does not respond according to the contract. c t =0; The process of determining the recommended user set based on the recommendation value of potential users and the target value of the demand response project specifically includes: Putting the first recommended value of the intended user Put into the first recommended value set , to Min-Max normalization is performed on all data to obtain the first updated user recommended value: ; Second recommended value of the intended user Put the second recommended value set , to Min-Max normalization of all data, get the second update user recommended value: ; according to Arrange all the potential user IDs in the non-ascending order of the potential user list to form a new potential user sequence. Calculate the expected cumulative value sequentially according to the order of the new potential user sequence. Determine the number of recommended users at timestamp t. K t This makes the former K t Estimated response volume per potential user The sum is less than or equal to D t -1 / 2, and the front K t Estimated response volume per user The sum is greater than D t -1 / 2, extract the top 1% of the new intent user sequence K t Each user ID is added to the recommended user set; wherein, D t = demand response project target value, D t = short-term load forecast value - forward contract electricity amount; Monitor whether the recommended users who have signed the demand response contract are fulfilling the contract, and update the auxiliary matrix, auxiliary vector, and historical response status of each recommended user in the recommended user set.

2. The online user recommendation method for electricity demand response as described in claim 1, characterized in that, The methods for obtaining the short-term load forecast values ​​include: Collect historical load data, historical weather data, and historical time characteristic data, and use time series forecasting methods or machine learning methods to perform short-term load forecasting to obtain short-term load forecast values.

3. An online recommendation system for electricity demand response users, used to execute the steps of the online recommendation method for electricity demand response users as described in any one of claims 1-2, characterized in that, Includes: a user-side module and a recommendation system module; The user-end module is used by user nodes to receive and respond to information sent by the recommendation system module during the execution of the demand response project. The recommendation system module includes: a deviation detection unit, an information interaction unit, a data acquisition unit, a user recommendation unit, and an online learning unit; The deviation detection unit is used to obtain the forward contract electricity volume and the short-term load forecast value. When the short-term load forecast value is detected to be greater than the forward contract electricity volume, the unit sends a demand response invitation to the user through the information interaction unit. The information interaction unit is used to send demand response invitations to users, obtain feedback intention information from users regarding demand response invitations, and determine the set of potential users based on user feedback intention information; it is also used to send demand response contracts to the user-end module and receive demand response contracts signed by users. The data acquisition unit is used to collect the timestamp set of each participating demand response item in the set of intended users. Historical response status Context data = ( , ),in, Characteristic parameters representing potential users, Indicates environmental parameters; The user recommendation unit is used to recommend users based on their contextual data and timestamps of participation in demand response projects, from the set of intended users. Auxiliary matrix Auxiliary vectors Historical response status The recommendation value of potential users in the potential user set is calculated using the user recommendation value calculation rule. The recommended user set is determined based on the recommendation value of potential users and the target value of the demand response project. The target value of the demand response project = short-term load forecast value - long-term contract electricity. The online learning unit is used to monitor whether the recommended users who signed the demand response contract are fulfilling their obligations under the contract, and to update the auxiliary matrix of each recommended user in the recommended user set. Auxiliary vectors Historical response status ;in, This indicates the actual response of the recommended user, when the recommended user responds according to the contract. c t =1, when the recommended user does not respond according to the contract. c t =0.

4. The online demand response customer recommendation system of claim 3, wherein, The recommendation system module also includes a data processing unit, which is used to clean and transform the data collected in the data acquisition unit.

5. The online demand response customer recommendation system of claim 3, wherein, The recommendation system module also includes a data storage unit, which stores the current number of times the organization has responded to demand-related projects. t total According to the intended user number user_id Store a collection of timestamps from projects that participated in the demand response. Context preprocessed by the data processing unit = ( , ) The auxiliary matrix updated by the online learning unit Auxiliary vectors Historical response status ,in, , d This refers to the data dimension within the context data. Data dimensions representing user parameters The data dimension representing environmental parameters, when the user ID... user_id If you have never participated in a requirements response project, then the timestamp set auxiliary matrix auxiliary vector It is also used to store the set of recommended users sent from the user recommendation unit and the demand response contract signed from the information interaction unit.

6. A computer-readable storage medium, characterized in that, It stores a computer program for online recommendations to users in demand response, wherein the computer program causes a computer to execute the online recommendation method for demand response as described in any one of claims 1 to 2.

7. An electronic device, comprising: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the programs comprising programs for performing the method for online recommendation of demand response users as claimed in any one of claims 1-2.

Images