Omnichannel power grid marketing management method and system based on data center

Through the omni-channel grid marketing management method of the data center, the LSTM long and short-term memory method is used to identify user categories and electricity loads, and dynamic adjustment of power supply in the distribution network is solved, which is inefficient in the dynamic distribution power supply strategy of users in the intelligent distribution network. The problem has improved the economic operation and efficiency of the distribution network.

CN119048184APending Publication Date: 2024-11-29STATE GRID SHANXI MARKETING SERVICE CENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411175108.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-26
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing smart distribution network technology is difficult to achieve dynamic distribution power supply strategies for different types of users, resulting in low distribution efficiency.

Method used

Using the omni-channel power grid marketing management method based on the data center, the user registration information is extracted and power load analysis is performed through the LSTM long and short-term memory method, the user category is identified and the active power is calculated, and the power supply of the distribution network is dynamically adjusted. .

Benefits of technology

It improves the safety and reliability of user electricity use, realizes the economic operation of the distribution network, saves losses and costs, and maximizes the utilization of benefits.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048184A_ABST
    Figure CN119048184A_ABST
Patent Text Reader

Abstract

The invention provides an omni-channel power grid marketing management method and system based on a data center. The method comprises the following steps: acquiring user registration information of a cloud database of the data center, power consumption data information of power distribution network users and electricity price information of different regions; keyword extraction is carried out, and the electricity utilization category of the user is identified; extracting electrical load amplitudes of different types of users; alternating current load periodic wave signals used by different types of users are extracted, and an analysis signal s (t) and an instantaneous phase theta u (t) of the alternating current load periodic wave signals are obtained through Hilbert transform; amplitude values and phases of current signals and voltage signals of different types of users are obtained, active power of the different types of users is calculated, and peak segments, valley segments and flat segments of the active power of the alternating current electric load used by the different types of users are identified; and S2, controlling the power distribution network to dynamically adjust the power supply quantities of the users in the different power utilization categories and areas identified in the step S2 in different time periods, and calculating the daily power utilization price according to the electricity price information of the different areas. According to the invention, after user category classification is accurately carried out, future electrical load prediction of different types of users is accurately carried out by using an intelligent technology, so that power distribution and power supply scheduling is reasonably arranged.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid power supply control, and specifically relates to an omni-channel power grid marketing management method and system based on a data center. Background Art

[0002] Smart distribution network is a key link in the construction of smart grid. It is directly connected to users and has the most direct impact on users. Its healthy development plays an extremely important role in users' use of high-quality, high-efficiency, high-reliability and high-safety power supply. At present, more and more intelligent technologies are emerging and applied, gradually forming the intelligent management of distribution network, and the current research pays more attention to the use of visualization technology to achieve the purpose of unattended operation. At the same time, it can also monitor the operation status of the system more clearly and intuitively to ensure the safe, reliable and high-quality operation of the system. The power grid distribution supplies power to different types of electricity users and conducts billing statistics for power marketing. The power marketing data sharing application platform is positioned to provide data analysis and shared application services for different fields and units. It is a strategic business architecture with a data service system integrating data planning, data acquisition, data governance, data analysis, data application and value realization as its content, with the goal of maximizing the value of data assets. Through data services, it meets the needs of data sharing, value mining, analysis application and integration across different levels and across different disciplines.

[0003] In the prior art, the Chinese patent application document with publication number CN114881254A discloses an autonomous service system for power marketing, whose supply and distribution network modules include central power supply network, hierarchical distribution, multi-type integration and regional distribution, and the user center module includes user login, identity authentication, power business application, payment business, power products, warranty and audit. However, it does not specify how to perform identity authentication and multi-type integration and regional distribution according to user login information to dynamically adjust the power supply and distribution strategy within 24 hours. In addition, the Chinese invention patent with publication number CN113327047B discloses a power marketing service channel decision-making method and system based on a fuzzy comprehensive model. Its technical solution constructs a comprehensive index volume of marketing service channels, assigns different indicators, and then assigns objective weights to obtain a fusion weight vector ω. The evaluation matrix is ​​obtained by calculating the score by the weighted average method, and the total evaluation matrix is ​​calculated with the fusion weight vector ω. Finally, the total evaluation matrix is ​​used to calculate the final evaluation grade and determine the service channel with the best grade. This technical solution is a decision-making method on how to obtain the best service channel, rather than a technical solution for omni-channel marketing management that extracts keywords from customer registration information for identity authentication and classification, and then further analyzes the power load curve of this type of user within 24 hours to make dynamic distribution strategy decisions. It cannot solve the technical problem of dynamic distribution of power supply to different types of users of the smart distribution network in omni-channel applications, resulting in low distribution efficiency. Summary of the invention

[0004] In view of the above-mentioned defects, the present invention provides an omni-channel power grid marketing management method and system based on a data hub. The present invention can accurately classify users, use intelligent technology to accurately predict the future power load of different types of users, and then reasonably arrange the power distribution and power supply scheduling plan, which can not only ensure the safety and reliability of users' power consumption, but also realize the economic operation of the distribution network, save losses and costs, and maximize the benefits.

[0005] The present invention provides the following technical solution: an omni-channel power grid marketing management method based on a data center, wherein the method uses an LSTM long short-term memory method to traverse the collected user registration information to identify the registration information, and comprises the following steps:

[0006] S1: Obtain user registration information of the data center cloud database, power consumption data information of distribution network users, and electricity price information of different regions;

[0007] S2: extracting keywords from the registration information extracted in step S1 to identify the electricity usage category of the user;

[0008] S3: Extract the power load amplitude of different types of users according to the power consumption data information of the distribution network users obtained in step S1; extract the AC load periodic wave signal used by different types of users, and use Hilport transform to obtain its analysis signal s(t) and its instantaneous phase θ u (t);

[0009] S4: Use the method in step S3 to sequentially obtain the amplitude and phase of the current signal and voltage signal of different categories of users, calculate the active power of different categories of users, identify the peak, valley and flat sections of the active power of AC load used by different categories of users, control the distribution network to dynamically adjust the power supply to users in different electricity consumption categories and regions identified in step S2 at different time periods, and calculate the daily electricity price based on the electricity price information of different regions.

[0010] Furthermore, the step S2 comprises the following steps:

[0011] S21: Move the window to segment the collected user registration information, extract the words that make up the registration information, and form a registration information word set W = {w1, w2, ..., w i …,w M},w i is the i-th registration information word extracted from the registration information, i=1,2,…,M; M is the total number of registration information words extracted;

[0012] S22: Calculate the probability p(w) that the i-th registration information word in step S21 appears in the registration information word set i ):

[0013]

[0014] Among them, w j For w i-L tow i+L The jth registration information word between; j∈[iL,i+L];

[0015] S23: Perform one-hot encoding on the i-th registration information word in the registration information word set extracted in step S21 i , and then construct the context word one-hot encoding set T of the i-th registration information word i :T i ={t i-L ,t i-L+1 ,…,t i-1 ,0,t i+1 ,…,t i+L-1 ,t i+L};

[0016] S24: Input a test word matrix V composed of arbitrary values ​​to perform context recognition on the i-th registration information word in the registration information word set W of step S21, and compare the i-th column in the test word matrix V with the context word one-hot encoding set T of the i-th registration information word constructed in step S23. i Perform product calculations;

[0017] The test word matrix V is an L×M matrix. Among them, v lm For the l×mth test word vector, l=1,2,…L; m=1,2,…,M;

[0018] The calculated value of the center word of the registration information word set W is

[0019] in, is the estimated value of the kth central word of the i-th registration information word, k∈[1,2L];

[0020] S25: Calculate the importance distribution coefficient g of the i-th registration information word i , the importance distribution coefficient of the central word of the i-th registration information word e i :

[0021]

[0022] Among them, softmax is the activation function; ψ(w i ,p(w i )) is based on w i and p(w i ) is a quasi-periodic function of the independent variable;

[0023] Estimated value set for the central word The average value of the estimated values ​​of the 2L central words in total,

[0024] S26: Further constructing a registration information extraction keyword selection threshold, R = g i e i , when 0.86<R≤1, select the i-th registration information word w i As the keyword for information extraction, the kth central word is used as the keyword to extract the central word, and the LSTM long short-term memory method is used to traverse the user registration information collected in the S1 step to perform registration information recognition.

[0025] Furthermore, the window length L for segmentation in step S21 is:

[0026]

[0027] Among them, b q is the total number of characters in the registration information of the qth user, q = 1, 2, ... Q; Q is the total number of users; B is the total number of characters in the registration information of Q users, Q = {b1, b2, ... b q ,…,b Q}; max(B) is the total number of characters in the user registration information with the longest total number of characters in set B, and min(B) is the total number of characters in the user registration information with the shortest total number of characters in set B.

[0028] Furthermore, the importance distribution coefficient e of the central word of the i-th registration information word is i middle

[0029] Furthermore, in the step S25, the quasi-periodic function θ(w i ,p(w i )) is calculated as follows:

[0030]

[0031] Where n is the probability p(w i ) The i-th registration information word w extracted from the registration information i discrete moments.

[0032] Furthermore, the user categories identified in step S2 are rural users, industrial raw tea users, and urban users; rural users are further divided into agricultural planting users, rural resident users, and agricultural breeding users; urban users are further divided into urban resident users and urban office users.

[0033] Furthermore, the S3 step includes the following steps:

[0034] S31: Extract the AC load periodic wave signal f used by the u-th category of users u (t):

[0035]

[0036] Where, d = 0, 1, 2, ..., D; d is the dth period of the AC load cycle wave used by the uth category user, A u is the positive period amplitude of the load cycle wave of the u-th category user, -A u is the negative period amplitude of the load period wave of the u-th category user, a u is the positive cycle width of the load cycle wave of the u-th category user, T 1u is a cycle of the load cycle wave of the u-th category user;

[0037] S32: The AC load periodic wave signal f of the u-th category of users extracted in step S31 is used u (t) Unit rectangular pulse wave Form of expression:

[0038]

[0039] in, The positive period of the load cycle wave of the u-th category user [dT 1u ,a u +dT 1u ] in the rectangular pulse wave, is the negative period of the load cycle wave of the u-th category user [a u +dT 1u ,(1+d)T 1u ] rectangular pulse wave; unit rectangular pulse wave The following equation is satisfied:

[0040] t0、t end They are the starting time and ending time of the unit rectangular pulse wave respectively;

[0041] S33: The u-th category of users in the expression form of step S32 uses the AC load periodic wave signal f u (t) Perform Hilbert transform H[f u (t)]:

[0042]

[0043] S34: According to the results of step S32 and step S33, the cycle wave analysis signal s(t) of the AC load used by the u-th category user is calculated:

[0044] s u (t) = f u (t)+γH[f u (t)]=A u exp(γθ u (t)); where γ is a complex number, γ 2 =-1;θ u (t) is the AC load cycle wave analysis signal s used by the u-th category user u The instantaneous phase of (t);

[0045] Solve to get the cycle wave analysis signal s of the AC load used by the u-th category user u The instantaneous phase θ of (t) u (t):

[0046]

[0047] Furthermore, the S4 step includes the following steps:

[0048] S41: Repeat the steps S31-S34 to obtain the amplitude of the periodic wave current signal of the AC load used by the u-th category of users Current signal instantaneous phase Further calculate the cycle wave current of AC load used by the u-th category user ω is the grid transmission angular frequency;

[0049] S42: Repeat the steps S31-S34 to obtain the voltage signal amplitude of the periodic wave of the AC load used by the u-th category of users. Current signal instantaneous phase Further calculate the cycle wave voltage of AC load used by the u-th category user

[0050] S43: The instantaneous phase of the periodic wave current signal of the AC load used by the u-th category of users obtained according to the steps S41 and S42 The instantaneous phase of the voltage signal Calculate the power factor of the AC load used by the u-th category of users

[0051] S44: According to the result of step S43, further calculate the active power P of the AC load used by the u-th category of users u :

[0052] S45: respectively construct and solve the first maximum value equation of the active power of the AC load used by the u-th category of users: argmax1 t P u and the second maximum equation: argmax2 t P u , respectively solve to obtain the highest peak value P of the active power of the AC load used by the u-th category of users within 24 hours u,max1 The first peak moment t max1 and the second highest peak value P of the AC load active power used by the uth category of users within 24 hours u,max2 The second peak moment t max2 ; The above equations are solved under the following constraints:

[0053] 0 <DT 1u ≤24h; P is the total power supplied by the distribution network; N is the total number of user categories;

[0054] Construct the valley value equation between two peak values ​​of the active power of the AC load used by the u-th category user: argmin1 t P u , the limiting condition is min{t max1 ,t max2} <t<max{t max1 ,t max2}, min is to solve the minimum time function between two moments, max is to solve the maximum time function between two moments; solve to obtain the valley value P between the two peaks of the active power of the AC load used by the u-th category user within 24 hours u,min1 The peak-to-valley time t min ;

[0055] S46: According to the solution result of step S45, determine two peak load power supply periods, a valley power supply period and an average power supply period when the u-th category of users use the active power of the AC electric load.

[0056] Further, the step S46 determines that the two peak load power supply periods of the u-th category user using the AC load active power are the first peak power supply load period The second peak power supply load section

[0057] The off-peak power supply period is

[0058] The average power supply period is

[0059] The present invention also provides an omni-channel power grid marketing management system based on a data hub using the above method, the system comprising a cloud data acquisition module, a power category identification module, a power load analysis signal and instantaneous phase calculation module, and a distribution network power supply dynamic adjustment control module;

[0060] The cloud data acquisition module is used to obtain user registration information of the data center cloud database, power consumption data information of distribution network users, and electricity price information of different regions;

[0061] The electricity usage category identification module is used to extract keywords from the registration information acquired by the cloud data acquisition module to identify which electricity usage category the user belongs to;

[0062] The power load analysis signal and instantaneous phase calculation module is used to extract the power load amplitude of different categories of users based on the power consumption data information of distribution network users obtained by the cloud data acquisition module; extract the AC load periodic wave signal used by different categories of users, and use Hilport transform to obtain its analysis signal s(t) and its instantaneous phase θ u (t);

[0063] The distribution network power supply dynamic adjustment control module is used to use the power load analysis signal and instantaneous phase calculation module to sequentially obtain the amplitude and phase of the current signal and voltage signal of different categories of users, calculate the active power of different categories of users, identify the peak, valley and flat sections of the active power of the AC load used by different categories of users, control the distribution network to dynamically adjust the power supply to users of different power consumption categories and regions identified in step S2 in different time periods, and calculate the daily electricity price based on the electricity price information of different regions.

[0064] The beneficial effects of the present invention are:

[0065] 1. In the process of using LSTM to traverse the collected user registration information, the application needs to extract keywords, and dynamically and specifically select the segmentation window length L according to the total number of characters in the registration information of different users after segmentation, so that the length of the small part of the sentence after segmentation is conducive to accurate keyword selection. The central word is identified according to the context words in the small part of the sentence after segmentation, and it is determined whether the central word is a useful keyword that can accurately identify the important information in the small part of the sentence. By respectively determining the probability p(w i ) reflects the probability of occurrence of the keyword, and the estimated value of the central word The average value of each element and the central word estimate The deviation of the i-th registration information word w reflects i The importance of the kth central word is then used to select the key word for translation, and then the influence of the key word as the central word on the accuracy of the semantic recognition translation result is comprehensively considered. i , the importance distribution coefficient of the central word of the i-th registration information word e i Then, the threshold limit selection is performed, which effectively improves the accuracy of LSTM long-term and short-term semantic recognition keyword selection, thereby improving the accuracy of subsequent user category recognition.

[0066] 2. The present invention analyzes the single-cycle signal through a rectangular pulse wave expression to obtain its phase and amplitude, thereby eliminating repeated feature information, amplifying the microscopic differences between different loads, and effectively extracting irregular mutation behaviors in the load signal.

[0067] 3. The method and system provided by the present invention convert real signals into complex plane signals, extract the amplitude and phase of current and voltage, and then obtain the power factor of the distribution network during the power consumption of the u-th user, and then effectively calculate the active power actually used by the user load, and then further solve to obtain two peak moments and valley moments, and further calculate the peak power load segment, valley power consumption time period and average power consumption time period of the u-th user, combine the currently collected load data and historical load data for analysis, and derive the load variation law, use intelligent technology to accurately predict the future power load of different types of users, and then reasonably arrange the distribution and power supply scheduling plan, which can not only ensure the safety and reliability of users' power consumption, but also realize the economic operation of the distribution network, save losses and costs, and maximize the efficiency utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The present invention will be described in more detail below based on embodiments and with reference to the accompanying drawings, wherein:

[0069] Figure 1 A flow chart of the omni-channel power grid marketing management method based on the data hub provided by the present invention;

[0070] Figure 2 A graph showing a change in the threshold value for selecting keywords extracted from registration information according to comparative examples 1 to 3 of the method provided by the present invention;

[0071] Figure 3 The instantaneous phase θ of the AC load cycle wave analysis signal extracted by the invention is u (t) The corresponding curve change diagram of the unit rectangular pulse wave;

[0072] Figure 4 The present invention provides an omni-channel power grid marketing management system based on a data center. DETAILED DESCRIPTION

[0073] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] like Figure 1 As shown, it is a flow chart of the omni-channel power grid marketing management method based on the data center provided by the present invention. The method provided by the present invention uses the LSTM long short-term memory method to traverse the collected user registration information to identify the registration information, including the following steps:

[0075] S1: Obtain user registration information of the data center cloud database, power consumption data information of distribution network users, and electricity price information of different regions;

[0076] S2: extract keywords from the registration information extracted in step S1 to identify the electricity usage category of the user;

[0077] S3: According to the power consumption data information of distribution network users obtained in step S1, the power load amplitude of different categories of users is extracted; the AC load periodic wave signal used by different categories of users is extracted, and the analysis signal s(t) and its instantaneous phase θ are obtained by Hilport transform u (t);

[0078] S4: Use the method in step S3 to obtain the amplitude and phase of the current signal and voltage signal of different categories of users in turn, calculate the active power of different categories of users, identify the peak, valley and flat sections of the active power of AC load used by different categories of users, control the distribution network to dynamically adjust the power supply to users in different electricity consumption categories and regions identified in step S2 at different time periods, and calculate the daily electricity price based on the electricity price information of different regions.

[0079] The LSTM long short-term memory method segments the source language sentence and converts it into a word vector sequence, then uses an encoder to process the source language word vector sequence, performs information processing, elimination and activation of the front-to-back RNN ​​neural unit, and synchronizes the information processing, elimination and activation of the right-back-to-front RNN neural unit, outputs a context vector, and uses the decoder to initialize the decoding of the context vector as the initial hidden state of the decoder LSTM, and further generates a word sequence of the target language until the translation of the source language sentence to the target language is completed. However, after the source language sentence is segmented and converted into a word vector sequence, when each RNN neural unit performs information processing, elimination and activation to output the context vector, due to the inaccurate source language segmentation, it is impossible to accurately identify whether the word vector obtained by the segmentation of the output context is the accurate keyword required for the accurate context semantic output. Therefore, it is necessary to use the LSTM long short-term memory method to convert and translate the user registration information and extract key user information, and select keywords that can help the LSTM long short-term memory method output the user registration information accurately. Therefore, as a preferred embodiment of the present invention, step S2 includes the following steps:

[0080] S21: Move the window to segment the collected user registration information, extract the words that make up the registration information, and form a registration information word set W = {w1, w2, ..., w i …,w M},w iis the i-th registration information word extracted from the registration information, i=1,2,…,M; M is the total number of registration information words extracted;

[0081] The window length L for segmentation is:

[0082]

[0083] Among them, b q is the total number of characters in the registration information of the qth user, q = 1, 2, ... Q; Q is the total number of users; B is the total number of characters in the registration information of Q users, Q = {b1, b2, ... b q ,…,b Q}; max(B) is the total number of characters in the user registration information with the longest total number of characters in set B, and min(B) is the total number of characters in the user registration information with the shortest total number of characters in set B;

[0084] S22: Calculate the probability p(w i ):

[0085]

[0086] Among them, w j For w i-L tow i+L The jth registration information word between; j∈[iL,i+L];

[0087] S23: Perform unique hot encoding on the i-th registration information word in the registration information word set extracted in step S21 i , and then construct the context word one-hot encoding set T of the i-th registration information word i :T i ={t i-L ,t i-L+1 ,…,t i-1 ,0,t i+1 ,…,t i+L-1 ,t i+L}; t i-L is the single-hot encoding of the first word in the context of the i-th registration information word, t i-L+1 is the unique hot encoding of the next-to-first context word of the i-th registration information word, t i-1 is the one-hot encoding of the most adjacent previous word of the i-th registration information word, t i+1 is the one-hot encoding of the nearest context word of the i-th registration information word, t i+L-1 is the single-hot encoding of the next-to-last word in the i-th registration information word, t i+L One-hot encoding of the last word in the context of the i-th registration information word;

[0088] S24: Input a test word matrix V composed of arbitrary values ​​to perform context recognition on the i-th registration information word in the registration information word set W of step S21, and compare the i-th column in the test word matrix V with the context word one-hot encoding set T of the i-th registration information word constructed in step S23. i Perform product calculations;

[0089] Among them, the test word matrix V is an L×M-order matrix,

[0090] Among them, v lm is the l×mth test word vector, which is the lth row and mth column element of the test word matrix V, l=1,2,…L; m=1,2,…,M;

[0091] The calculated value of the center word of the registration information word set W is

[0092] in, is the estimated value of the kth central word of the i-th registration information word, k∈[1,2L];

[0093]

[0094] S25: Calculate the importance distribution coefficient g of the i-th registration information word i , the importance distribution coefficient of the central word of the i-th registration information word e i :

[0095] When p(w i ) is larger, the importance distribution coefficient g of the i-th registration information word i The smaller the value, the more likely the i-th registration information word w i The less important it is;

[0096]

[0097] Among them, softmax is the activation function; ψ(w i ,p(w i )) is based on w i and p(w i ) is a quasi-periodic function of the independent variable;

[0098]

[0099] Where n is the probability p(w i ) The i-th registration information word w extracted from the registration information i discrete moments.

[0100] Estimated value set for the central word The average value of the estimated values ​​of the 2L central words in total, The average of the estimated value of the central word The smaller the deviation, the i The smaller the value of is, the more likely the i-th registration information word w i The kth central word has a smaller importance, therefore, the i-th registration information word w i The kth central word cannot be selected as the central word;

[0101] S26: Further constructing a registration information extraction keyword selection threshold, R = g i e i , the smaller R is, the smaller w i The less likely it is to be selected as the central word, when 0.86<R≤1, select the i-th registration information word w i As the keyword for information extraction, the kth central word is used as the keyword to extract the central word, and the LSTM long short-term memory method is used to traverse the user registration information collected in step S1 to perform registration information recognition.

[0102] Comparative Example 1

[0103] The steps of this comparative example are substantially the same as those of the above embodiment, except that the importance distribution coefficient g of the i-th registration information word is not calculated in step S25. i , the registration information extraction keyword selection threshold R=e in step S26 i .

[0104] Comparative Example 2

[0105] The steps of this comparative example are substantially the same as those of the above embodiment, except that the importance distribution coefficient e of the central word of the i-th registration information word is not calculated in step S25. i , the registration information extraction keyword selection threshold R=g in step S26 i .

[0106] Comparative Example 3

[0107] The steps of this comparative example are substantially the same as those of the above embodiment, except that the length of the window for segmentation in step S21 is

[0108] The change diagram of keyword selection threshold for registration information extraction of comparative examples 1-3 is as follows Figure 2 As shown, Figure 2As shown, in Example 3, the selection of the segmentation window length does not dynamically adapt to the normalization of the maximum and minimum values ​​in the total character number set B of the registration information of multiple users, which leads to a displacement deviation in the entire threshold change curve, which leads to deviation in the screening of keywords under the condition of this deviation threshold range limit; although there is no displacement deviation in Example 2 and Example 1, the change slope is relatively gentle, and more keywords are allowed to be not eliminated within the limited range [0.86,1]. Therefore, while the efficiency of keyword screening is too slow, the screening accuracy rate will also decrease due to too many keywords being left. In comparison, in the process of keyword extraction, the present application needs to segment different positions, and dynamically and specifically select the segmentation window length L according to the total number of characters in the registration information of different users after segmentation, so that the length of a small part of the sentence after segmentation is conducive to accurate keyword selection, thereby making the LSTM context output process more accurate and avoiding user category classification errors. According to the context words in the small part of the sentence after segmentation, the central word is identified, and it is judged whether the central word is a useful keyword that can accurately identify the important information in the small part of the sentence. The probability p(w i ) reflects the probability of occurrence of the keyword, and the estimated value of the central word The average value of each element and the central word estimate The deviation of the i-th registration information word w reflects i The importance of the kth central word is then used to select the key word for translation, and then the influence of the key word as the central word on the accuracy of the semantic recognition translation result is comprehensively considered. i , the importance distribution coefficient of the central word of the i-th registration information word e i Then, the threshold limit selection is performed, which effectively improves the accuracy of LSTM long-term and short-term semantic recognition keyword selection, thereby improving the accuracy of subsequent user category recognition.

[0109] Furthermore, according to the method of steps S21-S6, the collected user registration information is traversed using the LSTM long short-term memory method, and the identified user categories are rural users, industrial raw tea users, and urban users; rural users are further divided into agricultural planting users, rural resident users, and agricultural breeding users; urban users are further divided into urban resident users and urban office users.

[0110] In order to obtain accurate omni-channel power grid data information, it is necessary to accurately determine the peak, valley and flat start and end times of the active power effectively utilized by users at the power load end in the total power output of the distribution network within 0-24 hours, that is, the active power load curve, and then effectively determine the peak power consumption period and valley power consumption period of different categories of users, and then dynamically adjust the power distribution of the distribution network in different periods in the area. Step S3 includes the following steps:

[0111] S31: Extract the AC load periodic wave signal f used by the u-th category of users u (t):

[0112]

[0113] Where, d = 0, 1, 2, ..., D; d is the dth period of the AC load cycle wave used by the uth category user, A u is the positive period amplitude of the load cycle wave of the u-th category user, -A u is the negative period amplitude of the load period wave of the u-th category user, a u is the positive cycle width of the load cycle wave of the u-th category user, T 1u is a cycle of the load cycle wave of the u-th category user;

[0114] S32: The AC load periodic wave signal f of the u-th category user extracted in step S31 is used u (t) Unit rectangular pulse wave Form of expression:

[0115]

[0116] in, The positive period of the load cycle wave of the u-th category user [dT 1u ,a u +dT 1u ] in the rectangular pulse wave, is the negative period of the load cycle wave of the u-th category user [a u +dT 1u ,(1+d)T 1u ] rectangular pulse wave; unit rectangular pulse wave The following equation is satisfied:

[0117] t0、t end They are the starting time and ending time of the unit rectangular pulse wave respectively;

[0118] S33: For the u-th category of users expressed in the form of step S32, the AC load periodic wave signal f is used u(t) Perform Hilbert transform H[f u (t)]:

[0119]

[0120] S34: Calculate the cycle wave analysis signal s of the AC load used by the u-th category user according to the results of step S32 and step S33 u (t):

[0121] s u (t) = f u (t)+γH[f u (t)]=A u exp(γθ u (t)); where γ is a complex number, γ 2 =-1;θ u (t) is the AC load cycle wave analysis signal s used by the u-th category user u The instantaneous phase of (t);

[0122] According to s u (t) = f u (t)+γH[f u (t)], and the AC load cycle wave signal f used by the u-th category user in step S32 u (t) and the Hilbert transform H[f u (t)] can get s u (t):

[0123]

[0124] ; Then according to s u (t) = A u exp(γθ u (t)), and s obtained from the above equation u The expression of (t) is solved to obtain the periodic wave analysis signal s of the AC load used by the u-th category user u The instantaneous phase θ of (t) u (t):

[0125]

[0126] like Figure 3 As shown in FIG. 1 , the instantaneous phase θ of the user's AC load periodic wave analysis signal extracted by the present invention is u (t) Corresponding curve change diagram of unit rectangular pulse wave.

[0127] Longer signal input means more macroscopic phase space recursion, and microscopic differences are easily ignored. Since the load signal is a standard periodic signal, the corresponding rectangular pulse wave will also show obvious periodicity. The vertical distance between the positive cycle start time and the negative cycle end time of the rectangular pulse wave represents the length of a rectangular pulse wave cycle of the load signal. A single-cycle rectangular pulse wave can show more subtle difference information. Therefore, the present invention constructs the AC load periodic wave signal f of the u-th category user by adopting step S31 u (t), and expressed as a single rectangular pulse wave In the form of Hilbert transform, the one-dimensional real signal is converted into the analytical signal s on the two-dimensional complex plane. u (t), the complex modulus A of the analytical signal on the complex plane u The amplitude and phase angle represent the amplitude and phase of the signal respectively. Therefore, the analysis signal can be expressed as a complex exponential signal. u (t) = A u exp(γθ u (t)), we can get the cycle wave analysis signal s of the AC load used by the u-th category user u The instantaneous phase θ of (t) u (t). By analyzing the single-cycle signal through rectangular pulse wave expression and obtaining its phase and amplitude, the repeated information of the feature can be eliminated, the microscopic differences of different loads can be amplified, and the irregular mutation behavior in the load signal can be effectively extracted.

[0128] The daily load curve is a graph that describes the load changes of the power system or power grid within a day. It uses time as the horizontal axis and load size as the vertical axis, reflecting the changes in electricity demand in different time periods. The shape of the daily load curve is affected by many factors, including industrial production activities, commercial electricity consumption, residential electricity consumption habits, seasonal changes, etc. There may be differences in daily load curves in different regions and seasons. Therefore, it is necessary to accurately locate the peak load power consumption time domain, valley load power consumption area and average load power consumption area of ​​the load curves of different types of electricity users in time. Therefore, as another preferred embodiment of the present invention, step S4 includes the following steps:

[0129] S41: Repeat steps S31-S34 to obtain the amplitude of the periodic wave current signal of the AC load used by the u-th category of users Current signal instantaneous phase Further calculate the cycle wave current of AC load used by the u-th category user ω is the grid transmission angular frequency;

[0130] S42: Repeat steps S31-S34 to obtain the voltage signal amplitude of the periodic wave of the AC load used by the u-th category of users Current signal instantaneous phase Further calculate the cycle wave voltage of AC load used by the u-th category user

[0131] S43: The instantaneous phase of the periodic wave current signal of the AC load used by the u-th category user obtained according to step S41 and step S42 The instantaneous phase of the voltage signal Calculate the power factor of the AC load used by the u-th category of users

[0132] S44: According to the result of step S43, further calculate the active power P of the AC load used by the u-th category user u :

[0133] S45: respectively construct and solve the first maximum value equation of the active power of the AC load used by the u-th category of users: argmax1 t P u and the second maximum equation: argmax2 t P u , respectively solve to obtain the highest peak value P of the active power of the AC load used by the u-th category of users within 24 hours u,max1 The first peak moment t max1 and the second highest peak value P of the AC load active power used by the uth category of users within 24 hours u,max2 The second peak moment t max2 ; Obviously, P u,max1 >P u,max2 ; The above equations are solved under the following constraints:

[0134] 0 <DT 1u ≤24h; P is the total power supplied by the distribution network;

[0135] Construct the valley value equation between two peak values ​​of the active power of the AC load used by the u-th category user: argmin1 t P u , the limiting condition is min{t max1 ,t max2} <t<max{t max1 ,t max2}, min is to solve the minimum time function between two moments, max is to solve the maximum time function between two moments; solve to obtain the valley value P between the two peaks of the active power of the AC load used by the u-th category user within 24 hours u,min1 The peak-to-valley time t min ;

[0136] That is, if tmax1 >t max2 , then the limiting condition is t max2 <t<t max1

[0137] S46: According to the solution result of step S45, determine two peak load power supply periods, a valley power supply period and an average power supply period when the u-th category user uses the active power of the AC electric load.

[0138] Step S46 determines that the two peak load power supply periods of the u-th category user using the AC power load active power are the first peak power supply load period Right now

[0139] The second peak power supply load section

[0140] Right now

[0141]

[0142] The off-peak power supply period is

[0143] The average power supply period is

[0144] By accurately locating the first peak power supply load segment, the second peak power supply load segment, the valley power supply period, and the average power supply period, a peak-valley-flat load power consumption curve model is constructed for historical load data. Combined with the currently collected load data and historical load data, the load change law is obtained, and intelligent technology is used to accurately predict the future power load of different types of users, and then the distribution and power supply scheduling plan is reasonably arranged. This can not only ensure the safety and reliability of users' electricity use, but also realize the economic operation of the distribution network, save losses and costs, and maximize the utilization of benefits.

[0145] like Figure 4 As shown, the present invention also provides an omni-channel power grid marketing management system based on a data hub using the above method, the system includes a cloud data acquisition module, a power category identification module, a power load analysis signal and instantaneous phase calculation module and a distribution network power supply dynamic adjustment control module;

[0146] The cloud data acquisition module is used to obtain user registration information of the data center cloud database, power consumption data information of distribution network users, and electricity price information of different regions;

[0147] The electricity category identification module is used to extract keywords from the registration information obtained by the cloud data acquisition module to identify the electricity category of the user;

[0148] The power load analysis signal and instantaneous phase calculation module is used to extract the power load amplitude of different categories of users based on the power consumption data information of distribution network users obtained by the cloud data acquisition module; extract the AC load periodic wave signal used by different categories of users, and use Hilport transform to obtain its analysis signal s(t) and its instantaneous phase θ u (t);

[0149] The distribution network power supply dynamic adjustment control module is used to use the power load analysis signal and the instantaneous phase calculation module to sequentially obtain the amplitude and phase of the current signal and voltage signal of different categories of users, calculate the active power of different categories of users, identify the peak, valley and flat sections of the active power of the AC load used by different categories of users, control the distribution network to dynamically adjust the power supply to users of different power consumption categories and regions identified in step S2 at different time periods, and calculate the daily electricity price based on the electricity price information of different regions.

[0150] The present embodiment also provides a computer storage medium, which stores a computer program, and the method described above can be implemented when the processor executes the computer program. Computer-readable media can be considered to be tangible and non-temporary. Non-limiting examples of non-temporary tangible computer-readable media include non-volatile memory circuits (such as flash memory circuits, erasable programmable read-only memory circuits or mask read-only memory circuits), volatile memory circuits (such as static random access memory circuits or dynamic random access memory circuits), magnetic storage media (such as analog or digital tapes or hard drives) and optical storage media (such as CDs, DVDs or Blu-ray discs), etc. The computer program includes processor executable instructions stored on at least one non-temporary tangible computer-readable medium. The computer program may also include or rely on stored data. The computer program may include a basic input / output system (BIOS) that interacts with the hardware of a special-purpose computer, a device driver that interacts with a specific device of a special-purpose computer, one or more operating systems, user applications, background services, background applications, etc.

[0151] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0152] Although the present invention has been described with reference to preferred embodiments, various modifications may be made thereto and some of the technical processes / components therein may be replaced with equivalents without departing from the scope of the present invention. In particular, the various technical features mentioned in the various embodiments may be combined in any manner as long as there is no structural / methodological conflict. The present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. An omni-channel power grid marketing management method based on a data center, wherein the method uses the LSTM long short-term memory method to traverse the collected user registration information to identify the registration information, and is characterized in that: The following steps are involved: S1: Obtain user registration information of the data center cloud database, power consumption data information of distribution network users, and electricity price information of different regions; S2: extracting keywords from the registration information extracted in step S1 to identify the electricity usage category of the user; S3: Extract the power load amplitude of different types of users according to the power consumption data information of the distribution network users obtained in step S1; extract the AC load periodic wave signal used by different types of users, and use Hilport transform to obtain its analysis signal s(t) and its instantaneous phase θ u (t); S4: Use the method in step S3 to sequentially obtain the amplitude and phase of the current signal and voltage signal of different categories of users, calculate the active power of different categories of users, identify the peak, valley and flat sections of the active power of AC load used by different categories of users, control the distribution network to dynamically adjust the power supply to users in different electricity consumption categories and regions identified in step S2 at different time periods, and calculate the daily electricity price based on the electricity price information of different regions.

2. The omni-channel power grid marketing management method based on data center according to claim 1 is characterized in that: The step S2 comprises the following steps: S21: Move the window to segment the collected user registration information, extract the words that make up the registration information, and form a registration information word set W = {w1, w2, ..., w i ..., w M },w i is the i-th registration information word extracted from the registration information, i=1, 2, ..., M; M is the total number of registration information words extracted; S22: Calculate the probability p(w) that the i-th registration information word in step S21 appears in the registration information word set i ): Among them, w j For w i-L tow i+L The jth registration information word between; j∈[iL,i+L]; S23: Perform one-hot encoding on the i-th registration information word in the registration information word set extracted in step S21 i , and then construct the context word one-hot encoding set T of the i-th registration information word i :T i ={t i-L , t i-L+1 , ..., t i-1 ,0,t i+1 , ..., t i+L-1 , t i+L }; S24: Input a test word matrix V composed of arbitrary values ​​to perform context recognition on the i-th registration information word in the registration information word set W of step S21, and compare the i-th column in the test word matrix V with the context word one-hot encoding set T of the i-th registration information word constructed in step S23. i Perform product calculations; The test word matrix V is an L×M-order matrix, Among them, v lm For the l×mth test word vector, l=1,2,…L; m=1,2,…,M; The calculated value of the center word of the registration information word set W is in, is the estimated value of the kth central word of the i-th registration information word, k∈[1,2L]; S25: Calculate the importance distribution coefficient g of the i-th registration information word i , the importance distribution coefficient of the central word of the i-th registration information word e i : Among them, softmax is the activation function; ψ(w i ,p(w i )) is based on w i and p(w i ) is a quasi-periodic function of the independent variable; Estimated value set for the central word The average value of the estimated values ​​of the 2L central words in total, S26: Further constructing a registration information extraction keyword selection threshold, R = g i e i , when 0.86<R≤1, select the i-th registration information word w i As the keyword for information extraction, the kth central word is used as the keyword to extract the central word, and the LSTM long short-term memory method is used to traverse the user registration information collected in the S1 step to perform registration information recognition.

3. The omni-channel power grid marketing management method based on data hub according to claim 2 is characterized in that: The window length L for segmentation in step S21 is: Among them, b q is the total number of characters in the registration information of the qth user, q = 1, 2, ... Q; Q is the total number of users; B is the total number of characters in the registration information of Q users, Q = {b1, b2, ... b q ,…,b Q }; max(B) is the total number of characters in the user registration information with the longest total number of characters in set B, and min(B) is the total number of characters in the user registration information with the shortest total number of characters in set B.

4. The omni-channel power grid marketing management method based on data center according to claim 2 is characterized in that: The importance distribution coefficient e of the central word of the i-th registration information word i middle 5. The omni-channel power grid marketing management method based on data center according to claim 2 is characterized in that: In step S25, the quasi-periodic function θ(w i ,p(w i )) is calculated as follows: Where n is the probability p(w i ) The i-th registration information word w extracted from the registration information i discrete moments.

6. The omni-channel power grid marketing management method based on data center according to claim 5 is characterized in that: The user categories identified in step S2 are rural users, industrial raw tea users, and urban users; rural users are further divided into agricultural planting users, rural resident users, and agricultural breeding users; urban users are further divided into urban resident users and urban office users.

7. The omni-channel power grid marketing management method based on data center according to claim 1 is characterized in that: The S3 step includes the following steps: S31: Extract the AC load periodic wave signal f used by the u-th category of users u (t): Where, d = 0, 1, 2, ..., D; d is the dth period of the AC load cycle wave used by the uth category user, A u is the positive period amplitude of the load cycle wave of the u-th category user, -A u is the negative period amplitude of the load period wave of the u-th category user, a u is the positive cycle width of the load cycle wave of the u-th category user, T 1u is a cycle of the load cycle wave of the u-th category user; S32: The AC load periodic wave signal f of the u-th category of users extracted in step S31 is used u (t) Unit rectangular pulse wave Form of expression: in, The positive period of the load cycle wave of the u-th category user [dT 1u ,a u +dT 1u ] in the rectangular pulse wave, is the negative period of the load cycle wave of the u-th category user [a u +dT 1u ,(1+d)T 1u ] rectangular pulse wave; unit rectangular pulse wave The following equation is satisfied: t0、t end They are the starting time and ending time of the unit rectangular pulse wave respectively; S33: The u-th category of users in the expression form of step S32 uses the AC load periodic wave signal f u (t) Perform Hilbert transform H[f u (t)]: S34: According to the results of step S32 and step S33, the cycle wave analysis signal s(t) of the AC load used by the u-th category user is calculated: s u (t) = f u (t)+γH[f u (t)]=A u exp(γθ u (t)); where γ is a complex number, γ 2 =-1;θ u (t) is the AC load cycle wave analysis signal s used by the u-th category user u The instantaneous phase of (t); Solve to get the cycle wave analysis signal s of the AC load used by the u-th category user u The instantaneous phase θ of (t) u (t):

8. The omni-channel power grid marketing management method based on data center according to claim 7 is characterized in that: The S4 step includes the following steps: S41: Repeat the steps S31-S34 to obtain the amplitude of the periodic wave current signal of the AC load used by the u-th category of users Current signal instantaneous phase Further calculate the cycle wave current of AC load used by the u-th category user ω is the grid transmission angular frequency; S42: Repeat the steps S31-S34 to obtain the voltage signal amplitude of the periodic wave of the AC load used by the u-th category of users. Current signal instantaneous phase Further calculate the cycle wave voltage of AC load used by the u-th category user S43: The instantaneous phase of the periodic wave current signal of the AC load used by the u-th category of users obtained according to the steps S41 and S42 The instantaneous phase of the voltage signal Calculate the power factor of the AC load used by the u-th category of users S44: According to the result of step S43, further calculate the active power P of the AC load used by the u-th category user u : S45: respectively construct and solve the first maximum value equation of the active power of the AC load used by the u-th category of users: argmax1 t P u and the second maximum equation: argmax2 t P u , respectively solve to obtain the highest peak value P of the active power of the AC load used by the u-th category of users within 24 hours u,max1 The first peak moment t max1 and the second highest peak value P of the AC load active power used by the uth category of users within 24 hours u,max2 The second peak moment t max2 ; The above equations are solved under the following constraints: 0 <DT 1u ≤24h; P is the total power supplied by the distribution network; N is the total number of user categories; Construct the valley value equation between two peak values ​​of the active power of the AC load used by the u-th category user: argmin1 t P u , the limiting condition is min{t max1 ,t max2 } <t<max{t max1 ,t max2 }, min is to solve the minimum time function between two moments, max is to solve the maximum time function between two moments; solve to obtain the valley value P between the two peaks of the active power of the AC load used by the u-th category user within 24 hours u,min1 The peak-to-valley time t min ; S46: According to the solution result of step S45, determine two peak load power supply periods, a valley power supply period and an average power supply period when the u-th category of users use the active power of the AC electric load.

9. The omni-channel power grid marketing management method based on data hub according to claim 8 is characterized in that: The step S46 determines that the two peak load power supply periods of the u-th category user using the AC load active power are the first peak power supply load period The second peak power supply load section The off-peak power supply period is The average power supply period is 10. An omni-channel power grid marketing management system based on a data hub using the method according to any one of claims 1 to 9, characterized in that: The system includes a cloud data acquisition module, an electricity category identification module, an electricity load analysis signal and instantaneous phase calculation module, and a distribution network power supply dynamic adjustment control module; The cloud data acquisition module is used to obtain user registration information of the data center cloud database, power consumption data information of distribution network users, and electricity price information of different regions; The electricity usage category identification module is used to extract keywords from the registration information acquired by the cloud data acquisition module to identify which electricity usage category the user belongs to; The power load analysis signal and instantaneous phase calculation module is used to extract the power load amplitude of different categories of users based on the power consumption data information of distribution network users obtained by the cloud data acquisition module; extract the AC load periodic wave signal used by different categories of users, and use Hilport transform to obtain its analysis signal s(t) and its instantaneous phase θ u (t); The distribution network power supply dynamic adjustment control module is used to use the power load analysis signal and instantaneous phase calculation module to sequentially obtain the amplitude and phase of the current signal and voltage signal of different categories of users, calculate the active power of different categories of users, identify the peak, valley and flat sections of the active power of the AC load used by different categories of users, control the distribution network to dynamically adjust the power supply to users of different power consumption categories and regions identified in step S2 in different time periods, and calculate the daily electricity price based on the electricity price information of different regions.