Method for Characterizing Multi-Dimensional Energy User Profiles

Abstract

The present invention discloses a method for depicting a multi-dimensional energy user portrait, which includes: obtaining a database in an energy management platform, where the basic data of target users is stored; preprocessing the basic data in the database; performing correlation calculation on the basic data information of system users and the basic data information of target users to obtain the first portrait information of the target users; using statistical methods of coverage rate and time period preference to statistically analyze user behavior characteristics and determine the second portrait information of the target users; finally, forming the third portrait information of the target users through the energy consumption behavior of the target energy users, and predicting the final portrait information of the target users by integrating the first portrait information, the second portrait information, and the third portrait information through multi-dimensional calculation. The present invention uses rich and reliable data sources to not only reduce the situation of inaccurate information structure of the energy target user portrait, but also depict more accurate and complete user portrait information.

Description

Technical Field

[0001] The present invention relates to the technical field of energy Internet, and particularly relates to a method for depicting a multi-dimensional energy user portrait. Background Art

[0002] With the continuous deepening of the construction of the energy Internet, the energy-related industries have actively innovated from top to bottom to promote the efficient and reasonable utilization of energy. During the construction process of the energy Internet, the coordinated and complementary energy between various types of energy is enhanced, and the future integrated energy utilization system will gradually evolve into a user-centered integrated energy system. Among them, the importance of the user portrait is increasing day by day. Only by accurately depicting the user portrait and clearly analyzing the user's behavior trajectory can the value of data be better exerted and the value of energy use be maximized.

[0003] In the prior art, the methods for depicting user portraits mainly include the following categories. One is that the user fills in their own portrait data. The problem with this method is that the user's filling may be incorrect or the information filled in by the user is incomplete. The other is to use rules to create a customer user portrait, and draw a user portrait through the user's behavior data in the past month and the geographical location rules of daily communication. However, the above-mentioned techniques for depicting energy user portraits are not mature, resulting in inaccurate portrait results of energy users. Summary of the Invention

[0004] The main object of the present invention is to provide a method for depicting a multi-dimensional energy user portrait, aiming to solve the problem that the existing all-round information depicting technology for energy user portraits is not mature, resulting in inaccurate portrait results of energy users.

[0005] To solve the above problems, the present invention is achieved through the following technical solutions:

[0006] A method for depicting a multi-dimensional energy user portrait, comprising: obtaining a database in an energy management platform, where the database stores basic data of a target user. Preprocessing the basic data in the database.

[0007] Obtaining basic data information of the target user within a first preset time period.

[0008] Obtaining basic data information of a system user within a second preset time period; wherein, the first preset time period is less than the second preset time period, and the first preset time period is included in the second preset time period.

[0009] Performing correlation calculation on the basic data information of the system user and the basic data information of the target user to obtain first portrait information of the target user.

[0010] Calculate the second portrait information of the target user by using a statistical method based on coverage rate and time period preference.

[0011] Predict the third portrait information of the target user through the energy consumption behavior information of the target user by using a GRU algorithm model based on deep learning.

[0012] Determine the final portrait information of the target user by combining the first portrait information, the second portrait information, and the third portrait information.

[0013] Optionally, the step of preprocessing the basic data in the database includes:

[0014] Eliminate the missing values and outliers in the data information of the target user in the database.

[0015] Optionally, the system user is a user with corresponding portrait information. The portrait information includes tag information and corresponding tag content. The basic data information includes industry information, account number information, energy consumption information, electricity consumption information, energy supply unit information, power distribution location information, geographical location information, and time information. The energy consumption information includes carbon emissions and corresponding equipment names.

[0016] Optionally, the step of performing correlation calculation on the basic data information of the system user and the basic data information of the target user to obtain the first portrait information of the target user includes:

[0017] Correlate the industry information of the target user with the industry information of the system user, the account number information of the target user with the account number information of the system user, the energy consumption information of the target user with the energy consumption information of the system user, correlate the electricity consumption information of the target user with the electricity consumption information of the system user, the energy supply unit information of the target user with the energy supply unit information of the system user, the power distribution location information of the target user with the power distribution location information of the system user, correlate the geographical location information of the target user with the geographical location information of the system user, and correlate the time information of the target user with the time information of the system user to obtain a correlation comparison result; based on the correlation comparison result, determine the first portrait information of the target user.

[0018] Optionally, the step of determining the first portrait information of the target user based on the associated comparison result includes: based on the associated comparison result, marking system users whose distance from the geographical location of the target user is less than a preset distance as relevant users. Based on the portrait information of the relevant users, obtaining the label content that is the same between the target user and the relevant users, and marking it as relevant content. Setting the relevant content and the corresponding label information as the first portrait information of the target user.

[0019] Optionally, the step of calculating the second portrait information of the target user by using the statistical method of coverage rate and time period preference includes: defining a certain type of business behavior and the behavior state that occur to the target user at a certain moment as a behavior label T, and the behavior label T is the second portrait information. Representing the behavior characteristics represented by each behavior label T with frequency, average value, coverage rate, deviation degree, average time interval, periodic characteristics, and time period preference characteristics. The portrait information includes behavior label information and the corresponding label content. The business behaviors with the behavior label T include the following categories: the average monthly energy consumption of the target user, the highest annual energy consumption value, the lowest annual energy consumption value, the energy consumption change amount, the energy consumption change rate, the peak energy consumption period, and the payment situation.

[0020] Optionally, the step of calculating the second portrait information of the target user by using the statistical method of coverage rate and time period preference further includes:

[0021] If the coverage rate of the business behavior at of the jth behavior label that occurs to the target user u within a certain time period is represented by CreateRatio, then the calculation formula for the coverage rate CreateRatio is: j In the formula, ET - ST is the statistical time length, and sum(at

[0022]

[0023] , u) j represents the total number of times that the target user u has the business behavior at within the time period of ET - ST; j represents the jth behavior label selected from all the behavior labels of the target user u, and at ET-ST represents the business behavior of the jth behavior label of the target user u; sum(at j , u) j represents the total number of times that the target user u has the business behavior at i , u) ET-ST represents the total number of times that the target user u has the business behavior at i within the time period of ET - ST, and the value range of i is [0, n], where i represents the ith behavior label selected from all the behavior labels of the target user u, and at iThe business behavior with the i-th behavior label of the target user u; the business behavior at of the target user u with the j-th behavior label j The average time interval is represented by average(d), and the calculation formula of average(d) is:

[0024]

[0025] where n represents the number of times the target user u has the behavior at j the business behavior; d J represents the J-th time interval when the target user u has the behavior at j the business behavior; the deviation degree of the business behavior at of the target user u with the j-th behavior label j The calculation formula is:

[0026]

[0027] where n represents the number of times the target user u has the behavior at j the business behavior; d J represents the J-th time interval when the target user u has the behavior at j the business behavior; divide the start and end time range of the business behavior at of the target user u with the j-th behavior label into n behavior intervals D1, D2,... D j , then the periodicity period(at n , u) of the business behavior at is calculated as j j

[0028]

[0029] In the formula, sum(d I , u) represents the number of times the behavior interval d of the target user u appears, and the value range of I is [1, n]; d I represents the I-th time interval when the target user u has the behavior at I the business behavior; if the business behavior at j has no periodicity, it is represented by 0. If the number of times sum(d j , u) of the business behavior at of the target user u in this time period accounts for 60% of the total number of behavior occurrences, then the business behavior at j has a time period preference, and the time period preference feature TF is this time period. I , u) accounts for 60% of the total number of behavior occurrences, then the business behavior at j has a time period preference, and the time period preference feature TF is this time period.

[0030] Optionally, the step of predicting the predicted label information by the deep learning-based GRU algorithm to obtain the third portrait information of the target user includes: marking the label information to be predicted of the target user as the label to be predicted. The content of the label to be predicted includes: prediction of the energy consumption of the target user at a certain future moment, prediction of the energy consumption behavior, and prediction of the energy consumption change; establishing a deep learning-based GRU algorithm model; using the label to be predicted of the target user and the corresponding label content as training data to train the deep learning-based GRU algorithm model; predicting the label to be predicted of the target user and the corresponding label content through the trained deep learning-based GRU algorithm model, and using them as the third portrait information.

[0031] Optionally, the step of using the information to be predicted of the target user as training data to train the deep learning-based GRU algorithm model includes: the state h transmitted from the previous node t-1 and the input value x of the current node t t to obtain two gated states. After obtaining the gated states, use the reset gate to obtain the reset data h t-1’ , and then splice the data h t-1’ with the input value x t . Squeeze the spliced data into the range of [-1, 1] through the tanh activation function to obtain the candidate state h of the current node state ’ . Forget and remember the candidate state h of the current node state at the same time ’ .

[0032] Optionally, the gated states include an update gate state and a reset gate state; the update gate is used to control the degree to which the state information of the previous moment is retained in the current state, and the larger the value of the update gate, the more state information of the previous moment is retained; the reset gate is used to determine whether to combine the current state with the previous information, and the smaller the value of the reset gate, the more information is ignored.

[0033] The state update method of the current node state h t is as follows:

[0034] h t = z⊙h t-1 + (1 - z)⊙h'

[0035] where z is the update gate state, h ’ represents the candidate state at the current moment; x t represents the input value of the current node t; W represents the parameter of the memory gate neuron, which is learned during the training process.

[0036] Optionally, the step of determining the final portrait information of the target user by combining the first portrait information, the second portrait information, and the third portrait information includes: packing the first portrait information, the second portrait information, and the third portrait information to form the final portrait information corresponding to the target user.

[0037] The present invention has at least one of the following advantages:

[0038] The method for depicting a multi-dimensional energy user portrait proposed by the present invention determines the first portrait information of the target user by associating data information dimensions, determines the second portrait information of the target user by using statistical methods of coverage rate and time period preference, and finally obtains the label information to be predicted for the target energy user through deep learning, thereby forming the third portrait information of the target user. Through such multi-dimensional calculations and integrating the above first portrait information, second portrait information, and third portrait information, the final portrait information of the target user can be predicted. The rich and reliable data source not only reduces the inaccurate situation of the energy target user portrait information structure, but also can depict more accurate and complete user portrait information. The basic data information adopted by the present invention belongs to the static energy consumption information of the target user. Now, information such as energy suppliers is added, which is not only convenient for the use of the target user, but also provides an information basis for the future energy consumption service (i.e., the third portrait) of the supplier for the target user. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a flowchart of the first embodiment of a method for depicting a multi-dimensional energy user portrait proposed by the present invention;

[0040] Figure 2 It is an input-output structure diagram of the GRU algorithm of the sixth embodiment of a method for depicting a multi-dimensional energy user portrait proposed by the present invention;

[0041] Figure 3 It is a structural schematic diagram of the reset gate and the update gate of the seventh embodiment of a method for depicting a multi-dimensional energy user portrait proposed by the present invention;

[0042] Figure 4 It is a schematic diagram of the process of obtaining the current node state h t of the seventh embodiment of a method for depicting a multi-dimensional energy user portrait proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] The following further elaborates in detail on a method for depicting a multi-dimensional energy user profile proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. According to the following description, the advantages and features of the present invention will be clearer. It should be noted that the accompanying drawings are in a very simplified form and use non-precise scales, only for conveniently and clearly assisting in explaining the purpose of the embodiments of the present invention. In order to make the purpose, features, and advantages of the present invention more obvious and understandable, please refer to the accompanying drawings. It should be noted that the structures, scales, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those skilled in this technology to understand and read, and are not used to limit the limiting conditions for the implementation of the present invention. Therefore, they do not have technical substance significance. Any modification of the structure, change in the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope covered by the technical content disclosed by the present invention.

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the accompanying drawings required for describing the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on the structures shown in these drawings.

[0045] Embodiment 1

[0046] As Figure 1 shown, this embodiment provides a method for depicting a multi-dimensional energy user profile, including:

[0047] Step S100: Obtain the database in the energy management platform, where the database stores the basic data of the target user. The energy management platform can also be referred to as the "energy brain", which is a smart city energy cloud platform built using technologies such as big data analysis, artificial intelligence, and cloud computing.

[0048] Step S101: Preprocess the basic data in the database. In this embodiment, step S101 specifically includes: removing the missing values and abnormal values in the data information of the target user in the database.

[0049] It can be understood that the preprocessed data can be used for label definition and operations. In this embodiment, it includes basic data and energy consumption (i.e., static and dynamic data). The static data serves as the basis for the associated calculation of the first profile information, and the dynamic data serves as the basis for the second and third profile information for statistics, calculation, and prediction.

[0050] Step S102: Obtain the basic data information of the target user within the first preset time period.

[0051] Obtain the basic data information of the system user within the second preset time period; wherein, the first preset time period is less than the second preset time period, and the first preset time period is included in the second preset time period. Perform correlation calculation on the basic data information of the system user and the basic data information of the target user to obtain the first portrait information of the target user.

[0052] In this embodiment, the first preset time period is preferably 1 month here, but the present invention is not limited thereto. The basic data information here can be remotely obtained through a smart electricity meter.

[0053] For example, when obtaining target user A's energy consumption, the energy consumption information is {"emission factor": "3", "energy_name": "energy of Xiaoming's family"}, the electricity consumption information is {"electricity level": "first level", "consumption level": "high"}, and the geographical location information is {"wm629s9": "No. 1000, Area A"; traffic place name: Road A"}, and the first preset time is: June 1, 2021 - July 1, 2021.

[0054] Wherein, the system user is a user with corresponding portrait information, and the portrait information includes tag information and corresponding tag content; the basic data information includes industry information, household number information, energy consumption information, electricity consumption information, energy supply unit information, power distribution location information, geographical location information, and time information. The first preset time period is less than the second preset time period, and the first preset time period is included in the second preset time period, and the energy consumption information includes carbon emissions and the corresponding equipment name.

[0055] In this embodiment, the second preset time period is 6 months here, but the present invention is not limited thereto.

[0056] For example, the energy consumption information of a certain system user Y is {"emission factor": "3", "energy_name": "energy of System Y0's family"}, the electricity consumption information is {"electricity level": "first level", "consumption level": "low"}, and the geographical location information is {"wm629s9": "No. 10010, Area A"; traffic place name; Road A"}, and the second preset time is: January 1, 2021 - July 1, 2021.

[0057] For example: The tag information in the portrait information includes: basic data information, turnover, industry category, city classification, supply voltage, load nature, energy usage category, energy activity, energy consumption level, carbon emissions, energy proportion, electricity level, seasonal electricity peak, power supply quality perception, illegal electricity usage degree, electricity usage category, energy consumption type, meter reading method, meter reading cycle, bill release date, payment deadline date, and consumption level.

[0058] Each tag information corresponds to tag content. For example, the tag content of turnover is XX yuan; the tag content of power supply voltage is 220V, etc.

[0059] In this embodiment, the first portrait information here is the portrait information of the system user who is relatively close to the target user. For users at a relatively close distance, some portrait information is the same, such as the peak season of electricity consumption, electricity quantity level, power supply voltage, city classification, etc. in the basic data information.

[0060] Step S103: Calculate the second portrait information of the target user by using the statistical methods of coverage rate and time period preference.

[0061] The coverage rate represents the proportion of the number of occurrences of a certain business behavior within a certain time period to the total number of occurrences of similar business behaviors. The average time interval of behavior is the average of the time intervals of the behavior tags. The deviation degree is the standard deviation of the time intervals of the behavior tags, reflecting the time uniformity of the user's generation of a certain behavior. The lower the deviation degree, the more likely this behavior is a periodic behavior. Periodicity is used to measure whether a certain behavior of the user is periodic. The time period preference feature represents the time period preference for the generation of the user's behavior. The above features are all described from the historical state and the recent state, highlighting the time characteristics of a certain behavior.

[0062] In this embodiment, step S103 includes: defining a certain business behavior and its behavior state that occur to the target user at a certain moment as a behavior tag T, that is, using the behavior tag T as the second portrait information; representing the behavior characteristics represented by each behavior tag T with frequency, average value, coverage rate, deviation degree, average time interval, cycle feature and time period preference feature; the business behaviors with the behavior tag T include the following categories: the monthly average energy consumption of the target user, the highest annual energy consumption value, the lowest annual energy consumption value, the energy consumption change amount, the energy consumption change rate, the energy consumption peak period, and the payment situation.

[0063] In this embodiment, step S103 further includes: the business behavior at of the jth behavior tag occurs to the target user u within a certain time period j The coverage rate is represented by CreateRatio, and the calculation formula for the coverage rate CreateRatio is:

[0064]

[0065] In the formula, ET - ST is the statistical time length, and sum(at j , u) ET-ST represents that the target user u has the business behavior at within the time period of ET - ST jThe sum of times; j represents the j-th behavior label among all the behavior labels of the target user u selected, at j represents the business behavior of the j-th behavior label of the target user u; sum(at i , u) ET-ST represents the sum of the number of times the business behavior at occurs for the target user u during the period from ET to ST, where the value range of i is [0, n], and i represents the i-th behavior label among all the behavior labels of the target user u selected, at i represents the sum of the number of times the business behavior at occurs for the target user u during the period from ET to ST, where the value range of i is [0, n], and i represents the i-th behavior label among all the behavior labels of the target user u selected, at i represents the business behavior of the i-th behavior label of the target user u.

[0066] The business behavior at of the j-th behavior label occurs for the target user j The average time interval between occurrences is represented by average(d), and the calculation formula for average(d) is:

[0067]

[0068] where n represents the number of times the business behavior at occurs for the target user u; d j represents the J-th time interval when the target user u performs the business behavior at; J represents the J-th time interval when the target user u performs the business behavior at; j represents the J-th time interval when the target user u performs the business behavior at;

[0069] The deviation degree of the business behavior at of the j-th behavior label of the target user u j is The calculation formula is:

[0070]

[0071] where n represents the number of times the business behavior at occurs for the target user u; d j represents the J-th time interval when the target user u performs the business behavior at; J represents the J-th time interval when the target user u performs the business behavior at; j represents the J-th time interval when the target user u performs the business behavior at;

[0072] Divide the start and end time range of the business behavior at of the target user u with the j-th behavior label into n behavior intervals D1, D2,... D j n. n .

[0073] Then the periodicity period(at j , u) of the business behavior at j is calculated by the formula

[0074]

[0075] In the formula, sum(d I, u) represents the behavior interval d of the target user u I The number of occurrences, and the value range of I is [1, n]; d I represents the I-th time interval when the target user u has the business behavior at j If the business behavior at j has no periodicity, it is represented by 0.

[0076] If the number of times sum(d j of the business behavior at of the target user u in this time period I , u) accounts for 60% of the total number of occurrences of the behavior, then the business behavior at j has a time period preference, and the time period preference feature TF is this time period. Step S104: The GRU algorithm model based on deep learning predicts the third portrait information of the target user through partial portrait information of the system user.

[0077] Specifically, the portrait information of the target user obtained through the statistical methods of association calculation, coverage rate, and time period preference may still not be comprehensive enough, and some label information is still missing. Then, the partial labels of the system user with existing portrait information can be obtained through deep learning.

[0078] First, generate a GRU algorithm model based on deep learning. Then, use the labels to be predicted of the target user and the corresponding label contents as training data to train the GRU algorithm model based on deep learning. Here, the label content to be predicted is mainly the energy prediction, energy consumption behavior prediction, and energy consumption change prediction for a certain future moment.

[0079] Step S105: Determine the final portrait information of the target user by combining the first portrait information, the second portrait information, and the third portrait information.

[0080] Through the above steps, the first portrait information, the second portrait information, and the third portrait information of the target user can be obtained, and then the final portrait information of the target user can be obtained.

[0081] The method for depicting the multi-dimensional energy user portrait proposed in this embodiment determines the first portrait information of the target user by associating the data information dimension, calculates and determines the second portrait information of the target user by using the statistical methods of coverage rate and time period preference. Finally, the label information to be predicted of the target energy user is obtained through deep learning, and then the third portrait information of the target user is formed. Through such multi-dimensional calculations and integrating the above first portrait information, second portrait information, and third portrait information, the final portrait information of the target user can be predicted. The rich and reliable data source not only reduces the inaccurate situation of the energy target user portrait information structure, but also can depict more accurate and complete user portrait information.

[0082] Embodiment 2

[0083] Based on the first embodiment, the difference between this embodiment and the first embodiment is that the step S102 further includes the following steps:

[0084] Step S1021: Associate the industry information of the target user with the industry information of the system user, associate the account number information of the target user with the account number information of the system user, associate the energy consumption information of the target user with the energy consumption information of the system user, associate the electricity consumption information of the target user with the electricity consumption information of the system user, associate the energy supply unit information of the target user with the energy supply unit information of the system user, associate the power distribution location information of the target user with the power distribution location information of the system user, associate the geographical location information of the target user with the geographical location information of the system user, and associate the time information of the target user with the time information of the system user to obtain an association comparison result.

[0085] Step S1022: Determine the first portrait information of the target user based on the association result.

[0086] Specifically, the association result here is to compare the energy consumption information of the target user with the industry information of the system user, compare the energy consumption information of the target user with the energy consumption information of the system user, compare the energy consumption information of the target user with the energy supply unit information of the system user, compare the energy consumption information of the target user with the power distribution location information of the system user, compare the electricity consumption information of the target user with the electricity consumption information of the system user, compare the geographical location information of the target user with the geographical location information of the system user, and compare the time information of the target user with the time information of the system user.

[0087] Embodiment Three

[0088] Based on the above second embodiment, the difference between the method for depicting a multi-dimensional energy user portrait proposed in this embodiment and the above second embodiment is that the step S1022 includes the following steps:

[0089] Step S10221: Based on the association result, mark the system users whose distance from the geographical location of the target user is less than a preset distance as relevant users.

[0090] Specifically, the reasonable preset distance can be 1 kilometer, but the present invention is not limited thereto, that is, by comparing the geographical location information of the target user with the geographical location information of the system user, the system users whose distance from the geographical location of the target user is less than 1 kilometer are marked as relevant users.

[0091] If there are multiple system users with a distance less than the preset distance, the system user with the smallest distance is marked as the relevant user. Or the system users with closer energy consumption information and electricity consumption information are marked as relevant users.

[0092] Step S10222: Based on the portrait information of the relevant user, obtain the label content that the target user is consistent with the relevant user, and mark it as relevant content.

[0093] Specifically, since the target user and the relevant user are relatively close in distance, there will be some label contents that are the same, such as: peak season electricity consumption, electricity level, supply voltage, city classification; these label contents are marked as relevant content.

[0094] Step S10223: Set the relevant content and the corresponding label information as the first portrait information of the target user.

[0095] Specifically, that is, the peak season electricity consumption, electricity level, supply voltage, and city classification of the relevant user are used as the first portrait information of the target user.

[0096] Embodiment 4

[0097] In the fourth embodiment of, based on the first embodiment, step S104 in a method for depicting a multi-dimensional energy user portrait proposed in this embodiment further includes the following steps:

[0098] Step S1041: Mark the label information that the target user does not have and the relevant user has as the label to be predicted.

[0099] Specifically, for example: the portrait information of the target user does not have the label information of energy consumption level, and the portrait information of the relevant user has this label information, then the energy consumption level is the label to be predicted corresponding to the target user.

[0100] Step S1042: Establish a GRU algorithm model based on deep learning.

[0101] Step S1043: Use the label to be predicted and the corresponding label content of the relevant user as training data to train the GRU algorithm model based on deep learning.

[0102] Step S1044: Predict the label to be predicted and the corresponding label content of the target user through the trained GRU algorithm model based on deep learning, and use them as the third portrait information.

[0103] Specifically, the first portrait information and the second portrait information of the target user obtained through associated data information and the nearest neighbor algorithm dimension may still not comprehensively reflect the portrait of the target user, that is, some label information of the target user is still missing. In this case, some labels of the relevant users of the existing portrait information can be obtained through deep learning.

[0104] First, train the partial label information of the relevant users to generate a training module. The partial labels are the labels that the target user does not carry. In this patent, the system has predicted most of the portrait information of the target user through multiple dimensions, and the rest can be predicted through deep learning.

[0105] First, generate a GRU algorithm model based on deep learning. Then, train the model with the partial label information of the system users with existing portrait information. Some of the partial label information here is the label information that the target user does not have yet. For example, through the above steps, the system has predicted label information such as the industry classification, power level, turnover, and city classification of the target user. However, the target user does not have the label information about energy activities yet. In this case, the label information about energy activities and the corresponding label content of the system users associated with the target user can be used for deep learning to obtain the label information about energy activities of the target user and the corresponding label content. And the label information about energy activities and the corresponding label content are used as the third portrait information of the target user.

[0106] The GRU algorithm (Gate Recurrent Unit) is a type of Recurrent Neural Network (RNN). Similar to LSTM (Long-Short Term Memory), it is also proposed to solve problems such as long-term memory and gradients in backpropagation.

[0107] RNN is suitable for analyzing and processing time series data because RNN introduces a recurrent unit structure in the network and allows internal connections between hidden units, making it possible to explore the temporal relationships between non-consecutive data. However, RNN has the problem of gradient vanishing, which causes RNN to lose the ability to learn information from the more distant past as the time interval increases.

[0108] The proposed LSTM neural network solves the problem of gradient vanishing in RNN and has been widely used in the field of predicting time series data. In recent years, many variants have emerged according to different needs. As a variant of LSTM, GRU adopts a gated recurrent neural network structure, has fewer training parameters than LSTM, and at the same time maintains the prediction effect of LSTM.

[0109] The internal unit of the GRU algorithm is very similar to that of the LSTM. The difference is that GRU combines the input gate and the forget gate in the LSTM into a single update gate. Therefore, there are only two gate structures in GRU, namely the update gate and the reset gate. The update gate is used to control the degree to which the state information of the previous moment is retained in the current state. The larger the value of the update gate, the more state information of the previous moment is retained. The reset gate is used to determine whether to combine the current state with the previous information. The smaller the value of the reset gate, the more information is ignored.

[0110] As shown in the Figure 2 appendix, the input-output structure of the GRU algorithm is the same as that of the ordinary RNN.

[0111] There is an input value x t for the current node t, and a hidden state h t-1 passed down from the previous node. This hidden state contains the relevant information of the previous nodes.

[0112] Combining the input value x t and the hidden state h t-1 , the GRU algorithm will obtain the output y t of the current hidden node and the hidden state h t passed to the next node.

[0113] Two gated states are obtained through the state h t-1 transmitted from the previous node and the input value x t of the current node t.

[0114] Specifically, as shown in the Figure 3 appendix, where r controls the reset gate and z controls the update gate.

[0115] After obtaining the gated states, the reset gate is used to obtain the reset data h t-1’ , and then the data h t-1’ is concatenated with the input value x t . It can be understood from this that the state h t-1 becomes the value h t-1’ after being processed by the reset gate.

[0116] The concatenated data is scaled to the range of [-1, 1] through the tanh activation function to obtain the candidate state h ’ of the current node state.

[0117] Specifically, the process of obtaining the current node state h t is as shown in the Figure 4 appendix. Here, h tIt mainly contains the x of the current input value t data. Specifically for h t Added to the current hidden state, which is equivalent to "memorizing the state at the current moment".

[0118] The current node state candidate state h ’ Perform forgetting and memorization simultaneously.

[0119] Specifically, this step uses the previously obtained update gate z (update gate), which is called the "update memory stage".

[0120] Embodiment Five

[0121] In the fifth embodiment of [[ID=]], based on the fourth embodiment, in the method for depicting a multi-dimensional energy user portrait proposed in this embodiment, the gated state includes an update gate state and a reset gate state; the update gate is used to control the degree to which the state information at the previous moment is retained in the current state, and the larger the value of the update gate, the more state information at the previous moment is retained; the reset gate is used to determine whether to combine the current state with the previous information, and the smaller the value of the reset gate, the more information is ignored.

[0122] Embodiment Six

[0123] Based on the fifth embodiment, in the method for depicting a multi-dimensional energy user portrait proposed in this embodiment, the method for updating the state of the current node h t is as follows:

[0124] h t = z ⊙ h t-1 + (1 - z) ⊙ h'

[0125] where z is the update gate state, h ’ represents the candidate state at the current moment; x t represents the input value of the current node t; W represents the parameters of the memory gate neuron, which are learned during the training process; h ’ mainly contains the x of the current input t data, which is equivalent to memorizing the state at the current moment.

[0126] Embodiment Seven

[0127] Based on the first embodiment or the sixth embodiment, step S105 in the method for depicting a multi-dimensional energy user portrait proposed in this embodiment includes the following steps:

[0128] Pack the first portrait information, the second portrait information, and the third portrait information to form the final portrait information corresponding to the target user.

[0129] Specifically, when obtaining the portrait information of the target user through multi-dimensional calculations, there may be similar data information. Specifically, the system extracts the similar data information, comprehensively organizes the data information, and statistically obtains the most accurate data information.

[0130] Then, by screening and synthesizing the similar tag information of the first portrait information, the second portrait information, and the third portrait information; combining the first portrait information, the second portrait information, and the third portrait information to determine the final portrait information of the target user.

[0131] In summary, the rich and reliable data sources of the "Energy Brain" in this embodiment not only reduce the inaccurate situation of the energy target user portrait information structure, but also can depict more accurate and complete user portrait information. The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments.

[0132] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including several instructions to enable a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0133] The embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the purpose of the present invention and the scope protected by the claims. These all belong to the protection scope of the present invention.

[0134] Although the content of the present invention has been introduced in detail through the above preferred embodiments, it should be recognized that the above description should not be considered as a limitation of the present invention. After those skilled in the art read the above content, various modifications and substitutions of the present invention will be obvious. Therefore, the protection scope of the present invention should be defined by the appended claims.

Claims

1. A method for depicting a multi-dimensional energy user profile, characterized in that, include: Obtaining a database in the energy management platform, wherein the database stores basic data of the target user; Preprocessing the basic data in the database; Obtain basic data information of the target user within a first preset time period; Obtaining basic data information of the system user within a second preset time period; wherein the first preset time period is shorter than the second preset time period, and the first preset time period is included in the second preset time period; Performing correlation calculation on the basic data information of the system user and the basic data information of the target user to obtain first portrait information of the target user; Associating the target user's industry information with the system user's industry information, associating the target user's account number information with the system user's account number information, associating the target user's energy consumption information with the system user's energy consumption information, associating the target user's power consumption information with the system user's power consumption information, associating the target user's energy supply unit information with the system user's energy supply unit information, associating the target user's power distribution location information with the system user's power distribution location information, associating the target user's geographic location information with the system user's geographic location information, and associating the target user's time information with the system user's time information to obtain an association comparison result; Determining first portrait information of the target user based on the association comparison result; Calculating the second profile information of the target user using a statistical method of coverage and time period preference; A certain type of business behavior and behavior status of the target user at a certain moment is defined as a behavior label T, and the behavior label T is the second portrait information; The behavioral characteristics represented by each behavioral label T are expressed by frequency, average value, coverage, deviation, average time interval, period characteristics and time period preference characteristics; The business behaviors with the behavior label T include the following categories: the target user's average monthly energy consumption, the annual maximum energy consumption, the annual minimum energy consumption, the energy consumption change, the energy consumption change rate, the energy consumption peak period, and the payment status; Mark the label information that needs to be predicted for the target user as the label to be predicted; The content of the tags to be predicted includes: energy consumption prediction of the target user at a certain moment in the future, energy consumption behavior prediction, and energy consumption change prediction; Establish a GRU algorithm model based on deep learning; The target user's predicted label and the corresponding label content are used as training data to train the GRU algorithm model based on deep learning; The trained deep learning-based GRU algorithm model is used to predict the target user's pre-label and corresponding label content, which are used as the third profile information. The final portrait information of the target user is determined by combining the first portrait information, the second portrait information and the third portrait information.

2. The method for depicting a multi-dimensional energy user portrait according to claim 1, characterized in that: The step of preprocessing the basic data in the database includes: Missing values and outliers in the target user's data in the database are removed.

3. The method for depicting a multi-dimensional energy user portrait according to claim 1, wherein the system user is a user with corresponding portrait information; the portrait information includes tag information and corresponding tag content; the basic data information includes industry information, account number information, energy consumption information, electricity consumption information, energy supply unit information, power distribution location information, geographical location information, and time information; the energy consumption information includes carbon emissions and corresponding equipment names.

4. The method for depicting a multi-dimensional energy user portrait according to claim 3, wherein the step of determining the first portrait information of the target user based on the correlation comparison result includes: based on the correlation comparison result, marking system users whose distance from the geographical location of the target user is less than a preset distance as relevant users; based on the portrait information of the relevant users, obtaining the tag content that is the same between the target user and the relevant users, and marking it as relevant content; setting the relevant content and the corresponding tag information as the first portrait information of the target user.

5. The method for depicting a multi-dimensional energy user profile according to claim 4, wherein The step of calculating the second portrait information of the target user by using the statistical methods of coverage rate and time period preference further includes: The business behavior at of the j-th behavior label occurs for the target user u within a certain time period j If the coverage rate is represented by CreateRatio, the calculation formula for the coverage rate CreateRatio is as follows: where ET - ST is the statistical time length, sum(at j , u) ET-ST represents the total number of times that the business behavior at occurs for the target user u during the period of ET - ST j ; j represents the j-th behavior label among all the behavior labels of the selected target user u, and at j represents the business behavior of the j-th behavior label of the target user u; sum(at i ,u) ET-ST It represents the target user u’s business behavior at during the period of ET-ST. i The sum of the number of times, the value range of i is [0, n], i represents the i-th behavior label among all the behavior labels of the target user u, at i Indicates the business behavior of the target user u's i-th behavior label; The business behavior at of the j-th behavior label occurs for the target user u j If the average time interval of occurrence is represented by average(d), the calculation formula of average(d) is as follows: Among them, n represents the target user u at j Number of business activities; d J Indicates that the target user u has at j The Jth time interval during business behavior; The target user u performs the business behavior at of the jth behavior label j The deviation The calculation formula is: Among them, n represents the number of times the target user u performs the at j business behavior; d J represents the Jth time interval when the target user u performs the at j business behavior; The target user u performs the business behavior of the jth behavior label at j The start and end time range is divided into n behavior intervals D1, D2, ...D n , then the business behavior at j Periodicity period(at j ,u) is calculated as In the formula, sum(d I ,u) represents the behavior interval d of target user u I The number of occurrences, the value range of I is [1,n]; d I Indicates that the target user u has at j The first time interval during business activities; If the business behavior at j has no periodicity, it is represented by 0; If the target user u's business behavior at j The number of times it occurs in this time period sum(d I ,u) accounts for 60% of the total number of times the behavior occurs, then the business behavior at j It has a time period preference, and the time period preference feature TF is the time period.

6. The method for depicting a multi-dimensional energy user portrait according to claim 5, wherein the step of training the GRU algorithm model based on deep learning by using the to-be-predicted information of the target user as training data includes: The state h transmitted from the previous node t-1 and the input value x of the current node t t to obtain two gated states; After getting the gate state, use the reset gate to get the data h after reset t-1’ , and then the data h t-1’ With the input value x t Perform splicing; scaling the concatenated data to the range of [-1, 1] through the tanh activation function to obtain the candidate state h' of the current node state; simultaneously performing forgetting and memorization on the candidate state h' of the current node state.

7. The method for depicting a multi-dimensional energy user portrait according to claim 6, wherein the gated state includes an update gate state and a reset gate state; the update gate is used to control the degree to which the state information of the previous moment is retained in the current state, and the larger the value of the update gate, the more state information of the previous moment is retained; the reset gate is used to determine whether to combine the current state with the previous information, and the smaller the value of the reset gate, the more information is ignored; The state update method for the current node state h t is as follows: h t = z ⊙ h t-1 + (1 - z) ⊙ h' where z is the update gate state, h’ represents the candidate state at the current time; x t represents the input value of the current node t; W represents the parameter of the memory gate neuron, which is learned during the training process; the step of determining the final portrait information of the target user by combining the first portrait information, the second portrait information, and the third portrait information includes: packing the first portrait information, the second portrait information, and the third portrait information to form the final portrait information corresponding to the target user.

Images