CN113947444A - Electricity selling package recommending method considering multi-granularity hesitation fuzzy set and incomplete weight - Google Patents
Electricity selling package recommending method considering multi-granularity hesitation fuzzy set and incomplete weight Download PDFInfo
- Publication number
- CN113947444A CN113947444A CN202111359084.4A CN202111359084A CN113947444A CN 113947444 A CN113947444 A CN 113947444A CN 202111359084 A CN202111359084 A CN 202111359084A CN 113947444 A CN113947444 A CN 113947444A
- Authority
- CN
- China
- Prior art keywords
- user
- evaluation
- package
- load
- electricity
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000005611 electricity Effects 0.000 title claims abstract description 188
- 238000000034 method Methods 0.000 title claims abstract description 77
- 238000011156 evaluation Methods 0.000 claims abstract description 164
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 33
- 238000003064 k means clustering Methods 0.000 claims abstract description 27
- 239000011159 matrix material Substances 0.000 claims abstract description 21
- 238000012163 sequencing technique Methods 0.000 claims abstract description 7
- 238000011002 quantification Methods 0.000 claims abstract description 6
- 239000000126 substance Substances 0.000 claims description 36
- 235000019580 granularity Nutrition 0.000 claims description 29
- 239000013598 vector Substances 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 13
- 235000012054 meals Nutrition 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 6
- 239000006185 dispersion Substances 0.000 claims description 4
- 239000002245 particle Substances 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- -1 period Chemical class 0.000 claims description 2
- 238000002474 experimental method Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 10
- 230000006399 behavior Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 4
- 238000009826 distribution Methods 0.000 description 3
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 229910000831 Steel Inorganic materials 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000010959 steel Substances 0.000 description 2
- 239000004753 textile Substances 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002902 bimodal effect Effects 0.000 description 1
- 235000008429 bread Nutrition 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0255—Targeted advertisements based on user history
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/02—Computing arrangements based on specific mathematical models using fuzzy logic
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0282—Rating or review of business operators or products
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Primary Health Care (AREA)
- Human Resources & Organizations (AREA)
- Tourism & Hospitality (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Automation & Control Theory (AREA)
- Biomedical Technology (AREA)
- Fuzzy Systems (AREA)
- Molecular Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
Abstract
The invention discloses an electricity selling package recommending method considering multi-granularity hesitation fuzzy set and incomplete weight, which comprises the steps of establishing a user load information database based on a user electricity utilization characteristic index, and clustering users based on an improved K-means clustering algorithm to determine the category of a target user; considering incomplete weight information and a multi-granularity hesitation fuzzy language evaluation set provided by a user, and constructing an evaluation matrix of the user on the electric power selling package; respectively calculating the satisfaction degrees of the historical user and the target user to the electricity selling package; and obtaining the full-sequencing electricity selling package recommendation and the Top-M electricity selling package recommendation based on the user satisfaction quantification result. The method adopts the multi-granularity hesitation fuzzy language evaluation set to represent the evaluation information of the user, considers the incompleteness of the attribute weight information provided by the user for selling the electric package, not only can reflect the hesitation fuzziness of the user when evaluating the electric package, but also can enable the user to provide the evaluation information more flexibly, and more truly reflect the satisfaction degree of the user.
Description
Technical Field
The invention relates to the technical field of power information, in particular to an electric sales package recommendation method considering multi-granularity hesitation fuzzy set and incomplete weight.
Background
On 15 days 3 and 15 months 2015, the public center and the state department issue a plurality of opinions about further advanced power system reform, which marks that the power selling side is opened, and the single buying-single selling vertical integrated control mode of the traditional power market is converted into a multi-buying-multi-selling customized and marketized mode; in 2019, in 6 months, the national development and reform committee issues 'notice on fully releasing the power generation and utilization plan of the operational power users', the user coverage range of the retail market is further expanded, and the user develops from the original passive power supply receiving mode to the mode of automatically selecting power selling companies. Meanwhile, with the implementation of the concept of high-quality development, for power consumers, the power consumers are required to obtain diversified and convenient value-added services as well as reliability and safety of power supply; for the electricity selling companies, it is also expected to enhance the viscosity of existing users and attract a large number of new users through value-added services such as electricity selling packages, comprehensive energy services, energy-saving services, and the like. The electricity selling package is a necessary means for an electricity selling company to improve the income and obtain market competitiveness as service and operation mode innovation in a competitive electricity selling market environment. The foreign electric power markets of Germany, America, UK and the like provide thousands of electricity selling packages for users, and China also designs various forms of electricity price packages for the users to decide, such as: the method is characterized in that a power package which considers the peak clipping and valley filling effects of user behavior changes and influences on the power package is designed for industrial and commercial users; based on the user reference price decision and the user viscosity, the electricity selling package is optimally designed; and (4) respectively designing customized time-of-use electricity price packages and peak-valley combined power packages by considering the limited selection behaviors of users. However, in the face of a lot of electricity selling packages in the electricity selling market, how to accurately and scientifically recommend electricity packages meeting the needs of users to the users is a key point for improving the satisfaction of the users and increasing the market share of the users. In addition, in the conventional documents [1] and [2], a customized time-of-use electricity price package and a peak-valley combined power package are also designed in consideration of the limited selection behavior of the user.
The existing domestic and foreign recommendations for the electricity sales packages can be roughly divided into a direct recommendation method and an indirect recommendation method. Most of direct recommending methods for electricity selling packages are applied to an online recommending platform, and the basic idea is to compare the cost of each electricity selling package based on the electricity utilization condition of a user and recommend the electricity selling package with the lowest cost to the corresponding user, for example: platform iSelect, Power to Choose, Energy Made Easy, Check24, etc. The online recommendation method is simple and easy to implement, but the electricity selling package recommendation method only based on the cost of the electricity charge of the user ignores the diversity of user evaluation information, such as: green electricity, loyalty. Besides the direct cost-based recommendation method, the hierarchical clustering is carried out on the users based on the electricity utilization diversity characteristic quantity of the users, and stepped electricity price packages and stepped time-of-use electricity price packages are recommended for users with changeable electricity utilization behaviors and users with regular electricity utilization behaviors respectively. Although the recommendation method considers the electricity consumption behavior characteristics and the preference of the user, the hesitation fuzzy characteristics shown when the user evaluates the electricity selling package and the heterogeneity of evaluation information caused by the difference of knowledge and culture background of different users are still ignored, and a large error is brought to the accurate recommendation of the electricity selling package. In addition, considering that the user has limited knowledge of the property of the electricity selling package, the provided weight information has incomplete conditions, such as: the problem that how to process the diversified attribute weight language information and convert the incomplete weight information provided by the user into the determined attribute weight information is necessary to be solved by an electric power selling company for accurately recommending the electric power selling package for the user is that the attribute weight information possibly provided by the user is 'the attribute A is more important than the attribute B' or 'the attribute A is more important than the attribute B by 2 times'.
[1] Xiaobai, Zuianqi, Jianzhuo, etc. customized electricity price package design based on limited user selection behavior [ J ] power grid technology, 2021,45(3): 1050-.
Xiao Bai,Cui Hanqi,Jiang Zhuo,et al.Customized electricity price package design based on limited rational user selection behavior[J].Power System Technology,2021,45(3):1050-1058.
[2] Zhang Zhi, Lufeng, forest Intelligence, etc. consider the peak-valley combined power package design of the power selling company with limited user [ J ] power system automation, 2021,45(16): 114-.
Zhang Zhi,Lu Feng,Lin Zhenzhi,et al.Peak-valley combination electricity package design for electricity retailer considering bounded rationality of consumers[J].Automation of Electric Power Systems,2021,45(16):114-123.
Disclosure of Invention
The technical problem to be solved by the invention is to provide the method for recommending the electricity selling package considering the multi-granularity hesitation fuzzy set and the incomplete weight, so as to provide the recommendation of the electricity selling package for the user based on the user satisfaction quantization result.
In order to solve the technical problem, the invention provides an electricity selling package recommending method considering multi-granularity hesitation fuzzy set and incomplete weight, which comprises the following specific processes:
s1, establishing a user load information database based on the user electricity utilization characteristic indexes, clustering the users based on an improved K-means clustering algorithm, and determining the category of the target user;
step S2, considering the incomplete weight information and the multi-granularity hesitation fuzzy language evaluation set provided by the user, and constructing an evaluation matrix of the user for the electric sales package;
step S3, calculating the satisfaction of the historical user and the target user to the electric power selling package;
and step S4, obtaining the complete-sequence electricity selling package recommendation and the Top-M electricity selling package recommendation based on the user satisfaction quantification result.
The invention is an improvement of the method for recommending the electricity selling package considering the multi-granularity hesitation fuzzy set and the incomplete weight:
the specific process of step S1 is as follows:
step S1.1, building user load information database
Based on user A ═ { A1,A2,…,Ai,…,AIThe daily load data for a month is:wherein the content of the first and second substances,k is the total hours of a month; based on the monthly load rateNumber of hours of maximum utilizationMean value of peak-to-valley rate of working dayMean peak-to-valley rate of off-dayPeak load meanMean average of load rate in flat phaseMean value of load rate at valley periodObtain user AiLoad characteristic index vector ofAnd a user load information database corresponding to the user and the load characteristic index vector thereof is established;
(1) the monthly load rate:
wherein the content of the first and second substances,andare respectively user AiThe average load per month and the maximum load,for user AiDaily load data for the kth hour of the total hours of a month;
(2) number of hours of maximum utilization
Wherein the content of the first and second substances,for user AiElectricity consumption in whole month;
(3) mean value of peak-to-valley rate of working day
Wherein, KWIs the total number of days of the working day,the day maximum load on day i of the working day,the day minimum load for day i of the non-workday;
(4) mean peak-to-valley rate of off-day
Wherein, KNWIs the total number of days on the non-working day,the day maximum load on day j of the non-workday,the daily minimum load on day j of the non-workday;
(5) peak load mean
Wherein peak periods are defined as 08:00-11:00 and 18:00-21:00, T is the total number of days, and T is KW+KNW,For user A in the peak period of the t dayiThe average value of the loads of (a),is the user A on the t dayiThe load mean value of (1);
(6) mean average of load rate in flat phase
Wherein, the ordinary period is defined as 06:00-08:00, 11:00-18:00 and 21:00-22:00,for the tth balance period user AiThe load mean value of (1);
(7) mean load factor at off-peak
Wherein the valley time periods are defined as 22:00-24:00 and 00:00-06:00,for the user A in the trough period of the t dayiThe load mean value of (1);
step S1.2, clustering based on historical user load information database
Based on the user load information database established in the step S1.1, clustering historical users by adopting an improved K-means clustering algorithm, and specifically comprising the following steps:
step S1.2.1, making the clustering number C equal to 2, randomly selecting C users, and marking as C1,C2,…,Cj,…,CcUsing the load characteristic index vector as the clustering center
Step S1.2.3 for user AiAre compared and obtainedAnd will be user AiClassified as j, and the number of users belonging to the j is set as IjThe corresponding user is represented as
Step S1.2.4, calculating the square error J based on the classification result of step S1.2.3 as equation (8)c(t):
Wherein, t is the iteration number,is corresponding to Ai,jClass j user AiA load characteristic index vector of (a);
step S1.2.5, recalculating the cluster centers
Calculating the average distance D between the load characteristic index vectors of all users in each class and the clustering center, selecting the users with the distance to the clustering center being less than 2D in the class, and then recalculating the mean value of the load characteristic indexes of the selected users as a new clustering center;
step S1.2.6, determine whether formula (9) is satisfied:
Jc(t+1)-Jc(t)≥0 (9)
if not, t is t +1, and then the step S1.2.1 is returned to continue the iteration;
if the formula (9) is satisfied, the iteration is ended, and the clustering quality index I is calculated based on the clustering result of the t iterationqc.av:
Wherein, Iqc(i) When the number of clusters is c, the closeness degree of the ith user to the cluster center of the class to which the ith user belongs is represented as follows:
wherein the content of the first and second substances,represents the average distance between classes as user AiAverage value of the distance between the user load characteristic index phasor and other user load characteristic indexes except the category to which the user belongs;representing the average distance within a class, as user AiAverage of the phasor distances from the other user load characteristic indexes in the category to which the user belongs;
step S1.2.7, making the cluster number c ═ c +1, returning and executing step s 1.2.1-step S1.2.6 until the cluster number reaches the maximum value
Step S1.2.8, comparing the clustering quality index I under different clustering numbersqc.av,Iqc.avThe corresponding cluster number at the maximum is the optimal cluster number c;
step S1.3, calculating the similarity between the target user and the historical user based on Pearson correlation coefficient
The target user is a user needing to recommend the electricity selling package and is represented as W ═ W1,W2,…,Wn,…,WMAnd obtaining load characteristic index vectors of each target user based on daily load data of each user in a certain month by combining the formula (1) with the formula (7)Calculating the center of each cluster under the optimal cluster number cThe class corresponding to the minimum distance is the target user WnThe category to which it belongs;
set target user WnBelonging to the p-th class of historical users, wherein the number of the class of users is NpIs shown asThe similarity between the target user and the historical user is described by using the Pearson correlation coefficient:
wherein the content of the first and second substances,represents a target user WnAnd historical usersThe similarity of (2);andrespectively represent users WnAnd historical usersThe d-th load characteristic index of (1),andrespectively represent users WnAnd historical usersAverage value of load characteristic index.
The invention is further improved by taking the multi-granularity hesitation fuzzy set and incomplete weight into consideration as a recommendation method of the power selling package:
the specific process of the step 2 is as follows:
step S2.1, building an evaluation matrix of the user for the electricity sales package based on the hesitation fuzzy language set
The set of electricity selling set is T ═ T1,T2,…Tj,…,TJUser a ═ a }1,A2,…,Ai,…,AIThe evaluation attribute set for the electricity sales package isDescribing the evaluation information of the user on the electricity selling package by adopting hesitation fuzzy language:
let user AiLanguage ofSaid evaluation set isWherein g (A)i) Represents the granularity of the language evaluation set and is odd,represents the evaluation amount of the p-th language, p is 0,1, …, g (A)i) -1 symbolic corner marks for language evaluation quantities; language evaluation setThe elements are arranged in sequence:
user AiSelf-language-based evaluation setAnd evaluating the electricity selling package, wherein an evaluation matrix is as follows:
wherein the content of the first and second substances,represents user AiPair electric sales package TjProperty (2) ofGiven the hesitation fuzzy language evaluation information, J and K are respectively the J-th package and the K-th attribute;
step S2.2, determining the attribute weight of the electricity selling package considering the incomplete weight
Wherein "other" represents user AiThe incomplete weight information is provided in the form of,representation for the kth attributeUser A ofiThe degree of deviation between language evaluations given to each electricity sales package:
wherein the content of the first and second substances,p=0,1,…,g(Ai)-1,for user AiEvaluating the kth attribute in the qth package;
integrating the evaluation information of each user by adopting an Hesitation Fuzzy Language Weighted Average Operator (HFLWAO) to obtain the evaluation matrix of each user on the electric sales package
Wherein the content of the first and second substances,represents user AiPair electric sales package TjThe evaluation results of (1);
the HFLWAO is calculated as follows:
(1) based on electricity selling set meal TjThe K attributes of (a) are set as follows:
akthe evaluation result of the kth attribute;
wherein trunc (·) is an integer function; delta (-) represents a mapping between symbolic logo of language evaluation quantity and hesitation fuzzy language set, mhAs a result of the evaluation of the package by the user,in order to evaluate the degree of progress,to evaluate the degree of membership;
let beta be [0, g (A) ]i)-1]The equivalent information of beta is described by a binary group:
presence of a transformation function Δ-1(. h), converting the binary equivalent information into equivalent values:
Wherein the content of the first and second substances,for a two-tuple of language evaluation,ratio representing linguistic evaluation variables:
step S2.3, unification of multi-granularity hesitation fuzzy language evaluation information of the user on the electric power selling package
Converting language evaluation sets under different granularities used by a user for evaluating the power selling package into language evaluation sets under the same granularity, and determining the granularity as g (A)i) Evaluation setEvaluation information of (2)Conversion to g (A) on the basis of particle sizef) Evaluation setEvaluation information of (2)The process is as follows:
step S2.3.1, let g (×) -LCM (g (a)i)-1,g(Af) -1) +1, where LCM (·) represents the minimum common multiple;
Let thetae=0,e=0,1,…,g(*)-1;
Step S2.3.5 based onCalculating the normalized language evaluation variable ratio, as shown in formula (22):
Based on the steps S2.3.1-S2.3.6, an evaluation matrix of the user for the electricity sales package under uniform granularity is obtainedWherein the content of the first and second substances,representing user A at a uniform granularityiPair electric sales package TjEvaluation information of (2):
is at AfThe language evaluation at the granularity of the language,in order to evaluate the degree of membership newly,the evaluation degree is new.
The invention is further improved by taking the multi-granularity hesitation fuzzy set and incomplete weight into consideration as a recommendation method of the power selling package:
the specific process of the step 3 is as follows:
user A at uniform granularity based on formula (23)iPair electric sales package TjEvaluation information of (2)The expectation of the evaluation information of the user is obtained, and the satisfaction matrix of the user to the electric sales package is obtainedWherein the content of the first and second substances,for user AiPair electric sales package TjSatisfaction of (2):
target user WnPair electric sales package TjSatisfaction degree SCnjComprises the following steps:
the invention is further improved by taking the multi-granularity hesitation fuzzy set and incomplete weight into consideration as a recommendation method of the power selling package:
the full-sequencing electricity selling package recommendation specifically comprises the following steps:
calculating a root mean square error epsilon based on satisfaction results of the target users for each electricity selling package1nThen according to the root mean square error epsilon1nRecommending to the target user according to the order of the electricity selling packages corresponding from small to large for the target user to decide:
wherein J is the total number of the electricity sales packages and OnjAndrespectively represent for the target user WnAnd sell electric set meal TjThe actual sorting result of (a) and the sorting result obtained by the algorithm of the formula (25);
the Top-M electricity selling package recommendation specifically comprises the following steps:
based on the satisfaction result of the target user for each electricity selling package, the electricity selling company only recommends the packages ranked in the top M to the target user, and does not provide the ranking result;
recommend to target user WnThe first M sets of electricity sales packagesSet of packages with actual ranking as top MAccuracy of recommendation e2nComprises the following steps:
is a target user WnThe set of packages actually ranked top M,obtaining a target user W for an algorithmnThe packages with the top M are ranked, and M is the total number of recommended packages.
The invention has the following beneficial effects:
1. the invention provides an electricity selling package recommending method considering multi-granularity hesitation fuzzy sets and incomplete weights, and clusters users by utilizing an improved K-means clustering algorithm, so that the influence of noise points can be effectively reduced, and the clustering result can more effectively reflect the load characteristics of various users;
2. the invention provides an electric sales package recommendation method considering a multi-granularity hesitation fuzzy set and an incomplete weight, which adopts a multi-granularity hesitation fuzzy language evaluation set to represent evaluation information of a user, and considers the incompleteness of attribute weight information of an electric sales package provided by the user, so that the hesitation fuzzy performance of the user in the electric sales package evaluation can be reflected, users with different culture backgrounds can more flexibly and elastically provide evaluation information, the satisfaction degree of the user on the electric sales package can be more truly reflected, and a new thought is provided for an electric sales company to recommend the electric sales package;
3. the invention provides the electricity selling package recommending method considering the multi-granularity hesitation fuzzy set and the incomplete weight, which can ensure higher recommending accuracy and better recommending performance.
Drawings
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.
FIG. 1 is a schematic diagram of a process for building and clustering a user load information database according to the present invention;
FIG. 2 is a schematic diagram of an electricity selling package recommending process considering multi-granularity hesitation fuzzy sets and incomplete weight information according to the present invention;
FIG. 3 is a histogram showing the average daily load frequency distribution of each user in the park of experiment 1;
FIG. 4 is a schematic diagram of the average daily load rate, average daily peak-to-valley rate and language evaluation set granularity of each user in the campus of experiment 1;
FIG. 5 is a schematic diagram of cluster quality index values corresponding to different cluster numbers in experiment 1;
fig. 6 is a schematic diagram of the clustering results of the various types of users and the number of the various types of users in experiment 1;
FIG. 7 is a graph illustrating the satisfaction of the late-peak historical user with selling an electricity package in experiment 1;
FIG. 8 is a graph showing the similarity between the target user and the late peak type historical user in experiment 1;
fig. 9 is a schematic diagram of a result of quantifying satisfaction of the target user W1 for each electricity sales package in experiment 1;
FIG. 10 is a schematic diagram of the clustering centers obtained by the improved K-means clustering algorithm and the original K-means clustering algorithm in experiment 1;
FIG. 11 is a diagram of the optimal cluster number and cluster quality index for different noise ratios in experiment 1;
fig. 12 is a schematic diagram of the evaluation result of the satisfaction of various target users with the electricity sales package in experiment 1;
fig. 13 is a schematic diagram showing comparison of recommended ordering results of the electricity sales packages obtained by different methods in experiment 1.
Detailed Description
The invention will be further described with reference to specific examples, but the scope of the invention is not limited thereto:
1. User load information database building and clustering
1.1 building user load information database
The power utilization characteristics of different types of users have difference, the difference is reflected in the difference of user load characteristic indexes, and the power utilization characteristics are based on that a is equal to { A } A1,A2,…,Ai,…,AIDaily load data of a monthWherein the content of the first and second substances,k is the total hours of a certain month, and the electricity utilization characteristics of each user are described by taking the monthly load rate, the maximum utilization hours, the peak-to-valley difference rate mean value of working days and non-working days and the peak-to-valley period load rate mean value as characteristic indexes. With user AiFor example, the physical meanings and definitions of the characteristic indexes are as follows.
(1) Rate of load of the moon
The monthly load rate is used for reflecting the load change fluctuation condition of the user in the whole month, is the ratio of the average load of the month to the maximum load, and the calculation formula is shown as a formula (1).
Wherein the content of the first and second substances,andare respectively user AiThe average load per month and the maximum load,for user AiDaily load data for the kth hour of the total hours of a month;
(2) Number of hours of maximum utilization
The maximum utilization hours are used for reflecting the time utilization efficiency of the user load, and the calculation formula is shown in formula (2) as the ratio of the total monthly electricity consumption to the maximum monthly load:
wherein the content of the first and second substances,for user AiElectricity consumption in whole month, maximum utilization hoursThe larger; user AiThe more efficient the utilization of the load time of the whole month.
(3) Mean value of peak-to-valley rate of working day
The working day peak-valley difference rate mean value is used for reflecting the stability of the power utilization of the user in working days, is the mean value of the ratio of the difference value between the maximum load and the minimum load in working days and the ratio of the maximum load in working days, and the calculation formula is shown as a formula (3).
Wherein, KWIs the total number of days of the working day,the day maximum load on day i of the working day,the day minimum load for day i of the non-workday;
(4) Mean peak-to-valley rate of off-day
The non-working day peak-valley difference rate mean value is used for reflecting the stability of the non-working day power utilization of the user, is the mean value of the ratio of the maximum load and the minimum load of the non-working day to the maximum load of the non-working day, and has a calculation formula shown in a formula (4):
wherein, KNWIs the total number of days on the non-working day,the day maximum load on day j of the non-workday,the daily minimum load on day j of the non-workday;
mean peak-to-valley rate of off-dayThe larger, user AiThe greater the non-workday load fluctuations.
(5) Peak load mean
The peak period load rate mean value is used for reflecting the variation fluctuation condition of the load of the user in the peak period, the peak period is defined as 08:00-11:00 and 18:00-21:00, and is the mean value of the ratio of the peak period load mean value to the current day load mean value, and the calculation formula is shown as the formula (5):
wherein T is the total number of days, and satisfies that T is KW+KNW,For user A in the peak period of the t dayiThe average value of the loads of (a),is the user A on the t dayiThe load mean value of (1);
peak load meanThe larger, user AiThe more pronounced the fluctuation of the load during the peak period.
(6) Mean average of load rate in flat phase
The average value of the average period load rate is used for reflecting the variation fluctuation condition of the load of the user in the average period, the average period is defined as 06:00-08:00, 11:00-18:00 and 21:00-22:00, and is the average value of the ratio of the average value of the load of the average period to the average value of the load of the current day, and the calculation formula is shown as the formula (6):
wherein the content of the first and second substances,for the tth balance period user AiThe load mean value of (1);
mean average of load rate in flat phaseThe larger, user AiThe more pronounced the fluctuation of the load during the flat period.
(7) Mean load factor at off-peak
The mean value of the valley period load rate is used for reflecting the variation fluctuation condition of the load of the user in the valley period, the valley period is defined as 22:00-24:00 and 00:00-06:00, and is the mean value of the ratio of the mean value of the load of the valley period to the mean value of the load of the current day, and the calculation formula is shown as the formula (7):
wherein the content of the first and second substances,for the user A in the trough period of the t dayiLoad average of (2). Mean load factor at off-peakThe larger, user AiThe more pronounced the fluctuation of the load during the valley period.
Based on the user load characteristic index, the user A can be obtainediLoad characteristic index vector ofAnd a user load information database corresponding to the user and the load characteristic index vector thereof is established, so that support is provided for historical user clustering and calculation of similarity between the target user and the historical user.
1.2 clustering based on historical user load information database
Considering that the number of existing historical users is large, in order to reduce the calculation amount, users with similar load characteristics need to be clustered. Based on the built user load information database, clustering historical users by adopting an improved K-means clustering algorithm, and specifically comprising the following steps:
1) making the clustering number C equal to 2, randomly selecting C users, and marking as C1,C2,…,Cj,…,CcUsing the load characteristic index vector as the clustering center
3) For AiAre compared and obtainedAnd will be user AiClassified as j, and the number of users belonging to the j is set as IjThe corresponding user is represented as
4) Calculating a square error J based on the classification result of the step 3)c(t) as shown in formula (8), wherein t is the number of iterations;
wherein the content of the first and second substances,is corresponding to Ai,jClass j user AiLoad characteristic index vector of
5) Recalculating cluster centers
The original K-means clustering algorithm directly takes the mean value of each class as a new clustering center, and needs to be improved in order to reduce the influence of noise points. Firstly, calculating the average distance D between all user load characteristic index vectors and a cluster center in each class; then, selecting users in the category, wherein the distance between the users and the cluster center is less than 2D; and finally, recalculating the mean value of the load characteristic indexes of the selected users as a new clustering center. The operation can eliminate the influence of noise points, so that the clustering center can better reflect the characteristic of normal user load;
6) judging whether formula (9) is satisfied:
Jc(t+1)-Jc(t)≥0 (9)
if not, t is t +1, and then the step 1) is returned to continue the iteration;
if the formula (9) is satisfied, the iteration is ended, and the clustering quality index I is calculated based on the clustering result of the t iterationqc.avAs shown in formula (10);
wherein, Iqc(i) When the number of clusters is c, the closeness degree of the ith user to the cluster center of the class to which the ith user belongs depends on the intra-class distance and the inter-class distance, as shown in formula (11).
Wherein the content of the first and second substances,represents the average distance between classes as user AiAverage value of the distance between the user load characteristic index phasor and other user load characteristic indexes except the category to which the user belongs;representing the average distance within a class, as user AiAverage of the phasor distances from other user load characteristic indicators within the category to which they belong. Therefore, the larger the average distance between classes is, the smaller the average distance in the classes is, and the better the clustering quality is;
7) making the clustering number c equal to c +1, returning to the step 1), executing the steps 1) to 6) until the clustering number reaches the maximum value
8) Comparing clustering quality index I under different clustering numbersqc.av,Iqc.avMaximum time pairThe number of clusters is the optimal number of clusters c.
1.3 Pearson correlation coefficient-based calculation of similarity between target user and historical user
The target user is a user needing to recommend the electricity selling package and is represented as W ═ W1,W2,…,Wn,…,WMAnd (6) based on daily load data of each user in a certain month, combining the formula (1) and the formula (7), obtaining load characteristic index vectors of each target userCalculating the center of each cluster under the optimal cluster number cThe class corresponding to the minimum distance is the target user WnThe category to which it belongs.
Set target user WnBelonging to the p-th class of historical users, wherein the number of the class of users is NpIs shown asConsidering that the Pearson correlation coefficient can effectively measure the correlation and the closeness degree between the load characteristics of each user, the similarity between the target user and the historical user is described by using the coefficient, as shown in formula (12).
Wherein the content of the first and second substances,represents a target user WnAnd historical usersThe greater the similarity, the more similar the load characteristics of the two users are, and the more likely it is to select the same electricity sales package.Andrespectively represent users WnAnd historical usersThe d-th load characteristic index of (1),andrespectively represent users WnAnd historical usersAverage value of load characteristic index.
2 evaluation of selling electric combo by user considering multi-granularity hesitation fuzzy set and incomplete weight
2.1 construction of evaluation matrix of user for selling electric package based on hesitation fuzzy language set
Setting the set of electricity selling packages as T ═ T1,T2,…Tj,…,TJUser a ═ a }1,A2,…,Ai,…,AIThe evaluation attribute set for the electricity sales package isConsidering that quantitative information is difficult to give when a user evaluates the property of the electric power selling package and certain hesitation ambiguity is shown, a hesitation ambiguity language is adopted to describe the evaluation information of the user on the electric power selling package.
Let user AiIs a language evaluation set ofWherein g (A)i) Represents the granularity of the language evaluation set and is odd,represents the evaluation amount of the p-th language, p is 0,1, …, g (A)i) -1 symbolic corner mark for language evaluation quantity. Language evaluation setThe elements are arranged in sequence, namely: if p is>q, then:
user AiSelf-language-based evaluation setThe electricity sales package is evaluated according to the evaluation matrix ofWherein the content of the first and second substances,represents user AiPair electric sales package TjProperty (2) ofAnd giving hesitation fuzzy language evaluation information. It is to be noted that it is preferable that,may comprise a plurality ofThe language evaluation amount in (1), J and K are respectively the J-th package and the K-th attribute.
2.2 Electricity sales package Attribute weight determination with consideration of incomplete weight
The electricity selling package attribute weight reflects the importance degree of the attribute in the electricity selling package evaluation. Given that the user has limited knowledge of the importance of the properties of the electricity sales package, and that the weight information that may be provided is incomplete, such as: user AiThe weight information provided is: attribute 1 is more important than Attribute 2 Attribute 1 weight not exceedingThe difference between the importance of the attribute 3 and the importance of the attribute 4 is not less thanThe information is called incomplete weight information, and in order to ensure that the obtained weights can comprehensively reflect the importance degree of each attribute, the weight of each attribute is determined by adopting a dispersion maximization methodThe model is shown in formula (13).
Wherein "other" represents user AiIncomplete weight information is provided.Representation for the kth attributeUser A ofiThe degree of deviation between the language evaluations given to each electricity sales package is shown in equation (14).
Wherein the content of the first and second substances,p=0,1,…,g(Ai)-1,for user AiEvaluation of the kth Attribute for the qth package
Solution formula(13) The single-target optimization model can obtain the weight of each attribute of the electricity selling packageIn order to ensure the integrity and the interpretability of user evaluation information, a Hesitation Fuzzy Language Weighted Average Operator (HFLWAO) is adopted to integrate the evaluation information of each user, and an evaluation matrix of each user for selling the electric package is obtainedWherein the content of the first and second substances,represents user AiPair electric sales package TjThe method of determining (2) is represented by the following formula (15):
the calculation method of HFLWAO is as follows:
1) based on sell electric meal TjThe K attributes of (a) are set as follows:
akthe evaluation result of the kth attribute;
wherein, trunc (·) is an integer function, and a decimal part is directly removed; and delta (-) represents a mapping relation between the symbolic corner mark of the language evaluation quantity and the hesitation fuzzy language set.
mhAs a result of the evaluation of the package by the user,in order to evaluate the degree of progress,to evaluate the degree of membership;
with user AiLanguage evaluation set ofFor example, let β ∈ [0, g (A)i)-1]The equivalent information of β can be described by a binary, namely:
similarly, there is a transformation function Δ-1(. the binary equivalent information is converted into equivalent values, namely:
3) based on the result obtained in step 2), calculating by using a distributed language evaluation operator DAANamely:
wherein the content of the first and second substances,for a two-tuple of language evaluation,ratio representing linguistic evaluation variableFor example, the calculation formula is shown in formula (21).
2.3 unification of multiple-granularity hesitation fuzzy language evaluation information of users for selling electric packages
Considering the difference of culture and knowledge background of each user in the user load information database, the granularity of the language evaluation set used when different users evaluate the electric sales packages is different, namely: the user presents the characteristic of multi-granularity hesitation ambiguity on the evaluation information of the electric power selling package. Without loss of generality, in order to ensure the accuracy and fairness of the evaluation result, the language evaluation sets under different granularities need to be converted into the language evaluation sets under the same granularity. It is assumed that the particle size is g (A)i) Evaluation set
1) let g (═ LCM (g (A)) bei)-1,g(Af) -1) +1, where LCM (·) represents the minimum common multiple;
Let thetae=0,e=0,1,…,g(*)-1;
Based on steps 1) -6), an evaluation matrix of the user for the electricity sales package under uniform granularity can be obtainedWherein the content of the first and second substances,representing user A at a uniform granularityiPair electric sales package TjThe evaluation information of (2) is presented in the form of a binary group, namely:
is at AfThe language evaluation at the granularity of the language,in order to evaluate the degree of membership newly,to a new degree of evaluation
3. Electricity selling package recommendation and effect evaluation based on user satisfaction
The recommendation of the power selling company to the power selling package needs to consider the satisfaction degree of the user to the power selling package, and the higher the satisfaction degree is, the higher the recommendation success probability is. The satisfaction degree of the target user to the electric package selling can be obtained by firstly calculating the satisfaction degree of the user to the electric package selling in the historical user information database and then combining the similarity degree of the target user and the affiliated user.
Based on the obtained evaluation information of the users to the electricity selling package in the information database (namely, the users A under the unified granularity of the Chinese style (23) in the step 2.3)iPair electric sales package TjEvaluation information of (2)) The expectation of the evaluation information of the user is obtained, and a satisfaction matrix of the user to the electric sales package in the database can be obtainedWherein the content of the first and second substances,for user AiPair electric sales package TjSatisfaction of, calculating methods such asAnd (3) formula (24).
The satisfaction degree of the target user on the electric sales package depends on the satisfaction degree of the user on the electric sales package in the historical user library and the similarity degree of the target user and the historical user. With target user Wn and the pth type history userFor example, target user WnPair electric sales package TjSatisfaction degree SCnjAs in formula (25):
based on the calculation result of the satisfaction degree of the target user for each electricity selling package, two recommendation methods are provided: and (4) performing full-sequencing recommendation and Top-M recommendation, and respectively comparing and evaluating the recommendation effect of the electricity selling package through the root mean square error and the intersection.
(1) Full-rank recommendation
Based on the satisfaction result of the target user on each electricity selling package, the electricity selling company sorts the electricity selling packages, and recommends all the electricity selling packages and the corresponding sorting result to the target user for the target user to choose.
To measure the performance of the proposed full-rank recommendation method, the root mean square error is calculated based on the ranking results. With a target user WnFor example, the root mean square error ε1nThis can be calculated by equation (26).
Wherein J is the total number of the electricity sales packages and OnjAndare respectively provided withRepresenting a target user WnAnd sell electric set meal TjThe actual ranking result of (i.e., the user's own provided ranking result) and the ranking result of the algorithm (i.e., the ranking result of equation (25)). Root mean square error epsilon1nThe smaller the better the performance of the proposed electric sales package recommendation algorithm.
(2) Top-M recommendation
Based on the satisfaction result of the target user for each electricity selling package, the electricity selling company only recommends the packages ranked in the top M to the target user, and does not provide the ranking result.
To measure the performance of the proposed Top-M recommendation method, the first M sets of electricity sales packages obtained by the proposed algorithm are comparedSet of packages with actual ranking as top MAnd calculating the recommendation accuracy. With a target user WnFor example, the accuracy ε2nCan be calculated using equation (27):
is a target user WnThe set of packages actually ranked top M,obtaining a target user W for an algorithmnThe packages with the top M are ranked, and M is the total number of recommended packages.
Experiment 1:
1. case background
The method for recommending the electricity sales package considering the multi-granularity hesitation fuzzy set and the incomplete weight, which is described in the embodiment 1, takes a certain large park in the east of China as an analysis trial object, and users in the park comprise residents, industries and businessesWaiting for various types of users, based on 2020/1/1-2020/1/31 load data collected by the smart meter, and taking 700 bits of user A as { A ═ A }1,A2,…,A700Verification is performed on the basis of a total of 21700 load curve data for that month. Wherein, 600 users are randomly selected as history users, and 100 users are selected as target users. The histogram of the load frequency distribution of each user per day is averaged, as shown in fig. 3. The electricity selling company provides the park users with an electricity selling set of T ═ T1,T2,…,T10As shown in table 1:
table 1 electricity selling package provided by electricity selling company for park user
During the trial period, the electricity selling company carries out questionnaire survey on each user in the park and evaluates the attribute set based on the electricity selling packageAs shown in table 2:
TABLE 2 evaluation Attribute set for electricity sales packages
And collecting the evaluation information of the power selling package of the user based on the language evaluation set of each user. According to statistics, the language evaluation set granularity of the users in the park for the electricity sales packages can be divided into 3, 5, 7 and 9, as shown in formulas (28) to (31), and the average daily load rate and the average daily peak-valley difference rate of each user and the corresponding language evaluation set granularity are shown in fig. 4.
2. Historical user cluster analysis
Based on the expressions (1) - (7), calculating the load characteristic indexes of 600 historical users in the park, and clustering the historical users according to the improved K-means clustering algorithm, wherein the clustering quality indexes corresponding to different clustering numbers are shown in FIG. 5.
As shown in fig. 5, when the cluster number is 5, that is: when c is 5, the clustering quality index value is the largest, the clustering effect is the best, and the clustering result and the number of various users are respectively a late peak type, a peak flat type, a peak avoidance type, a heavy industry type and a double peak type as shown in fig. 6. Wherein, the late peak type is a working group which mainly uses the electricity load of residents, works in the daytime and goes home after leaving work at 19:00 night; the peak-flat type is mainly used by office users, businesses, office buildings and other work places, and users who get on work in the morning at about 9:00 and get off work at night at about 19: 00-20: 00; the peak avoiding type aims at night places such as bars, KTVs, bread factories, traffic lighting and the like or peak-shifting electricity users; heavy industrial types are in the steel industry, the metallurgical industry, etc. which are constantly maintained at a high level for loads; the user who uses electricity continuously in the morning and reduces the power consumption moderately after lunch is mainly the users in agriculture, textile industry and the like who work for a fixed and regular time.
Then, based on the historical user clustering result, the distance between each target user characteristic index phasor and each clustering center is calculated, and the category to which each target user belongs is determined, as shown in fig. 8.
3. Evaluation of historical users on electricity sales packages
A certain user A1 in the late peak pattern isExample, its language evaluation set is L5The user gives an evaluation matrix according to the attributes of each electricity selling packageComprises the following steps:
byIt can be known that the user shows a certain hesitation fuzzy characteristic when evaluating the electricity sales package, and the evaluation information shows that the state is also the same, and the evaluation behavior of the user can be reflected more truly.
Binding evaluation matrixWith user A1The provided electricity selling package attribute weight information is shown as a formula (32), and based on a formula (13), the established dispersion maximization model is shown as a formula (33).
Using the link () of MATLAB to obtain the weight of each attribute asEvaluating information for users on the basis of the aboveCarry out integration to obtain user A1The evaluation results of the electricity sales package are shown in table 3:
table 3 user a1Self-language-based evaluation setEvaluation result of electricity sales package
Further, the language evaluation information under different granularities is converted into the evaluation information under the same granularity. Herein with L7As the unified evaluation granularity, based on section 2.3, the evaluation result of the user on the electricity sales package under the unified granularity is obtained, as shown in table 4:
table 4 user a at unified language evaluation granularity1Evaluation result of electricity sales package
4. User satisfaction evaluation for electricity sales package
Based on the obtained uniform granularity, the user A1The evaluation result of the electricity selling package is expected to obtain the user A1The satisfaction with each electricity sales package is shown in table 5.
TABLE 5 user A1Satisfaction degree of each electricity selling package
Similarly, the satisfaction evaluation results of other historical users for the electricity selling packages can be obtained, and as shown in fig. 7, the satisfaction of the target user for each electricity selling package is calculated and obtained based on the formula (25) by combining the similarity between the target user and the historical user belonging to the category. With target users W of late peak type1For example, the similarity result with the history user of the late peak type is shown in fig. 8, and the satisfaction quantification result for each electricity sales package is shown in fig. 9.
As can be seen from fig. 7, for the same electricity selling package, the satisfaction evaluation results of different historical users in the late peak type are different, and the satisfaction of the same historical user for different electricity selling packages is also different. From FIG. 8, the eye isTarget user W1The similarity between the users and the late peak historical users is more than 97.5%, and it can be seen that the improved K-means clustering algorithm provided by the embodiment 1 can effectively classify the users with the same load characteristics into one class.
Based on the obtained historical user satisfaction, and the target user W1The similarity with the historical user is further obtained1Satisfaction degree for each electricity selling package. As can be seen from FIG. 9, target user W1Pair electric sales package T7The most satisfactory level of (c) is 587.4434. And in practice, target user W1The electricity consumption period is mainly concentrated on 19:00-22:00, and the electricity consumption period is advocated by low-carbon action and electricity sales package T7The properties of (1) are identical. Therefore, the obtained result is consistent with the reality, and the correctness of the provided user satisfaction quantification method is demonstrated.
5. Recommendation and performance evaluation of electricity selling package
Based on the satisfaction quantitative result of the target user to each electricity selling package, a full-sequencing recommendation method and a Top-M recommendation method can be respectively adopted to recommend the electricity selling packages, wherein M is 5. Still taking the target user W1 as an example, the electricity selling package and the sorting result provided by the electricity selling company to the target user W1 are shown in table 6 by adopting the full sorting recommendation method. The actual ordering results for each electricity sales package are also shown in Table 6.
TABLE 6 result of electric power selling company adopting full-sorting recommendation method for target user W1 recommendation
As shown in Table 6, the ordering results of the electricity sales packages obtained by the method provided by the invention are mostly consistent with the actual ordering results, except for electricity sales packages T9And T10The sorting results are consistent except the differences. The root mean square error ε can be calculated from equation (26)110.447. Electricity selling set meal T9And T10The reason why the sorting results are different is that the user W uses electricity at a cost equivalent to the electricity consumption cost of the user W1More concern about the environmental protection degree and increment of the packageService type, which is higher in maintenance demand than in failure diagnosis prediction demand for equipment for residential users, for home appliances, and thus, for electricity sales package T10Is higher than T9It is demonstrated that the ordering result obtained by the method of example 1 is in fact.
Similarly, by adopting the Top-5 recommendation method, the electricity selling company directly serves as the user W1Recommended electricity selling package setThe recommendation accuracy ε can be calculated from equation (27)21100%. It can be seen that the accuracy of the recommendation method of example 1.
6. Improved K-means clustering result comparison
To demonstrate the effectiveness of the improved K-means clustering algorithm herein, 5% random noise was superimposed on the collected campus user load data to obtain improved K-means clustering results and clustering centers at the optimal clustering number (c ═ 5), as shown in fig. 10. And calculating the load data after the random noise is superimposed by using an original K-means clustering algorithm to obtain a clustering center of each calculation result, as shown in FIG. 10. As can be seen from fig. 10, the result obtained by using the original K-means clustering algorithm is greatly affected by random noise, and the clustering centers except (a) and (c) are all shifted, however, the improved K-means clustering algorithm can effectively reduce the influence of noise points, and better reflect the load distribution characteristics of users.
In order to further explain the robustness of the improved K-means clustering algorithm to noise, random noise with different proportions is respectively superposed on the collected user load data to obtain the corresponding optimal clustering number c and clustering quality index Iqc.avAs shown in fig. 11. As shown in fig. 11, when the superimposed noise ratio is less than 30%, the optimal clustering numbers obtained by the improved K-means clustering algorithm are all 5, and the optimal clustering numbers start to deviate when the noise ratio reaches 10% in the original K-means clustering algorithm. In addition, as the proportion of the superimposed noise increases, the corresponding clustering quality indexes under the two clustering algorithms are reduced, but the clustering quality index corresponding to the improved K-means clustering algorithm is higher than that of the original K-means clustering algorithmAlgorithm and the descending speed is slower. It can be seen that the improved K-means clustering algorithm of embodiment 1 is robust to noise.
7. Comparing recommended sorting results of electricity selling packages of various target users
By using the method for recommending the electricity sales packages provided in embodiment 1, the satisfaction evaluation results of various target users for the electricity sales packages are calculated, as shown in fig. 12. As can be seen from fig. 12, for the same type of target users, the satisfaction trends of the users for the electricity sales packages are the same, and the ordering results of the electricity sales packages are approximately consistent. The satisfaction degree of different types of target users to the same electric sales package is greatly different. Therefore, the method for recommending the electricity selling packages provided by the embodiment 1 can meet the requirements of different types of users on the electricity selling packages. Wherein, the late peak type user mainly using the residential electricity load and the peak-flat type user mainly using the commercial and office buildings pay attention to the electricity cost, therefore, the electricity selling set meal T with lower cost and more rewarding electricity quantity7And T3Is more satisfactory. For peak avoiding users, the power consumption time is more special than that of other users, the peak avoiding users are typical peak-off users, the peak-valley time setting of the power selling package is relatively concerned, and the power selling package T9Compared with T10The electricity charge is more saved, therefore, the electric-power selling package T is more favored9. In addition, heavy industrial users mainly in the steel and iron and metallurgical industries are very sensitive to power quality disturbance, and the loss caused by disturbance is far greater than the electricity cost, so that the attention degree of the heavy industrial users on power quality value-added service is greater than other attributes of the electricity sales package, and the heavy industrial users on the electricity sales package T4The satisfaction degree of the method is the maximum and is in line with the reality. Similarly, a bimodal customer production facility, mainly in the textile industry, is sensitive to voltage sags, improving the immunity of the facility effectively reduces the risk of customer losses, and therefore, it is suitable for selling a package of electricity T8The satisfaction of (2) is greater.
8. Comparison of recommended sorting results of electricity selling packages
To further prove the effectiveness of the method for recommending the electric sales packages provided in embodiment 1 of embodiment 1, the actual ordering results of the electric sales packages of various target users are combined with the ordering obtained by the method for recommending the electric sales packages provided in embodiment 1Comparing the results of the sorting results obtained by only considering the cost, the sorting results obtained by clustering by using the original K-means and the results obtained by directly adopting the electric sales package recommendation method provided by the embodiment 1 without clustering, and respectively calculating the mean values epsilon of the root mean square errors of various target users under the full-sorting recommendation and the Top-5 recommendation1And the average value of accuracy ∈2As shown in fig. 13.
As can be seen from fig. 13, the sorting result obtained by the method for recommending an electric package sold in embodiment 1 is not much different from the actual sorting result, although the sorting of 2 to 4 packages is different, the maximum root mean square error is 0.6325, the electric packages sold in the top 5 are consistent with the actual package, and the accuracy reaches 100%, thereby illustrating the accuracy of the method for recommending an electric package sold in embodiment 1.
Compared with the actual sequencing result, the root mean square error of the recommendation result considering only the cost of the electricity selling package is larger, the maximum is 4.8374, the recommendation accuracy is not higher than 60%, therefore, the evaluation of the electricity selling package by the user has diversity, and under the condition that the cost is not greatly different, the value-added service, the reward policy and the like contained in the electricity selling package are also the attention points of the user.
In addition, if the original K-means clustering algorithm is used to cluster the historical users, as shown in fig. 10, except for the late peak users and the peak avoidance users, the clustering centers of other types of users are shifted, as shown in fig. 13, the root mean square errors of the late peak users and the peak avoidance users are 0.4472 and 0.7746 respectively, and the accuracy is 100%, while the root mean square errors of the other types of users are 1.7321, 3.1623 and 0.8944 respectively, which are higher than the root mean square error of the result obtained by the method for recommending the electric package sold in embodiment 1, and the recommendation accuracies are 80%, 60% and 80%, respectively, and are lower than 100%. And if the clustering of the historical users is not considered, the electricity sales packages are recommended directly according to the similarity between the target user and all the historical users, the root mean square error corresponding to the recommendation result is large, the minimum is 1, the maximum is 3.3466, the recommendation accuracy is not more than 80%, and the recommendation performance is poor.
To sum up, the improved K-means clustering algorithm provided by the method for recommending an electricity sales package in embodiment 1 clusters users, can more accurately cluster users with similar load characteristics into one category, and considers the multiple attributes of an electricity sales package, the multi-granularity hesitation fuzzy characteristic of user evaluation information and incomplete weight information, so that the root mean square error of the recommendation result obtained by the method for recommending an electricity sales package is smaller, the recommendation accuracy is higher, and the recommendation performance is better.
Finally, it is also noted that the above-mentioned lists merely illustrate a few specific embodiments of the invention. It is obvious that the invention is not limited to the above embodiments, but that many variations are possible. All modifications which can be derived or suggested by a person skilled in the art from the disclosure of the present invention are to be considered within the scope of the invention.
Claims (5)
1. The method for recommending the electricity selling package considering the multi-granularity hesitation fuzzy set and the incomplete weight is characterized by comprising the following steps of: the specific process is as follows:
s1, establishing a user load information database based on the user electricity utilization characteristic indexes, clustering the users based on an improved K-means clustering algorithm, and determining the category of the target user;
step S2, considering the incomplete weight information and the multi-granularity hesitation fuzzy language evaluation set provided by the user, and constructing an evaluation matrix of the user for the electric sales package;
step S3, calculating the satisfaction of the historical user and the target user to the electric power selling package;
and step S4, obtaining the complete-sequence electricity selling package recommendation and the Top-M electricity selling package recommendation based on the user satisfaction quantification result.
2. The method of claim 1, wherein the method further comprises:
the specific process of step S1 is as follows:
step S1.1, building user load information database
Based on user A ═ { A1,A2,…,Ai,…,AIThe daily load data for a month is:wherein the content of the first and second substances,k is the total hours of a month; based on the monthly load rateNumber of hours of maximum utilizationMean value of peak-to-valley rate of working dayMean peak-to-valley rate of off-dayPeak load meanMean average of load rate in flat phaseMean value of load rate at valley periodObtain user AiLoad characteristic index vector ofAnd a user load information database corresponding to the user and the load characteristic index vector thereof is established;
(1) the monthly load rate:
wherein the content of the first and second substances,andare respectively user AiThe average load per month and the maximum load,for user AiDaily load data for the kth hour of the total hours of a month;
(2) number of hours of maximum utilization
Wherein the content of the first and second substances,for user AiElectricity consumption in whole month;
(3) mean value of peak-to-valley rate of working day
Wherein, KWIs the total number of days of the working day,the day maximum load on day i of the working day,the day minimum load for day i of the non-workday;
(4) mean peak-to-valley rate of off-day
Wherein, KNWIs the total number of days on the non-working day,the day maximum load on day j of the non-workday,the daily minimum load on day j of the non-workday;
(5) peak load mean
Wherein peak periods are defined as 08:00-11:00 and 18:00-21:00, T is the total number of days, and T is KW+KNW,For user A in the peak period of the t dayiThe average value of the loads of (a),is the user A on the t dayiThe load mean value of (1);
(6) mean average of load rate in flat phase
Wherein, the ordinary period is defined as 06:00-08:00, 11:00-18:00 and 21:00-22:00,for the tth balance period user AiThe load mean value of (1);
(7) mean load factor at off-peak
Wherein the valley time periods are defined as 22:00-24:00 and 00:00-06:00,for the user A in the trough period of the t dayiThe load mean value of (1);
step S1.2, clustering based on historical user load information database
Based on the user load information database established in the step S1.1, clustering historical users by adopting an improved K-means clustering algorithm, and specifically comprising the following steps:
step S1.2.1, making the clustering number C equal to 2, randomly selecting C users, and marking as C1,C2,…,Cj,…,CcUsing the load characteristic index vector as the clustering center
Step S1.2.3 for user AiAre compared and obtainedAnd will be user AiClassified as j, and the number of users belonging to the j is set as IjThe corresponding user is represented as
Step S1.2.4, calculating the square error J based on the classification result of step S1.2.3 as equation (8)c(t):
Wherein, t is the iteration number,is corresponding to Ai,jClass j user AiA load characteristic index vector of (a);
step S1.2.5, recalculating the cluster centers
Calculating the average distance D between the load characteristic index vectors of all users in each class and the clustering center, selecting the users with the distance to the clustering center being less than 2D in the class, and then recalculating the mean value of the load characteristic indexes of the selected users as a new clustering center;
step S1.2.6, determine whether formula (9) is satisfied:
Jc(t+1)-Jc(t)≥0 (9)
if not, t is t +1, and then the step S1.2.1 is returned to continue the iteration;
if the formula (9) is satisfied, the iteration is ended, and the clustering quality index I is calculated based on the clustering result of the t iterationqc.av:
Wherein, Iqc(i) When the number of clusters is c, the closeness degree of the ith user to the cluster center of the class to which the ith user belongs is represented as follows:
wherein the content of the first and second substances,represents the average distance between classes as user AiAverage value of the distance between the user load characteristic index phasor and other user load characteristic indexes except the category to which the user belongs;representing the average distance within a class, as user AiAverage of the phasor distances from the other user load characteristic indexes in the category to which the user belongs;
step S1.2.7, making the cluster number c ═ c +1, returning and executing step s 1.2.1-step S1.2.6 until the cluster number reaches the maximum value
Step S1.2.8, comparing the clustering quality index I under different clustering numbersqc.av,Iqc.avThe corresponding cluster number at the maximum is the optimal cluster number c;
step S1.3, calculating the similarity between the target user and the historical user based on Pearson correlation coefficient
The target user is a user needing to recommend the electricity selling package and is represented as W ═ W1,W2,…,Wn,…,WMAnd obtaining load characteristic index vectors of each target user based on daily load data of each user in a certain month by combining the formula (1) with the formula (7)Calculating the center of each cluster under the optimal cluster number cThe class corresponding to the minimum distance is the target user WnThe category to which it belongs;
set target user WnBelonging to the p-th class of historical users, wherein the number of the class of users is NpIs shown asThe similarity between the target user and the historical user is described by using the Pearson correlation coefficient:
wherein the content of the first and second substances,represents a target user WnAnd historical usersThe similarity of (2);andrespectively represent users WnAnd historical usersThe d-th load characteristic index of (1),andrespectively represent users WnAnd historical usersAverage value of load characteristic index.
3. The method of claim 2, wherein the method further comprises:
the specific process of the step 2 is as follows:
step S2.1, building an evaluation matrix of the user for the electricity sales package based on the hesitation fuzzy language set
The set of electricity selling set is T ═ T1,T2,…Tj,…,TJUser a ═ a }1,A2,…,Ai,…,AIThe evaluation attribute set for the electricity sales package isDescribing the evaluation information of the user on the electricity selling package by adopting hesitation fuzzy language:
let user AiIs a language evaluation set ofWherein g (A)i) Represents the granularity of the language evaluation set and is odd,represents the evaluation amount of the p-th language, p is 0,1, …, g (A)i) -1 symbolic corner marks for language evaluation quantities; language evaluation setThe elements are arranged in sequence:
user AiSelf-language-based evaluation setAnd evaluating the electricity selling package, wherein an evaluation matrix is as follows:
wherein the content of the first and second substances,represents user AiPair electric sales package TjProperty (2) ofGiven the hesitation fuzzy language evaluation information, J and K are respectively the J-th package and the K-th attribute;
step S2.2, determining the attribute weight of the electricity selling package considering the incomplete weight
Wherein "other" represents user AiThe incomplete weight information is provided in the form of,representation for the kth attributeUser A ofiThe degree of deviation between language evaluations given to each electricity sales package:
wherein the content of the first and second substances, for user AiEvaluating the kth attribute in the qth package;
integrating the evaluation information of each user by adopting an Hesitation Fuzzy Language Weighted Average Operator (HFLWAO) to obtain the evaluation matrix of each user on the electric sales package
Wherein the content of the first and second substances,represents user AiPair electric sales package TjThe evaluation results of (1);
the HFLWAO is calculated as follows:
(1) based on electricity selling set meal TjThe K attributes of (a) are set as follows:
akthe evaluation result of the kth attribute;
wherein trunc (·) is an integer function; delta (-) represents a mapping between symbolic logo of language evaluation quantity and hesitation fuzzy language set, mhAs a result of the evaluation of the package by the user,in order to evaluate the degree of progress,to evaluate the degree of membership;
let beta be [0, g (A) ]i)-1]The equivalent information of beta is described by a binary group:
presence of a transformation function Δ-1(. h), converting the binary equivalent information into equivalent values:
Wherein the content of the first and second substances,for a two-tuple of language evaluation,ratio representing linguistic evaluation variables:
step S2.3, unification of multi-granularity hesitation fuzzy language evaluation information of the user on the electric power selling package
Converting language evaluation sets under different granularities used by a user for evaluating the power selling package into language evaluation sets under the same granularity, and determining the granularity as g (A)i) Evaluation setEvaluation information of (2)Conversion to g (A) on the basis of particle sizef) Evaluation setEvaluation information of (2)The process is as follows:
step S2.3.1, let g (×) -LCM (g (a)i)-1,g(Af) -1) +1, where LCM (·) represents the minimum common multiple;
Let thetae=0,e=0,1,…,g(*)-1;
Step S2.3.5 based onCalculating the normalized language evaluation variable ratio, as shown in formula (22):
step S2.3.6, obtaining the unified language evaluation information based on the formula (22)
Based on the steps S2.3.1-S2.3.6, an evaluation matrix of the user for the electricity sales package under uniform granularity is obtainedWherein the content of the first and second substances,representing user A at a uniform granularityiPair electric sales package TjEvaluation information of (2):
4. The method of claim 3, wherein the method further comprises:
the specific process of the step 3 is as follows:
user A at uniform granularity based on formula (23)iPair electric sales package TjEvaluation information of (2)The expectation of the evaluation information of the user is obtained, and the satisfaction matrix of the user to the electric sales package is obtainedWherein the content of the first and second substances,for user AiPair electric sales package TjSatisfaction of (2):
target user WnPair electric sales package TjSatisfaction degree SCnjComprises the following steps:
5. the method of claim 4, wherein the method further comprises:
the full-sequencing electricity selling package recommendation specifically comprises the following steps:
calculating a root mean square error epsilon based on satisfaction results of the target users for each electricity selling package1nThen according to the root mean square error epsilon1nSell electricity according to the correspondence from small to largeAnd recommending the package ordering to the target user for the target user to decide:
wherein J is the total number of the electricity sales packages and OnjAndrespectively represent for the target user WnAnd sell electric set meal TjThe actual sorting result of (a) and the sorting result obtained by the algorithm of the formula (25);
the Top-M electricity selling package recommendation specifically comprises the following steps:
based on the satisfaction result of the target user for each electricity selling package, the electricity selling company only recommends the packages ranked in the top M to the target user, and does not provide the ranking result;
recommend to target user WnThe first M sets of electricity sales packagesSet of packages with actual ranking as top MAccuracy of recommendation e2nComprises the following steps:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111359084.4A CN113947444A (en) | 2021-11-17 | 2021-11-17 | Electricity selling package recommending method considering multi-granularity hesitation fuzzy set and incomplete weight |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111359084.4A CN113947444A (en) | 2021-11-17 | 2021-11-17 | Electricity selling package recommending method considering multi-granularity hesitation fuzzy set and incomplete weight |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113947444A true CN113947444A (en) | 2022-01-18 |
Family
ID=79338190
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111359084.4A Pending CN113947444A (en) | 2021-11-17 | 2021-11-17 | Electricity selling package recommending method considering multi-granularity hesitation fuzzy set and incomplete weight |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113947444A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116562990A (en) * | 2023-07-03 | 2023-08-08 | 湖北国网华中科技开发有限责任公司 | Electricity selling transaction service recommendation method and device |
CN116596640A (en) * | 2023-07-19 | 2023-08-15 | 国网山东省电力公司营销服务中心(计量中心) | Recommendation method, system, equipment and storage medium for electric power retail electric charge package |
CN116596640B (en) * | 2023-07-19 | 2024-06-21 | 国网山东省电力公司营销服务中心(计量中心) | Recommendation method, system, equipment and storage medium for electric power retail electric charge package |
-
2021
- 2021-11-17 CN CN202111359084.4A patent/CN113947444A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116562990A (en) * | 2023-07-03 | 2023-08-08 | 湖北国网华中科技开发有限责任公司 | Electricity selling transaction service recommendation method and device |
CN116562990B (en) * | 2023-07-03 | 2023-10-27 | 湖北国网华中科技开发有限责任公司 | Electricity selling transaction service recommendation method and device |
CN116596640A (en) * | 2023-07-19 | 2023-08-15 | 国网山东省电力公司营销服务中心(计量中心) | Recommendation method, system, equipment and storage medium for electric power retail electric charge package |
CN116596640B (en) * | 2023-07-19 | 2024-06-21 | 国网山东省电力公司营销服务中心(计量中心) | Recommendation method, system, equipment and storage medium for electric power retail electric charge package |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Fu et al. | Clustering-based short-term load forecasting for residential electricity under the increasing-block pricing tariffs in China | |
Tsekouras et al. | Two-stage pattern recognition of load curves for classification of electricity customers | |
CN111582911A (en) | Multi-user and power grid friendly interactive power utilization system | |
Mahmoudi-Kohan et al. | A three-stage strategy for optimal price offering by a retailer based on clustering techniques | |
CN105117810A (en) | Residential electricity consumption mid-term load prediction method under multistep electricity price mechanism | |
CN101266674A (en) | Method and system for determination of an appropriate strategy for supply of renewal energy onto a power grid | |
Bandyopadhyay et al. | Individual and aggregate electrical load forecasting: One for all and all for one | |
Lai et al. | Individualized pricing of energy storage sharing based on discount sensitivity | |
Wen et al. | Demand side management in smart grid: A dynamic-price-based demand response model | |
Panapakidis et al. | Implementation of modified versions of the K-means algorithm in power load curves profiling | |
CN106992942A (en) | A kind of SDN resource pricing methods based on resource load and user's request | |
Zhou et al. | Smart energy management | |
Xu et al. | Spatial-temporal load forecasting using AMI data | |
Kojury-Naftchali et al. | Identifying susceptible consumers for demand response and energy efficiency policies by time-series analysis and supplementary approaches | |
CN116611862A (en) | Prediction method for monthly marketized purchase electric quantity of power supply company | |
CN113947444A (en) | Electricity selling package recommending method considering multi-granularity hesitation fuzzy set and incomplete weight | |
Lu et al. | Research on creating multi-attribute power consumption behavior portraits for massive users | |
CN112288496A (en) | Load classification calculation method and tracking analysis method for power industry | |
CN110599059A (en) | Incremental power distribution network user-oriented electricity selling decision method and system | |
Wang et al. | Three-dimensional maturity model of regional power users against the background of the ubiquitous power internet of things | |
Wang et al. | Optimization of retail packages adapted to the electricity spot market with the goal of carbon peak and carbon neutrality | |
Jin et al. | Comprehensive evaluation of impacts of connecting distributed generation to the distribution network | |
Maniar et al. | Validity index based improvisation in reproducibility of load profiling outcome | |
Mabuggwe et al. | Design and development of an intelligent tool for retail electric provider plan selection | |
Zhang et al. | Joint planning of residential electric vehicle charging station integrated with photovoltaic and energy storage considering demand response and uncertainties |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |