KR102133741B1 - Server Having a Function of Managing Power Demand and Method of Managing Power Demand Thereof - Google Patents

Abstract

According to the present invention, the storage unit for storing the event information including the subscriber information including the power consumption of the subscriber's at least one power consumption device and the subscriber's personal demand response event participation record, and the date of the demand response event; Receiving unit for receiving a request signal for power savings; A selection unit for selecting a subscriber who wants to request power savings by using the subscriber information and event information as the request signal is received; And a transmitting unit that transmits guide information on the reduction in power use to the subscriber terminal of the selected subscriber, wherein the selecting unit generates a credit scoring model, evaluates a subscriber's credit score according to the credit scoring model, and uses the power. A server having a power demand management function is provided that selects subscribers who wish to request savings according to the evaluated credit score.

Description

A server having a power demand management function, and a server having a power demand management method thereof

The present invention relates to a power demand management method, and more specifically, in a target selection process for selecting a subscriber who is likely to participate in a demand response event, it is easy to understand the selection process for general subscribers who do not have energy-related knowledge. It relates to a server having a power demand management function that provides a method of selecting an invitation target subscriber in case of a demand response event capable of providing gender, and a power demand management method thereof.

Electricity consumption in the household sector is steadily increasing and is expected to account for 41% of total power consumption in the next 20 years. Moreover, the household sector can have a significant impact because it not only accounts for a large portion of the power consumption, but also can play an important role in improving social awareness of energy use through educational activities in the household sector. In this context, Residential Demand Response is considered an important tool to solve the imbalance between electricity demand and supply.

There are two types of housing demand response: price-based demand response and incentive-based demand response. Among them, incentive-based demand response can provide the flexibility for subscribers to choose whether or not to participate in the demand response, and there is no need for large-scale investment in a dedicated infrastructure in the initial stage. Therefore, incentive-based demand response is advantageous for introduction due to low burden on subscribers, and it can provide both efficiency and reliability in the initial application of demand response.

This incentive-based demand response is generally operated in the form of an invitation-based event that induces the participation of subscribers by sending a scheduled invitation at a time when power reduction is most required. In this case, incentive-based demand response cannot force subscribers to participate. Therefore, it can be helpful in two ways to introduce the'selective process', which sends invitations mainly to subscribers who are likely to participate, to incentive-based demand response.

First, the target selection process can exclude participants in demand reactions that receive only compensation without substantially contributing to power reduction. Because people's power consumption patterns are very diverse, some simply participate in all demand response events, do not make efforts to reduce power, and receive compensation even if they consume power according to their existing power consumption patterns. This problem occurs because the usage pattern of residential demand response participants is more diverse than that of business demand response participants. Therefore, the demand response program operator can improve the operational efficiency by providing an option to detect and exclude such demand response participants by utilizing the target selection process.

Second, the target selection process can increase the satisfaction of participants in demand response by appropriately controlling the frequency of sending invitations and notifications. In order for the incentive-based demand response to work, an invitation must be sent by default, and a reminder message to remind the point of reduction can help increase the efficiency of the demand response for housing. However, if there are many invitations and notifications, there is a drawback that can annoy some demand response participants, which in turn can lead to complaints about complaints or notifications. For this problem, the target selection process can determine and control the number of notifications to a specific demand response participant.

However, most of the existing studies focused on optimizing the amount of compensation, and there were not many studies on the target selection process. This means that even though the operation efficiency and demand response participant satisfaction can be improved through the target selection process, the opportunity for participation by some consumers can be discriminatoryly limited due to selective invitation, which may lead to fairness and legal disputes.

In particular, in the case of applying the target selection process, the supplier of the demand response program should guarantee fairness to the subscriber, make the process transparent, and explain the causal relationship for compensation of the demand response. In addition, the demand response for housing needs an easy and concise explanation because it deals with a large number of general subscribers who have no knowledge of the energy industry. Therefore,'Explainability' is a very important requirement when operating demand response for housing.

However, the process of selecting a subject is usually implemented through machine learning in order to maximize efficiency. Algorithms implemented through machine learning are accompanied by mathematical algorithms, so they are basically inexplicable. In particular, most of the demand response subscribers are general energy non-professionals, so it is very difficult to understand machine learning algorithms.

Accordingly, there is an urgent need for an electric power demand management system capable of ensuring explanatoryness while utilizing an efficient target selection process through machine learning when operating a demand response program.

In order to solve the above problems, the present invention is a server, communication device and system having a power demand management function capable of ensuring explanation while utilizing an efficient target selection process when operating a demand response program, and power demand management thereof I want to provide a method.

In order to solve the above technical problem, a server having a power demand management function according to the present invention includes subscriber information including power usage for at least one power consumption device of a subscriber and a record of participation in the personal demand response event of the subscriber, A storage unit for storing event information including the date of the demand response event; Receiving unit for receiving a request signal for power savings; A selection unit for selecting a subscriber who wants to request power savings by using the subscriber information and event information as the request signal is received; And a transmitting unit that transmits guide information on the reduction in power use to the subscriber terminal of the selected subscriber, wherein the selecting unit generates a credit scoring model, evaluates a subscriber's credit score according to the credit scoring model, and uses the power. Subscribers who want to request savings can be selected based on the evaluated credit score.

On the other hand, the power demand management method according to the present invention, an event including the subscriber information including the power usage of the subscriber's at least one power consumption device and the subscriber's personal demand response event participation record, and the date of the demand response event Storing information; Receiving a request signal for power savings; Upon receiving the request signal, selecting a subscriber who wants to request power savings by using the subscriber information and event information; And transmitting guide information on the reduction in power use to the communication device of the selected subscriber, wherein the selecting step includes generating a credit scoring model and evaluating a subscriber's credit score according to the credit scoring model. Therefore, it may include the step of selecting a subscriber who wishes to request the reduction of power use according to the evaluated credit score.

In addition, the method for selecting a subscriber for a demand response event according to the present invention includes: subscriber information including power usage for at least one power consumption device of the subscriber and a record of participation of the individual's personal demand response event, and the date of the demand response event. A storage unit for storing event information including; Receiving unit for receiving a request signal for power savings; A selection unit for selecting a subscriber who wants to request power savings by using the subscriber information and event information as the request signal is received; And In the server having a power demand management function including a transmission unit for transmitting the guide information for the power use reduction to the subscriber terminal of the selected subscriber, the information value for the plurality of features generated based on the subscriber information Selecting features based thereon; Roughly classifying the features by obtaining a logit value for the selected features; And generating a credit scoring model by generating a score range for each range of the rough classification through machine learning.

Accordingly, according to the present invention, a server, a communication device and a system having a power demand management function capable of ensuring explanation while utilizing an efficient target selection process when operating a demand response program, and a power demand management method thereof are provided. .

In addition, according to the present invention, since it is possible to encourage participation in demand response events to subscribers who are more likely to participate in demand response events, reduction in power use can be stably performed.

1 is a configuration diagram of a system having a power demand management function according to an embodiment of the present invention.
2 is a flowchart of a power demand management method according to an embodiment of the present invention.
3 is a configuration diagram of a server having a power demand management function according to an embodiment of the present invention.
4 is a configuration diagram of a subscriber terminal having a power demand management function according to an embodiment of the present invention.
5 is a flowchart of a method for selecting a subscriber for a demand response event according to an embodiment of the present invention.
6 shows features for generating a credit scoring model according to an embodiment of the present invention.
7 illustrates features selected through a selection process among the features of FIG. 6 according to an embodiment of the present invention.
8 is a view for explaining the concept of explanatory properties according to the present invention.

The following is merely illustrative of the principles of the invention. Therefore, those skilled in the art can implement various principles included in the concept and scope of the invention and implement the principles of the invention, although not explicitly described or illustrated in the specification. In addition, all conditional terms and examples listed herein are intended to be understood in principle only for the purpose of understanding the concept of the invention, and should be understood as not limited to the examples and states specifically listed in this way. do.

The above objects, features, and advantages will become more apparent through the following detailed description in connection with the accompanying drawings, and accordingly, those skilled in the art to which the invention pertains can easily implement the technical spirit of the invention. .

In addition, in the description of the invention, when it is determined that the detailed description of the known technology related to the invention may unnecessarily obscure the subject matter of the invention, the detailed description will be omitted. It will be described below with reference to the accompanying drawings.

Hereinafter, a server, a communication device and a system having a power demand management function according to an embodiment of the present invention, and a power demand management method thereof will be described with reference to FIGS. 1 to 6.

Before describing the present invention, terms that are mainly mentioned in the specification will be defined.

First, the request signal for power use reduction may include information on at least one of the amount of power required for power saving, the time zone for the power saving, and the region for which power saving is required. For example, the request signal may include information to reduce 10,000 KWh (reduction required power amount) for 2 pm to 5 pm (time required for reduction) on May 1, 2015. Furthermore, the request signal may further include information designating a specific area (for example, Suwon-si, Gangnam-gu, Seoul, etc.) as a reduction request area.

Next, the subscriber information is raw data for generating features, and records a personal event of the subscriber who subscribed to the demand response program (eg, whether to participate in each past event, the highest reduction target success, the reduction target and usage), and Electricity consumption data collected through a power measurement device (eg, a smart meter) per unit time for a certain period of time for each demand response program subscriber is included.

Next, the information on the reduction in power usage includes information on at least one of the amount of power savings required for a subscriber selected to invite to a demand response event, the power usage status of the corresponding subscriber, the time period of the reduction request, and the expected compensation according to the reduction. It may include, and may be provided for the entire power consumption device or individual power consumption devices used by the subscriber.

Lastly, the information on the compensation for the reduction in power usage may include information on discounts on electricity usage or points available when payment of electricity usage is paid in proportion to the amount of electricity actually saved to correspond to the guidance information. .

1 is a block diagram of a system having a power demand management function according to an embodiment of the present invention, and FIG. 2 is a flow chart of a power demand management method according to an embodiment of the present invention.

As shown in FIG. 1, the system 700 having a power demand management function displays an invitation to reduce power use received from the demand response operation server 300, the power measurement device 200, and the server 100. Or it may include a subscriber terminal 400 for outputting, or a power supplier server 600 transmitting a request signal for saving power use to the demand response operation server 300.

Further, according to an embodiment of the present invention, the system 700 having a power demand management function may be additionally connected to the sponsor server 610.

In this case, the demand response operation server 300, the power measurement device 200, and the subscriber terminal 400 are connected through a network 100, such as the Internet.

The demand response operation server 300 may receive power consumption of power consumption devices 500 inside the house in real time from the power measurement device 200 through the network 100. In addition, the demand response management server 300 may select a target for participation in the demand response program among subscribers based on the power consumption.

In addition, the demand response operation server 300 may receive a request signal for power usage reduction from the power supplier server 600. When receiving the request signal, the demand response operation server 300 may select a target for participation in the demand response program based on the power consumption and the profile of the subscriber. Thereafter, the demand response operation server 300 invites the subscriber terminal 400 to participate in the demand response event with respect to the selected participation target (Invitation), for example, a reminder for participation in the demand response event, etc. Notification can be sent.

In addition, the demand response management server 300 may provide compensation to subscribers participating in the demand response event. In this case, the demand response management server 300 may be connected to the sponsor server 610 and provide coupons, cash, and the like provided by the sponsor server 610 to the subscribers participating in the demand response event.

The power measuring device 200 supplies power consumption of the power consuming devices 500 inside the house to the demand response operation server 300 in real time.

On the other hand, as described above, the subscriber terminal 400 may receive invitations and notifications from the demand response operation server 300 and guide the user to participate in the demand response event. In addition, various information provided by the demand response operation server 300 may be displayed. The subscriber terminal 400 may be, for example, a mobile phone, a tablet, a personal computer (PC), and various types of devices capable of communication.

Hereinafter, the operation of each component for carrying out the present invention will be described with reference to FIG. 2.

The demand response operation server 300 receives a request signal for power usage reduction from the power supplier server 600 (S210).

Here, the power supplier server 600 is a server managed by an electric power supply source (for example, Korea Electric Power Corporation) or an electric power supply agency, and may perform overall operations related to electric power supply.

Upon receiving the request signal, the demand response operation server 300 selects a subscriber who wants to request power savings through a target selection process using subscriber information (S220).

Here, the demand response operation server 300 may store subscriber information, update the subscriber information by reflecting the subscriber's power usage status/pattern, or update the subscriber information according to the subscriber's request.

The demand response management server 300 includes a participation rate, a maximum number of consecutive participations, a cumulative reward, and a maximum number of consecutive non-participations. At least one subscriber to participate in a demand response event may be selected according to a predetermined criterion in consideration of success in the event (Success in the Last Event, failure to participate in the last event, or power reduction failure). Accordingly, the demand response operation server 300 may select a subscriber who is likely to reduce power usage by participating in a demand response event in response to a request for reducing power usage through a target selection process. Alternatively, the demand-response operation server 300 may select a subscriber who is expected to have a high response to a request to reduce power use.

With regard to the process of selecting a target for selecting a subscriber, it will be described later in detail with reference to FIG. 3.

The demand response operation server 300 transmits an invitation message including information on saving power use to the subscriber terminal device 400 selected in the selection step S220 (S230).

Here, the subscriber terminal 400 is a communication device of a subscriber registered in the demand response operation server 300 for data communication with the demand response operation server 300, and a mobile device such as a smart phone, laptop, tablet PC, and It may include all of the fixed-type home appliance (for example, TV, refrigerator, air conditioner, etc.) equipped with a communication function.

The demand response operation server 300 monitors power consumption by at least one power consuming device 500 of the subscriber selected in the selection step S220 (S240).

Therefore, the demand response operation server 300 may check whether the actual power consumption has been reduced and the amount of power saved in the power consumption device of the selected subscriber so as to correspond to the guide information transmitted in the transmission step (S230) (S250). .

When the demand response operation server 300 determines that the power usage has been reduced to correspond to the guidance information in the confirmation step S250, for example, the demand response operation server 300 may provide a compensation corresponding to the reduction of the power usage (S260).

The demand response operation server 300 may determine the compensation to be given to the corresponding subscriber in consideration of the actual amount of power saved to correspond to the guide information.

For example, the demand response management server 300 may provide compensation in proportion to the amount of power saved. In addition, the demand-response operation server 300 may reward power usage fees, and may provide rewards in a manner of discounting power usage fees or paying points that can be used to pay electricity usage fees.

Alternatively, the compensation according to an embodiment of the present invention may include cash, gifts, coupons, etc. provided by the sponsor server 610 other than the server related to the power supplier. In this case, the demand response management server 300 may receive cash, gifts, coupons, etc. in electronic form from the sponsor server 610. Alternatively, the demand response management server 300 may accumulate electronic conversion points in the account of the sponsor to correspond to cash, gifts, and coupons provided by the sponsor. The demand response management server 300 calculates the amount of power use reduction corresponding to cash, gifts, coupons, or conversion points provided for each sponsor. In addition, the sponsor server 610 may obtain a carbon dioxide emission right from the carbon dioxide emission trading server 620 based on the calculated amount of power use reduction. In this case, "based on the power savings" means all cases that are converted to carbon dioxide savings based on the power savings, and those skilled in the art can convert the power savings to carbon dioxide savings using conventional methods. In addition, the carbon dioxide emission trading server 620 confirms that power reduction has been made as much as corresponding to cash, gifts, coupons, etc. provided by the sponsor server 610 from the power supplier server 600 or the demand response operation server 300. Can.

The demand response operation server 300 may transmit the information (hereinafter, compensation information) related to the compensation given in the granting step (S260) to the corresponding subscriber station 400 (S270 ).

Accordingly, the subscriber of the subscriber station 400 can check whether the actual power usage fee is discounted and the degree of discount by his own power saving act. This can exert an effect of encouraging subscribers to actively participate in power saving when power saving is required.

3 shows a detailed configuration diagram of the demand response operation server 300 having the power demand management function shown in FIG. 1.

According to FIG. 3, the demand response management server 300 includes a storage unit 310 for storing subscriber information, a receiving unit 321 for receiving a request signal for saving power use, and the subscriber information by receiving the request signal Selection unit 331 for selecting a subscriber who wishes to request a reduction in power use by using, a transmitting unit 322 for transmitting information on the reduction in power use to the communication device of the selected subscriber, the power of the selected subscriber It may include a monitoring unit 332 for monitoring the power consumption by the consumption device and a compensation management unit 333 for granting a predetermined compensation to the corresponding subscriber when the power consumption by the power consumption device decreases according to the guide information. have.

Here, the receiving unit 321 and the transmitting unit 322 may be implemented as respective communication modules or may be implemented as one communication module 320. Also, the selection unit 331, the monitoring unit 332, and the compensation management unit 333 may be implemented as respective modules or may be implemented as one control module 330.

The storage unit 310 may store subscriber information and event information, and in particular, information about power usage for each subscriber. For example, as subscriber information, it is possible to store information about power consumption by the total power consumption device or individual power consumption device for each subscriber, power usage profile for each hour, and power saving history according to guide information. More specifically, the power saving history according to the guidance information corresponds to previous guidance information including the compensation method and the amount of money, whether or not actual power consumption has been reduced, the amount of power savings, information about the power consumption device that has performed the power reduction, etc. It may include.

The event information includes event day information (date, start time, weather, temperature and basic incentive amount) common to all demand response participants.

The reception unit 321 may receive a request signal for power usage reduction from the power supply-related server 600 prior to the start of a period/time zone where power usage reduction is required.

For example, the power supply-related server 600 may transmit a request signal one week before, one day before, or an hour before the start of a period/time zone in which power use reduction is required. This is to induce an active power use reduction action by providing information about the power use reduction to the subscriber before the period/time when the power use reduction is required.

The selection unit 331, the participation rate (Participation Rate), the maximum number of consecutive participation (Maximum Number of Consecutive Participation), cumulative compensation (Cumulative Rebate), the maximum number of consecutive non-participation (Maximum Number of Consecutive Non-Participation) and past events According to a predetermined criterion in consideration of success of the (Success in the Last Event, failure to participate in the last event or power reduction), at least one subscriber who wishes to request reduction of power use may be selected. Accordingly, the demand response operation server 300 may select a subscriber that is expected to have a high response to a request for reducing power use.

Hereinafter, with reference to FIG. 5, a process of selecting a subject, which is a detailed operation of the selection unit 331, will be described.

The selection unit 331 according to an embodiment of the present invention is designed to increase explanability by integrating a credit scoring algorithm while maximizing efficiency by utilizing machine learning.

The selection unit 331 has two modes, a credit scoring learning mode and a credit score evaluation mode. In this case, the credit scoring learning mode is a mode for generating a credit scoring model for selecting invitees to respond to demand based on existing subscriber information, and the credit score evaluation mode is based on the generated credit scoring model for individual subscribers' credit scores. This mode is to evaluate and select the invitees for demand response events.

Before explaining the two modes of the selection unit 331, the concept of explainability and credit scoring, which are important concepts for the present invention, will be described in detail.

Demand response In the program Explanatory importance

With the increase in automated decision making using big data and machine learning, many demand response participants have concerns about privacy. To address this concern, the EU's'GDPR (General Data Protection Regulation)' and'Communication on Delivering a New Deal for Energy Consumers (hereinafter referred to as New Deal)' Is being implemented.

GDPR was passed in May 2016 and is effective from 2018. It is intended to provide transparency and fairness when personal data is used for automated decision making. According to the GDPR, data subjects have the right to monitor and understand all processes. In other words, the model used for automatic decision-making should be designed in a descriptive way so that participants in demand responses can provide answers when asking questions or forwarding protests.

According to the EU's New Deal (2015), for the same reason, descriptive modeling in the energy market is needed. In the energy market, all decision-making processes related to energy services must be made transparent to demand response participants, and demand response participants have the right to freely monitor and change program options at any time. Therefore, the demand response and target selection process must also follow the same regulations, and in particular, the housing demand response program should be available to the general public because a simpler explanation is available.

Explanatory Justice

The efficiency and accuracy of demand response can use various measurement techniques such as Area Under Receiver Operation Characteristics (AURC) and Area Under Precision-Recall Curve (AUPRC).

However, explainability is not easy to define because it depends on the person's background and knowledge base. Because of the subjective nature of explainability, most of the existing research indirectly defines and evaluates explainability. For example, explanatoryity is according to existing studies (e.g., Learning Interpretable Models, S. Ruping, Dortmund Univ. (PhD Thesis) 2006, Towards a Rigorous Science of Interpretable Machine Learning, Finale Doshi-Velez, arXiv, 2017). It is defined in a qualitative way.

According to the definition of the explanatoryness of the previous study in FIG. 8, the explanatoryness can be defined using an empirical method, but not a quantitative metric. In particular, according to the previous studies above, the only way to guarantee the ability to explain is to use a model that is already known to be explainable. Therefore, in the present invention, credit scoring, which is known as an explainable approach, is adopted to enhance explanatoryness in response to demand.

Credit scoring

Credit scoring, devised by Durand in 1941, sets scores for each range through logistic regression or linear programming of features generated from historical data of demand-response subscribers. Credit scoring uses a simple and familiar form of scorecards so that subscribers to demand responses who do not have expertise in statistics can understand. Primarily credit scoring provides banks with minimal financial risks and easy to explain. For example, when you visit a bank to get a mortgage loan, the banker uses a credit scoring model to predict the person's loan repayment potential and determine the amount, interest rate, and duration to borrow. For rating purposes, overdue over the past three months is awarded -10 points, +50 points if the average salary is over $5,000, +20 points if you own a car, etc. In this case, if the demand response participant is denied the loan, the demand response participant may complain about the fairness of the bank.

To address this issue, the banking sector has the National Credit Act of 2005 states in article 62, where banks have a legal obligation to disclose decision-making processes and clearly explain the causality of each process. '). In other words, a model that can be explained should be used in situations where demand response participants can ask about fairness.

In addition to meeting the explanatory requirements, credit scoring has been actively used in the industry for over 60 years due to its high predicted hit ratio. There are two reasons for high predictive performance. First, it controls the overfitting by extracting only the features necessary for prediction. Second, nonlinearity is handled by adaptively selecting a range of different scores.

Because credit scoring performs classification based on probability basis, prediction performance is improved compared to the simple case of using raw data as it is. To improve the predictive performance of credit scoring, machine learning can be applied to feature selection or feature creation stages. However, even if machine learning is adopted, the form of a simple scorecard should be maintained so that the explanation can be maintained. In the present invention, machine learning is applied to several steps such as function selection and credit score range designation, but the final model is designed to remain a score card that can be understood by general demand response participants.

Hereinafter, the credit scoring learning mode of the target selection process of the selection unit 331 will be described in detail with reference to FIG. 5.

Credit Scoring Model for Selecting Participants in Demand Response

1. Feature Selection (Step S710)

Choosing features for credit scoring is a step in selecting some important baseline features needed to maximize efficiency while maintaining clarity. Credit scoring for demand response selects a reference feature based on the information value defined by Equation 1 below.

(Equation 1)

Here, IV is an information value, and Pr() is a probability function. Equation 1 above is a value that measures how well two groups (Y=0 or Y=1) are separated by a feature (x). Here, features that are targets for feature selection are illustrated in FIG. 6. Generally, features having an information value of 0.1 or more can be said to be important features.

In this case, the selection unit 331 may select, for example, a predetermined number of features, for example, five features, in order of high information value. Alternatively, as described above, all features having an information value of 0.1 or more may be selected. In the embodiment of the present invention, five of the 56 features shown in Fig. 6 were selected. Here, the information on the 56 features shown in FIG. 6 may be calculated from individual subscriber information and event information by the selection unit 331.

2. Fine Classing (Step S720)

Detailed classification is a step of dividing a feature having continuous values into classification values in a range of 20 or less. For example, if the reference feature is power consumption, if the data corresponding to 95% of the total subscriber's power usage data is between 250 kWh and 450 kWh, the range of one detailed classification is 10 kWh ((450 kWh-250 kWh)/20 ) Will have a gap. In total, if a feature has less than 20 values (for example, the number of participation of a specific user during the last 10 demand response events), the number of participation in the integer itself is subject to detailed classification.

3. Coarse Classing Step (Step S730)

The selection unit 331 merges some of the classifications from the detailed classification and performs classification adjustment to the rough classification. While detailed classification provides a simple and linear way to define a classification, it is important to explainability to maintain a good balance between the number of values corresponding to the classification (Population). The coarse classification merges the classification steps into more stable and statistically important classifications.

In this case, credit scoring approximates the probabilities of the two groups by logistic regression. At this time, the logit value of logistic regression has a linear relationship with the SCORE variable. (See Equation 4 below) The SCORE variable is the score value of each classification for each feature. Equations 2 and 3 are steps for calculating the logit value of Equation 4.

(Equation 2)

(Equation 3)

(Equation 4)

In this case, X represents each feature, range 1 corresponds to the range of the feature classification step obtained in the detailed classification, and β ₀ and β ₁ are constants of logistic regression. In addition, the left side of equation (4) is a logit value. The selection unit 331 rounds the calculated logit value using the detailed classification step to increase the explanatoryness by using a smaller number of classification steps, and then combines the classification steps having the same value.

4. Score adjustment step through machine learning (S740)

The selection unit 331 finally generates a credit scoring model through score adjustment through machine learning.

According to the present invention, the score (Score) is calculated by dividing a value between 0 and 100 into 0.05 steps, which is a scaling factor. That is, a value divided by 0 to 100 in 2000 steps is set as a score value, and then a meaningful score range is set. In the embodiment of the present invention, for example, the score range was divided based on the increase in area under receiver operation characteristics (AUROC) of 0.001.

In addition, in the embodiment of the present invention, machine learning was performed based on the GBM (Gradient Boosting Method). Machine learning methods may include various methods such as Logistic Regression (LR), Neural Network, and Naive Bayes (NB), but the results of the review show that machine learning based on GBM has the best performance.

The final feature selected according to the embodiment of the present invention and the classification step (Range) and score classification according to the features are shown in FIG. 7.

Referring to FIG. 7, among the 56 demand response-related features shown in FIG. 6, five selected features are a participation rate, a maximum number of consecutive participations, a cumulative reward, and a maximum It is the maximum number of consecutive non-participation and success in the last event (failure to participate in the last event or power failure).

Selection Division (331) Credit score evaluation mode

The selection unit 331 selects a subscriber to be invited to a demand response event among subscribers based on features and scores selected in the credit evaluation scoring learning mode as described above.

For example, if the participation rate of the demand response event up to now is 0.6, the maximum number of consecutive participation is 3 times, the cumulative reward is 500, the maximum number of consecutive non-participation is 6, and the user participated in the last event, see FIG. 7. You will get a credit score of 75+28.35+17.75+9.1+40=95.95 points.

Meanwhile, referring to FIG. 3 again, the transmitter 322 may transmit the guide information (①) for the reduction in power use to the subscriber terminal 400 selected by the selector 331.

For example, the guide information may include a time period in which power saving is required, the amount of power required to be saved, the amount of power that can be saved for the entire power consuming device or an individual power consuming device, and a guide for reducing power use (for example, the air conditioning temperature is changed to several degrees Lower, increase the refrigerator refrigeration/freezing temperature, unplug the unused electronics, etc.), and compensation information for power savings.

The monitoring unit 332 may monitor power consumption by the power consumption device of the selected subscriber after transmission of the guide information. This is to grasp whether the use of power by the subscriber's power consuming device has actually decreased or how much the amount of power has been reduced to correspond to the guide information, and to compensate later.

When the power consumption by the power consumption device of the selected subscriber decreases according to the transmitted guide information, the compensation management unit 333 may provide a predetermined compensation to the corresponding subscriber in proportion to the reduced power usage. More specifically, the compensation management unit 333 may provide compensation for power usage fees.

For example, the compensation management unit 333 may provide compensation in a manner of discounting electricity usage fees or paying points that can be used to pay electricity usage fees. Alternatively, the reward management unit 333 may award cash, gifts, coupons, etc. received from the sponsor server 610 as rewards.

Furthermore, the transmitting unit 322 may transmit information on power saving according to subscriber-specific guide information (hereinafter, saving information) to the power supply-related server 600, and the power supply-related server 600 may reduce information. Compensation corresponding to whether or not the power is included and the amount of power saved can be determined and information about it can be transmitted to the demand response management server 300. Accordingly, the compensation management unit 333 may grant compensation to the corresponding subscriber based on the information on compensation received from the server 600 related to power supply.

On the other hand, the compensation management unit 333, in advance, determines whether or not to reduce the power according to the guidance information and the criteria for awarding compensation based on the amount of power saved, and may provide compensation to the corresponding subscriber based on the award criteria.

The transmitter 322 may transmit the grant information (hereinafter referred to as reward information) ② to the communication device 400 of the corresponding subscriber.

4 is a block diagram of a subscriber terminal 400 having a power demand management function shown in FIG. 1.

According to FIG. 4, the subscriber terminal 400 includes a communication unit 410 receiving guidance information on power usage reduction from the demand response operation server 300, and a control unit for executing a power-related application upon receiving the guidance information ( 420) and an output unit 430 outputting the guide information by executing the power-related application. In addition, the communication device 400 may further include a storage unit 450 that stores power-related applications, guide information, compensation information, and the like.

Here, the control unit 420 may control the operation of at least one of the communication unit 410, the output unit 430, the input unit 440, and the storage unit 440.

Here, the power-related application means an application for managing power use of power-consumer devices owned by a subscriber of the subscriber terminal 400, and is installed at the time of production of the subscriber terminal 400 or downloaded from an external server by subscriber selection Can be installed.

When receiving the guide information, the control unit 420 may automatically execute the power-related application or execute the power-related application only when an execution command is received from the subscriber after providing the notification of the guide information reception to the subscriber.

For example, the guidance information reception notification may display brief contents of the guidance information, or may output a vibration/lamp/notification sound for simply notifying whether reception is received.

In addition, the communication unit 410, receives information on compensation (hereinafter referred to as compensation information) provided in correspondence to the reduction in power use according to the guide information from the demand response operation server 300, and the output unit 430, the power The received compensation information may be output by executing the related application.

Regarding the notification of receiving the compensation information, a description of the notification of receiving the above-described guide information can be cited, and thus detailed description thereof is omitted.

Accordingly, according to the present invention, a server, a communication device and a system having a power demand management function capable of ensuring explanation while utilizing an efficient target selection process when operating a demand response program, and a power demand management method thereof are provided. .

In addition, according to the present invention, since it is possible to encourage participation in demand response events to subscribers who are more likely to participate in demand response events, reduction in power use can be stably performed.

The above description is merely illustrative of the technical spirit of the present invention, and those of ordinary skill in the art to which the present invention pertains can make various modifications, changes, and substitutions without departing from the essential characteristics of the present invention. will be.

Therefore, the embodiments and the accompanying drawings disclosed in the present invention are not intended to limit the technical spirit of the present invention, but to explain, and the scope of the technical spirit of the present invention is not limited by these embodiments and the accompanying drawings. . The scope of protection of the present invention should be interpreted by the claims below, and all technical spirits within the scope equivalent thereto should be interpreted as being included in the scope of the present invention.

Claims (27)

A storage unit for storing subscriber information including power usage for at least one power consumption device of a subscriber and a record of participation of a demand response event by the subscriber, and event information including a date of a demand response event;
Receiving unit for receiving a request signal for power savings;
A selection unit for selecting a subscriber who wants to request power savings by using the subscriber information and event information as the request signal is received;
And a transmitter for transmitting guide information on the reduction in power use to the subscriber terminal of the selected subscriber,
The selection unit generates a credit scoring model, evaluates a subscriber's credit score according to the credit scoring model, and selects a subscriber who wants to request power savings according to the evaluated credit score,
The credit scoring model selects features based on information values for a plurality of features generated based on the subscriber information, and divides the features into a plurality of classification steps, and then, among the plurality of classification steps At least some of them are merged and roughly classified into fewer classification steps.
Server with power demand management. According to claim 1,
The credit scoring model is created by obtaining a logit value for the selected features and roughly classifying the features, and then generating a score range for each range of the rough classification through machine learning.
Server with power demand management. According to claim 2,
The schematic classification is generated by integrating the detailed classified classification range when the selected features are classified and rounded up to the logit value for the detailed classified classification range to have the same logit value.
Server with power demand management. The method of claim 3,
The detailed classification range is 20 or less for one feature,
Server with power demand management. According to claim 2,
The selected features include Participation Rate, Maximum Number of Consecutive Participation, Cumulative Rebate, Maximum Number of Consecutive Non-Participation and Success in Past Events Whether (Success in the Last Event, failed to participate in the last event or power reduction),
Server with power demand management. According to claim 1,
The selection unit generates a plurality of features based on the subscriber information and the event information
Server with power demand management. According to claim 1,
A monitoring unit monitoring power consumption by the selected at least one power consumption device; And
When the power consumption by the power consumption device is reduced according to the guide information, further comprising a compensation management unit for granting a predetermined compensation to the subscriber,
Server with power demand management. The method of claim 7,
The transmitting unit transmits the information on the granted compensation to the subscriber terminal of the corresponding subscriber,
Server with power demand management. In a server including a storage unit, a reception unit, a selection unit, and a transmission unit,
Storing, by the storage unit, subscriber information including power usage for at least one power consumption device of the subscriber and a record of participation of the subscriber's personal demand response event, and event information including a date of the demand response event;
Receiving a request signal for power saving by the receiving unit;
Selecting a subscriber who wants to request power savings by using the subscriber information and the event information as the request signal is received by the selection unit;
And transmitting guide information on the reduction in power use to the communication device of the selected subscriber by the transmitter,
The selecting step includes generating a credit scoring model, and evaluating a subscriber's credit score according to the credit scoring model to select a subscriber who wants to request power savings according to the evaluated credit score.
In the generating of the credit scoring model, features are selected based on information values of a plurality of features generated based on the subscriber information, and divided into a plurality of classification steps for the selected features, and then the Comprising at least a part of the plurality of classification steps to roughly classify into fewer classification steps,
How to manage power demand. The method of claim 9,
Generating the credit scoring model,
Roughly classifying the features by obtaining a logit value for the selected features; And
Generating a score range for each range of the rough classification through machine learning,
How to manage power demand. The method of claim 10,
The rough classification step,
Classifying the selected features in detail; And
Obtaining a logit value for the classified classification range; And
And generating the schematic classification by integrating the detailed classified classification range when the logit value is rounded to have the same logit value.
How to manage power demand. The method of claim 11,
The detailed classification range is 20 or less for one feature,
How to manage power demand. The method of claim 10,
The selected features include Participation Rate, Maximum Number of Consecutive Participation, Cumulative Rebate, Maximum Number of Consecutive Non-Participation and Success in Past Events Whether (Success in the Last Event, failed to participate in the last event or power reduction),
How to manage power demand. The method of claim 9,
The selecting step further includes generating a plurality of features based on the subscriber information and the event information,
How to manage power demand. The method of claim 9,
Monitoring power consumption by the subscriber's power consuming device; And
If the power consumption of the power consumption device is reduced according to the guide information, comprising the step of granting a predetermined compensation to the subscriber
How to manage power demand. The method of claim 15,
The transmitting step includes transmitting the information on the granted compensation to the subscriber terminal of the corresponding subscriber,
How to manage power demand. A storage unit that stores subscriber information including power usage of at least one power consumption device of the subscriber and a record of participation of the subscriber's personal demand response event, and event information including a date of the demand response event; Receiving unit for receiving a request signal for power savings; A selection unit for selecting a subscriber who wants to request power savings by using the subscriber information and event information as the request signal is received; And In the server having a power demand management function comprising a transmitting unit for transmitting the guide information for the power use reduction to the subscriber terminal of the selected subscriber,
Selecting features based on information values for a plurality of features generated based on the subscriber information, and classifying the selected features into a plurality of classification steps;
Obtaining a logit value for the selected features and roughly classifying them into fewer classification steps by merging at least some of the plurality of classification steps of the selected features; And
Including the step of generating a credit scoring model by generating a score range for each range of the rough classification through machine learning;
How to select subscribers for demand response events. The method of claim 17,
The rough classification step,
Classifying the selected features in detail; And
Obtaining a logit value for the classified classification range; And
And generating the schematic classification by integrating the detailed classified classification range when the logit value is rounded to have the same logit value.
How to select subscribers for demand response events. The method of claim 18,
The detailed classification range is 20 or less for one feature,
How to select subscribers for demand response events. The method of claim 18,
The selecting of the features includes generating a plurality of features based on the subscriber information and the event information,
How to select subscribers for demand response events. The method of claim 20,
The selected features include Participation Rate, Maximum Number of Consecutive Participation, Cumulative Rebate, Maximum Number of Consecutive Non-Participation and Success in Past Events Whether (Success in the Last Event, failed to participate in the last event or power reduction),
How to select subscribers for demand response events.


According to claim 1,
The information value is defined by Equation 1 below,
(Equation 1)

Here, IV is an information value, Pr() is a probability function, x is a feature, and Y=1 and Y=0 are groups to be distinguished by feature x,
Server with power demand management.
According to claim 2,
The logit value is calculated by the following equation (4),
(Equation 4)

(Where X is a feature, range 1 is a feature classification step obtained from subclassification, the logistic regression constant, and the left side is the logit value)
Server with power demand management.
The method of claim 9,
The information value is defined by Equation 1 below,
(Equation 1)

Here, IV is an information value, Pr() is a probability function, x is a feature, and Y=1 and Y=0 are groups to be distinguished by feature x,
How to manage power demand.
The method of claim 10,
The logit value is calculated by the following equation (4),
(Equation 4)

(Where X is a feature, range 1 is a feature classification step obtained from subclassification, the logistic regression constant, and the left side is the logit value)
How to manage power demand.
The method of claim 17,
The information value is defined by Equation 1 below,
(Equation 1)

Here, IV is an information value, Pr() is a probability function, x is a feature, and Y=1 and Y=0 are groups to be distinguished by feature x,
How to select subscribers for demand response events.
The method of claim 17,
The logit value is calculated by the following equation (4),
(Equation 4)

(Where X is a feature, range 1 is a feature classification step obtained from subclassification, the logistic regression constant, and the left side is the logit value)
How to select subscribers for demand response events.

Images