CN111126498A - Customer electricity consumption behavior characteristic analysis method based on classification analysis - Google Patents
Customer electricity consumption behavior characteristic analysis method based on classification analysis Download PDFInfo
- Publication number
- CN111126498A CN111126498A CN201911361606.7A CN201911361606A CN111126498A CN 111126498 A CN111126498 A CN 111126498A CN 201911361606 A CN201911361606 A CN 201911361606A CN 111126498 A CN111126498 A CN 111126498A
- Authority
- CN
- China
- Prior art keywords
- load
- electricity consumption
- factors
- consumption behavior
- rule
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/80—Management or planning
- Y02P90/82—Energy audits or management systems therefor
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Educational Administration (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Artificial Intelligence (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- Primary Health Care (AREA)
- General Health & Medical Sciences (AREA)
- Water Supply & Treatment (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a customer electricity consumption behavior characteristic analysis method based on classification analysis, which comprises the following steps: step one, establishing various user load characteristic indexes of power users in various power consumption modes; analyzing the influence factors of the power consumption behavior mode, and then extracting the main influence factors of the p user power consumption behavior modes by using a principal component analysis method; and step three, analyzing the customer electricity consumption behavior based on a decision tree classification method. The method has the advantages of fine analysis result, high accuracy, and capability of going deep into the user level, and is beneficial to the refinement of the formulation of the demand response strategy.
Description
Technical Field
The invention relates to the technical field of power grid load characteristic analysis, in particular to a customer power utilization behavior characteristic analysis method based on classification analysis.
Background
Load characteristic analysis is the basis of intelligent power grid research, and the traditional power system electricity utilization characteristic analysis method mainly has two types: firstly, analyzing according to influence factors, namely extracting dependent variables one by one on the premise that the other variables are kept unchanged, and describing the influence degree of the dependent variables qualitatively or quantitatively, wherein the objects are regional power grids which are far inferior to the research precision and depth of a power supply side and cannot meet the requirement of pushing intelligent demand side management; and secondly, performing classification analysis according to industries, namely analyzing the power utilization characteristics of various industries or a user refined to a certain type, and qualitatively or quantitatively obtaining the influence of various types of users on the power utilization characteristics of the regional power grid. But its accuracy is not high and it is difficult to perform load prediction based on the analysis result.
In general, the existing load characteristic analysis technology has low data utilization rate, coarse analysis results and low precision, cannot go deep into a user level, and is not beneficial to the refinement of the requirement response strategy. The rapid development of big data technologies brings more opportunities and challenges to the load analysis work. The new idea and method for load characteristic analysis in the big data environment are significant.
Disclosure of Invention
The technical problem to be solved by the invention is to establish a customer electricity consumption behavior characteristic analysis method based on classification analysis, define a novel load characteristic index, extract main electricity consumption behavior influence factors by using a principal component analysis method, and classify electricity consumption behaviors according to different characteristics by using a decision tree classification method.
The technical scheme of the invention is as follows:
a customer electricity consumption behavior feature analysis method based on classification analysis comprises the following steps:
the method comprises the following steps: establishing various user load characteristic indexes of power users in various power utilization modes;
step two: analyzing the influence factors of the power consumption behavior mode, and then extracting the main influence factors of p user power consumption behavior modes by using a principal component analysis method;
step three: and (3) analyzing the customer electricity consumption behavior based on a decision tree classification method.
Preferably, the user load characteristic index in the first step includes a traditional load characteristic index, a load importance level index, a flexible power utilization characteristic index and a short-term load demand response index.
Preferably, the conventional load characteristic indexes include daily load rate, daily maximum/small load, daily average load, daily peak-valley difference characteristic indexes, and also include indexes related to demand response.
Preferably, the load importance level index includes four levels, I level index: the load is ensured safely; grade II indexes: the main productive load; grade III indexes are as follows: auxiliary production load; grade IV indexes are as follows: non-productive load.
Preferably, the flexible electricity consumption characteristic index is based on electricity consumption characteristic analysis, and a load decomposition model is established to calculate the rigid load and the flexible load in the load:
L＝Lbasic+Lweather
in the formula: l is the total load; l isbasicThe basic load reflects a certain general development trend of the load in a longer duration, and has certain stability, periodicity and seasonality; l isweatherThe load is a meteorological sensitive load, reflects the influence of meteorological factors such as temperature, humidity, rainfall, wind speed and cloud cover on load change, is a load component which fluctuates up and down along with the change of the meteorological factors, and is mainly reflected as a cooling and heating load.
Preferably, the short-term demand response indexes of the load comprise load adjustability, load price flexibility, demand response speed index climbing rate and demand response capacity index.
Preferably, the internal and external factors of the step two, which influence the electricity consumption behavior of the client, are user factors, system factors, environmental factors and policy factors.
Preferably, the method for extracting the main influence factors of the electricity consumption behavior pattern of the user in the second step comprises the following steps:
1) taking m observation days, wherein n observation sample matrixes of the influence factors of the power consumption behaviors to be analyzed are as follows:
wherein the element xijThe meaning of (1) is that the observed value of the ith observation day on the jth influence factor to be analyzed, each variable is subjected to standardized transformation, and the transformation formula is as follows:
in formula (II), x'ijThe observation value after the standardized transformation on the jth influence factor to be analyzed on the ith observation day is obtained;
the average value of the j to-be-analyzed influence factors on the m observation days;
2) obtaining a correlation matrix R, matrix elements RjgThe calculation formula of (2) is as follows:
in formula (II), x'igThe observation value is subjected to standardized transformation on the g influence factor to be analyzed on the ith observation day;
3) the characteristic equation of the correlation matrix R is | λ I-R | ═ 0, and the characteristic equation is solved to obtain the characteristic root λ of the R matrixjJ — 1,2, … …, n, calculating the variance contribution rate of each factor to be analyzed:
in the formula, αjJ is 1,2, … …, n for the variance contribution rate of the jth influencing factor to be analyzed;
4) taking the first p to-be-analyzed factors with larger variance contribution rate as principal components, namely extracting p main influence factors of the user electricity consumption behavior mode, namely the characteristic quantity of the user electricity consumption behavior mode, and meeting the following requirements:
preferably, in the third step, in order to extract the rule from the decision tree, a rule is created for each path from the root to the leaf node, a logic AND of each splitting criterion on a given path forms a front piece of the rule, AND a back piece of the rule formed by the leaf node of the class prediction is stored;
each rule extracted implies a logical OR between them, which are mutually exclusive and exhaustive since they are extracted directly from the tree; for a given rule precursor, any condition that does not improve the estimation accuracy of the rule can be pruned, thereby generalizing the rule.
Preferably, the rule induction of the sequential covering algorithm is used in the third step; extracting IF-THEN rules directly from training data using a sequential override algorithm, without having to generate a decision tree, the names of the algorithm are derived from rules that are learned sequentially one at a time, where each rule of a given class ideally overrides many of the meta-progenitors of that class; the sequential overlay algorithm includes AQ, CN2 and RIPPER.
The invention has the beneficial effects that:
the method deeply utilizes the information contained in the load data of the power consumer, can accurately analyze the power consumption behavior of the consumer, is beneficial to accurately predicting the load by the power grid, has fine analysis result and high accuracy, can go deep into the user level, is beneficial to the refinement of the formulation of a demand response strategy, improves the energy utilization efficiency, further improves the stability of the power grid, and provides better service for the power grid.
Drawings
FIG. 1 is a schematic block diagram of the method of the present invention.
Fig. 2 is a diagram of factors influencing the electricity consumption behavior of the user according to the present invention.
Detailed Description
As shown in fig. 1-2, the present invention provides a customer electricity consumption behavior feature analysis method based on classification analysis, which comprises the following specific implementation steps:
the method comprises the following steps: establishing various user load characteristic indexes of the power users in various power utilization modes:
1. the traditional load characteristic index is as follows: daily load rate, daily maximum/small load, daily average load, daily peak-to-valley difference, etc.
The daily load rate is used for describing daily load curve characteristics and representing the imbalance of the load in one day, and the higher load rate is beneficial to the economic operation of the power system, and the definition formula is as follows:
the numerical value of the daily load rate is related to the nature, category and composition of users, production shift and proportion of various types of electricity (domestic electricity, power electricity and process electricity) in the system, and is also related to measures for adjusting load. With the development of an electric power system, the constitution of users, the electricity utilization mode and the process characteristics can be changed, the proportion occupied by various users can be changed, and the daily load rate can be changed accordingly.
2. Load importance rating
The electric equipment of the power consumer is various in types and different in importance, and the load importance of the power consumer can be divided into four levels according to the loss degree caused by power failure or power shortage of the electric equipment, as shown in table 1:
TABLE 1 load importance rating
The different industries have different power utilization characteristics and load composition characteristics, the power utilization equipment of the different industries can be classified according to the definition of the importance level in the table 1, demand response resources with adjustable capacity are selected from non-productive loads and auxiliary productive loads, the proportion or the capacity of the part of loads is obtained through industry research, representative terminal power utilization equipment which can be used as the demand response resources is summarized on the basis, and subsequent research is carried out on specific equipment.
Taking a power consumer in a certain business industry as an example, the time periods in which the loads are mainly concentrated are as follows: 8: 30-22: 00, the main electric equipment is corridor lighting, office lighting, a central air conditioner, a split air conditioner, a ceiling air conditioner, a water boiler, a heat pump, a water chilling unit, an electric boiler, an elevator, monitoring equipment, an alarm system, a computer and the like, and the classification and importance grades are shown in table 2:
TABLE 2 commercial user Equipment Classification and importance level
The commercial users have smaller electrical loads, mainly equipment such as office lighting, fans, air conditioners, elevators and the like, and the electrical loads are larger in summer and winter. The temporary stop or the rotation control of non-production loads such as commercial escalators, water boilers, brightening billboards, office air conditioners and other equipment does not influence the normal operation of users, can be used as demand response resources, and the proportion of the demand response resources is easy to obtain through research and development.
3. Electrical characteristic index for flexibility
"Flexible load" is defined as: the load can be transferred or reduced by a certain technical means, and the process has certain cost-effectiveness, and the time span meets certain requirements.
In the past, the load is often regarded as a passively controlled physical terminal, and the electricity characteristic index is expressed by a fixed value. With the construction of smart grids and the implementation of demand-side management, loads that are originally regarded as rigid gradually exhibit certain elasticity, and many consumers have reducible or transferable loads in their power consumption devices, as shown in table 3:
TABLE 3 example of interruptible consumer
Therefore, defining an index of 'rigid load to flexible load ratio', the level of the flexible load of the power consumer can be evaluated, and the method is a basis for evaluating the load demand response capability and the energy-saving potential.
The patent decomposes the load to obtain the basic load and the meteorological sensitive load, and roughly represents the flexible load by the meteorological sensitive load. On the basis of electric characteristic analysis, a load decomposition model is established to calculate rigid load and flexible load in the load:
L＝Lbasic+Lweather(3)
in the formula: l is the total load; l isbasicThe basic load reflects a certain general development trend of the load in a longer duration, and has certain stability, periodicity and seasonality; l isweatherThe load is a meteorological sensitive load, reflects the influence of meteorological factors (temperature, humidity, rainfall, wind speed, cloud cover and the like) on load change, is a load component which fluctuates up and down along with the change of the meteorological factors, and is mainly reflected as a cooling and heating load.
1) Base load separation
The base load includes a trend component and a periodic component. The trend component reflects a certain continuous development trend of the load in a period of time, and mainly reflects the trend that the power load gradually deviates or approaches to the load under the influence of factors such as economy and the like in load prediction, and is not influenced by factors such as weather, holidays and random factors; the periodic component reflects seasonal and periodic cyclical variations of the load.
According to the method for combined modeling of the time series, firstly, an exponential curve model is used for fitting the trend component of the historical load, then the annual cycle component and the cycle component are added one by one, and the model formula is as follows:
in the formula: a. the1，B1Is an exponential term coefficient; cj，Dj，EjIs the annual cycle term coefficient; fk，Gk，HkIs the periodic term coefficient;ω1＝2π/365，ω2＝2π/7。
2) meteorological sensitive load separation
Lweather＝L-Lbasic(5)
4. Short-term load demand response indicator
In order to facilitate analysis of the resource potential of the load side resource applied to load scheduling, novel power utilization characteristic indexes such as load adjustability, load price elasticity and demand side response potential need to be established on the basis of analyzing the traditional power utilization characteristic indexes.
(1) Load adjustability
Most load-side resources have load-tunability, but their values are not the same. Some users have too little elasticity and are not suitable for participating in load scheduling; some users are too costly to provide load side resources to achieve the desired benefits. Therefore, the load adjustability index is established, and the load side resource suitable for scheduling is favorably mined.
Load tunability refers to the rate at which the user load can be adjusted without participating in any load scheduling project, and is mainly considered from two aspects:
technically adjustable: whether the direct or indirect control of the load can be realized through the smart grid technology;
is economically adjustable: whether subsidies obtained by load adjustment can compensate losses or not, some industrial load production processes have strict process limitations, and if the influence of transfer on enterprises is large, load scheduling is not feasible economically.
(2) Price elasticity of load
The load of the power terminal has certain elasticity, the load change is caused by the price change, therefore, under the excitation of different electricity price measures, different types of power consumer loads have different transfer or reduction rates which can be expressed by the load price elasticity, namely
The power demand of users has a certain relation with the price, but the trend and the amplitude of the change of the demand of various users along with the electricity price are different; the load price elasticity is relative change of electricity consumption caused by relative change of electricity price, and can be used for measuring the sensitivity degree of different electric equipment to price, and the electric equipment with higher load price elasticity is more sensitive to price change, so that the load can be controlled by using a price-based demand response strategy.
(3) Demand response speed index-ramp rate
In the economic dispatching of the power system, the climbing rate of the unit needs to be considered comprehensively, the climbing rate of the entity unit refers to the output which can be increased or decreased per unit time of each generator unit, and the climbing rate of the load is defined as the power which can be increased or decreased per unit time of the terminal power consumption equipment, and the unit is MW/min.
Under the condition that the electric equipment accounts for a large proportion, the demand response speed index can be obtained by researching the internal characteristics of various electric equipment; in addition, a demand response speed index can be obtained through model identification according to historical data.
(4) Demand response capacity indicator
The demand response capacity index is defined as the power which can be reduced by the electric equipment at a certain moment through a demand response means, and is related to the installation degree of time, excitation and intelligent control equipment:
the demand response capacity index is based on a load curve, and if the electric equipment is in a shutdown state within a certain period of time, the demand response capacity is 0, so that the load curve of the electric equipment can be regarded as a maximum output constraint, and the value is a time-varying quantity;
the reduction degree of the power of the electric equipment is in direct proportion to the excitation size, so that the demand response capacity index is in direct proportion to the excitation size;
the effect of the electric equipment participating in the demand response project is also related to the installation degree of the intelligent control equipment, such as an intelligent electric meter with a negative control function, a high-level measurement system and the like, which is a technical condition for ensuring that the electric equipment can be remotely controlled and influences the magnitude of the demand response capacity index.
Step two: analyzing the influence factors of the power consumption behavior patterns, and extracting the main influence factors of the p user power consumption behavior patterns by using a principal component analysis method
1. Research on influence factors of power utilization mode of user
The internal and external factors influencing the electricity utilization behavior of the client are many, and mainly comprise four types of users, systems, environments and policies, and the influence factors of each type can be further refined, so that a three-layer structure can be presented, as shown in fig. 2.
(1) Internal factors
1) User factor
The user factors refer to factors related to user behaviors and buildings and mainly comprise building envelope structures and user response willingness.
a) The building envelope structure mainly refers to the heat preservation and insulation capacity of the building, and is particularly characterized in that when a response strategy is executed, the indoor temperature and humidity increase/decrease speed is achieved. Generally speaking, the better the building envelope, the less heat is exchanged between the indoor and outdoor, and the less influence on the response willingness of the user is.
b) The user's willingness to respond can be understood as two aspects, one is whether the user is willing to participate in the response; second, the upper and lower limits of the indoor temperature at which the user feels comfortable.
2) System factor
The system factors refer to factors related to participation in electric power response, and mainly comprise five types, namely system type, system personnel management level, system automation level, system working condition and system response strategy (inlet and outlet water temperature, running frequency of a cooling/freezing water pump and the like).
a) System type (including system installed capacity): because of the diversification of the system operation principle, the ice/water cold accumulation air conditioner is mainly mentioned, ice is made and stored in the ice/water storage device by utilizing the low valley load electric power at night, the stored cold energy is released in the daytime, and the air conditioner electricity load and the installed capacity of an air conditioning system are reduced at the peak time of a power grid.
b) The system personnel management level refers to the awareness level of system personnel in the aspects of energy conservation and demand response and the capacity of optimizing the operation of the air conditioning system.
3) The automation level of the system mainly measures the automation capacity of the air conditioning system, including whether the energy management system, the control equipment, the fine measurement device and the like exist. Generally, the higher the automation level, the more convenient the user's response, the positive impact on the willingness to respond, and the higher the reliability and speed of the response.
4) The system working condition level refers to basic states of the system, such as the running state, the running efficiency, the running life and the like. The general air conditioner with longer operation life has higher energy.
(2) External factors
1) Environmental factors
The environmental factors refer to factors related to temperature and humidity, population, response occurrence time and the like, and part of the factors have randomness.
a) The temperature and humidity refer to the temperature and humidity of the external environment. Under the same comfort level, when the outside temperature is higher, the output of the air conditioning system is larger.
b) The crowd quantity refers to the crowd quantity in a building, and has certain regularity, for example, for office buildings, more workers are in Monday-Friday offices, and fewer workers are at weekends; the store, on the contrary, peaks the number of people on weekends. But overall, the population size is also highly random.
c) The response occurrence time refers to the trigger time of the demand response event. If the situation occurs in summer at noon and during the peak of the electricity consumption of the air conditioner host, negative effects can be caused on the response willingness and the response capability of the user, and the participation willingness is low.
2) Policy factors
The policy factors refer to national and local policies or regulations related to the demand response, and the incentives, the electricity price, the energy conservation and other policies are used for guiding the enthusiasm of participating in the demand response and encouraging the user to participate in the demand response.
a) The incentive policy mainly refers to incentive fees given by government, electric power companies and other organizations when demand response is implemented, for example, in enterprises with voluntary load interruption in a peak period, the accumulated interruption of 1 hour per 1 ten thousand kilowatts is supplemented by 1 ten thousand yuan, which is equivalent to 1 kilowatt-hour electricity compensation of 1 yuan.
b) The electricity price policy mainly refers to an electricity price scheme for prompting users to peak load shifting and valley load shifting, for example, peak electricity prices can bring rapid reduction of peak load, and time-of-use electricity prices can guide users to transfer unnecessary electricity to low-valley periods, so that the effect of peak load shifting is achieved.
c) The energy-saving policy is a series of policies established for improving energy utilization rate, controlling energy consumption and reducing pollutant emission, such as contract energy management, and is an energy-saving business mode for paying the total cost of an energy-saving project through reduced energy cost.
The factors influencing the flexible load response are diversified, some factors can be quantified, and some factors can be only analyzed qualitatively, but the factors can more or less influence the implementation of the demand response and can be embodied in response characteristics.
2. Extracting main influence factors of user electricity consumption behavior pattern
(1) Taking m observation days, wherein n observation sample matrixes of the influence factors of the power consumption behaviors to be analyzed are as follows:
wherein the element xijThe meaning of (1) is that the observed value of the ith observation day on the jth influence factor to be analyzed, each variable is subjected to standardized transformation, and the transformation formula is as follows:
in formula (II), x'ijThe observation value after the standardized transformation on the jth influence factor to be analyzed on the ith observation day is obtained;
the average value of the j to-be-analyzed influence factors on the m observation days;
2) obtaining a correlation matrix R, matrix elements RjgThe calculation formula of (2) is as follows:
in formula (II), x'igThe observation value is subjected to standardized transformation on the g influence factor to be analyzed on the ith observation day;
3) the characteristic equation of the correlation matrix R is | λ I-R | ═ 0, and the characteristic equation is solved to obtain the characteristic root λ of the R matrixjJ — 1,2, … …, n, calculating the variance contribution rate of each factor to be analyzed:
in the formula, αjJ is 1,2, … …, n for the variance contribution rate of the jth influencing factor to be analyzed;
4) taking the first p to-be-analyzed factors with larger variance contribution rate as principal components, namely extracting p main influence factors of the user electricity consumption behavior mode, namely the characteristic quantity of the user electricity consumption behavior mode, and meeting the following requirements:
step three: customer electricity consumption behavior analysis based on decision tree classification method
(1) Extracting rules from a decision tree
Decision tree classification is a popular classification method and is well known for accuracy. IF-THEN rules can be extracted using the decision tree, building a rule-based classifier, which is easier to understand, especially when the decision tree is very large.
To extract rules from the decision tree, one rule is created for each path from the root to the leaf node, the logical AND along each split criterion on a given path forms the front piece of the rule (IF part), AND the leaf node holding the class prediction forms the back piece of the rule (THEN part).
Each rule extracted implies a logical OR between them, which are mutually exclusive and exhaustive since they are extracted directly from the tree. Mutual exclusion means that there is no possibility of rule conflicts, since no two rules are triggered by the same primitive ancestor. Exhaustive means that there is one rule for each attribute-value combination, such that the rule set does not require a default rule.
For a given rule precursor, any condition that does not improve the estimation accuracy of the rule can be pruned, thereby generalizing the rule.
(2) Rule induction using sequential override algorithm
IF-THEN rules can be extracted directly from the training data using the sequential covering algorithm without having to generate decision trees. The names of the algorithms are derived from rules being learned sequentially (one at a time), with each rule of a given class ideally covering many of the meta-progenitors of that class (and hopefully not covering the meta-progenitors of other classes). The sequential covering algorithm is the most widely used method of mining the disjunctive set of classification rules. There are many sequential overlay algorithms, popular ones including AQ, CN2 and the recently proposed RIPPER.
The strategy of the algorithm is as follows: learning a rule at a time, deleting the ancestor covered by the rule each time a rule is learned, and repeating the process for the remaining ancestors, such sequential learning of rules in contrast to decision tree induction. Since each path to a leaf in a decision tree corresponds to a rule, decision tree generalization can be viewed as learning a set of rules simultaneously.
The sequential override algorithm learns rules for one class at a time, ideally when learning rules for class C, it is desirable to override all (or many) of the training metaprogenitors for class C, and none (or very little) of the metaprogenitors for other classes. In this way, the learned rules have high accuracy. Rules need not be high coverage because a class may have multiple rules, so that different rules may cover different ancestors of the same class.
Basic sequence overlay algorithm:
inputting: d-type labeled metaancestor data set; att _ vals-the set of all attributes and their possible values.
And (3) outputting: a set of IF-THEN rules.
The method comprises the following steps:
(1) rule-set { }; // the initial set of learned rules is null
(2) for each class cdo
(3)repeat
(4) Rule ═ lear _ one _ Rule (D, Att _ vals, c); // finding the best rule for the current class
(5) Deletion of primitive progenitors of Rule coverage from D
(6) Until termination condition satisfied// the quality of the rule, if there are no more training primitive progenitors or returns, is below a user-specified threshold
(7)Rule-set＝{Rule-set,Rule}
(8)end
(9) Return Rule-set
The step (4) adopts a greedy depth-first strategy, and when a new attribute test is added to the current rule, the test which can improve the quality attribute of the rule most is selected according to the training sample, for example, the accuracy can be selected as the quality measurement. The greedy search does not allow backtracking, and at each step, heuristically adds the selection that looks best at the time. In this process, the result is not ideal if we involuntarily make a bad choice. To reduce the chance of this happening, the best k attributes may be selected to test into the current rule instead of one. Thus, we can perform a beam search of width k, maintaining k best candidates at each step, rather than one.
Learn _ one _ Rule needs to measure the quality of the Rule and each time an attribute test is considered, it must check whether adding the test to the conditions of the current Rule results in an improved Rule. Here, we use statistical significance checking to determine whether the effect of a rule is not due to chance factors, but rather predicts a true relationship between attribute values and classes. The check compares the distribution of observed classes of the metaancestor covered by the rule against the expected class distribution generated by the random prediction of the rule. We wish to assess whether the observed difference between these two distributions is likely to be accidental, and the likelihood statistics can be used
Where m is the number of classes, for primitive progenitors that satisfy the rules, fi is the observed frequency of class i in those primitive progenitors, and ei is the expected frequency of the rule for making a random prediction of class i. The statistics obey a chi-square distribution of m-1. The higher the likelihood, the more significantly the difference of the regular correct predictor number compared to the random guesser. That is, the performance of the rules is not contingent, and the likelihood helps identify rules with significant coverage.
Claims (10)
1. A customer electricity consumption behavior feature analysis method based on classification analysis is characterized by comprising the following steps:
the method comprises the following steps: establishing various user load characteristic indexes of power users in various power utilization modes;
step two: analyzing the influence factors of the power consumption behavior mode, and then extracting the main influence factors of p user power consumption behavior modes by using a principal component analysis method;
step three: and (3) analyzing the customer electricity consumption behavior based on a decision tree classification method.
2. The customer electricity consumption behavior feature analysis method based on classification analysis as claimed in claim 1, wherein the user load characteristic index in step one comprises a traditional load characteristic index, a load importance level index, a flexible electricity consumption characteristic index and a short-term load demand response index.
3. The method as claimed in claim 2, wherein the conventional load characteristic indexes include daily load rate, daily maximum/small load, daily average load, daily peak-to-valley difference characteristic indexes, and indexes related to demand response.
4. The customer electricity consumption behavior feature analysis method based on classification analysis as claimed in claim 2, wherein the load importance level index includes four levels, I level index: the load is ensured safely; grade II indexes: the main productive load; grade III indexes are as follows: auxiliary production load; grade IV indexes are as follows: non-productive load.
5. The customer electricity consumption behavior feature analysis method based on classification analysis as claimed in claim 2, wherein the flexible electricity consumption characteristic index is based on electricity consumption characteristic analysis, and a load decomposition model is established to calculate the rigid load and the flexible load in the load:
L＝Lbasic+Lweather
in the formula: l is the total load; l isbasicIs a basic load which reflects a certain general development trend of the load in the duration time and has stability, periodicity and seasonality; l isweatherThe load is a meteorological sensitive load, reflects the influence of meteorological factors such as temperature, humidity, rainfall, wind speed and cloud cover on load change, is a load component which fluctuates up and down along with the change of the meteorological factors, and is reflected as a cooling and heating load.
6. The method as claimed in claim 2, wherein the demand response indexes of the short term load include load adjustability, load price flexibility, demand response speed index ramp rate and demand response capacity index.
7. The method for analyzing the characteristics of the customer electricity consumption behaviors based on the classification analysis as claimed in claim 2, wherein the second step is implemented by internal and external factors, namely user factors, system factors, environmental factors and policy factors, which influence the electricity consumption behaviors of customers.
8. The customer electricity consumption behavior feature analysis method based on classification analysis as claimed in claim 7, wherein the method for extracting the main influence factors of the user electricity consumption behavior pattern in the second step comprises the following steps:
1) taking m observation days, wherein n observation sample matrixes of the influence factors of the power consumption behaviors to be analyzed are as follows:
wherein the element xijThe meaning of (1) is that the observed value of the ith observation day on the jth influence factor to be analyzed, each variable is subjected to standardized transformation, and the transformation formula is as follows:
in formula (II), x'ijThe observation value after the standardized transformation on the jth influence factor to be analyzed on the ith observation day is obtained;
the average value of the j to-be-analyzed influence factors on the m observation days;
2) obtaining a correlation matrix R, matrix elements RjgThe calculation formula of (2) is as follows:
in formula (II), x'igThe observation value is subjected to standardized transformation on the g influence factor to be analyzed on the ith observation day;
3) the characteristic equation of the correlation matrix R is | λ I-R | ═ 0, and the characteristic equation is solved to obtain the characteristic root λ of the R matrixjJ — 1,2, … …, n, calculating the variance contribution rate of each factor to be analyzed:
in the formula, αjJ is 1,2, … …, n for the variance contribution rate of the jth influencing factor to be analyzed;
4) taking the first p to-be-analyzed factors with larger variance contribution rate as principal components, namely extracting p main influence factors of the user electricity consumption behavior mode, namely the characteristic quantity of the user electricity consumption behavior mode, and meeting the following requirements:
9. the method for analyzing customer electricity consumption behavior characteristics based on classification analysis as claimed in claim 1, wherein, in step three, in order to extract rules from the decision tree, a rule is created for each path from the root to the leaf node, the logical AND of each splitting criterion on a given path forms the front piece of the rule, AND the back piece of the leaf node forming the rule of class prediction is stored;
each rule extracted implies a logical OR between them, which are mutually exclusive and exhaustive since they are extracted directly from the tree; for a given rule precursor, any condition that does not improve the estimation accuracy of the rule is pruned, thereby generalizing the rule.
10. The customer electricity consumption behavior feature analysis method based on classification analysis as claimed in claim 1, wherein, in step three, rule induction of sequential covering algorithm is used; extracting IF-THEN rules directly from training data using a sequential override algorithm, without generating a decision tree, the names of the algorithm resulting from the rules being learned sequentially one at a time, wherein each rule of a given class ideally overrides many of the meta-progenitors of that class; the sequential overlay algorithm includes AQ, CN2 and RIPPER.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911361606.7A CN111126498A (en) | 2019-12-25 | 2019-12-25 | Customer electricity consumption behavior characteristic analysis method based on classification analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911361606.7A CN111126498A (en) | 2019-12-25 | 2019-12-25 | Customer electricity consumption behavior characteristic analysis method based on classification analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111126498A true CN111126498A (en) | 2020-05-08 |
Family
ID=70502639
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911361606.7A Pending CN111126498A (en) | 2019-12-25 | 2019-12-25 | Customer electricity consumption behavior characteristic analysis method based on classification analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111126498A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111783827A (en) * | 2020-05-27 | 2020-10-16 | 中能瑞通（北京）科技有限公司 | Enterprise user classification method and device based on load data |
| CN112581012A (en) * | 2020-12-25 | 2021-03-30 | 国网北京市电力公司 | Electricity customer classification method participating in demand response |
| CN113420733A (en) * | 2021-08-23 | 2021-09-21 | 北京黑马企服科技有限公司 | Efficient distributed big data acquisition implementation method and system |
| CN115983430A (en) * | 2022-12-02 | 2023-04-18 | 成都市迈德物联网技术有限公司 | Method and system for managing and optimizing comprehensive energy system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103872680A (en) * | 2014-03-24 | 2014-06-18 | 国家电网公司 | Method for evaluating interaction capacity of flexible loads |
| CN105260798A (en) * | 2015-10-21 | 2016-01-20 | 中国电力科学研究院 | Big data miner for multi-dimensional load characteristic analysis |
-
2019
- 2019-12-25 CN CN201911361606.7A patent/CN111126498A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103872680A (en) * | 2014-03-24 | 2014-06-18 | 国家电网公司 | Method for evaluating interaction capacity of flexible loads |
| CN105260798A (en) * | 2015-10-21 | 2016-01-20 | 中国电力科学研究院 | Big data miner for multi-dimensional load characteristic analysis |
Non-Patent Citations (3)
| Title |
|---|
| 周倩: "基于基因表达式编程的分类与聚类研究", 《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑》 * |
| 章博 等: "基于决策树和数据驱动的零电量用户筛选方法", 《基于决策树和数据驱动的零电量用户筛选方法》 * |
| 马尚才 等: "《决策支持与知识发现》", 31 March 2005 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111783827A (en) * | 2020-05-27 | 2020-10-16 | 中能瑞通（北京）科技有限公司 | Enterprise user classification method and device based on load data |
| CN112581012A (en) * | 2020-12-25 | 2021-03-30 | 国网北京市电力公司 | Electricity customer classification method participating in demand response |
| CN113420733A (en) * | 2021-08-23 | 2021-09-21 | 北京黑马企服科技有限公司 | Efficient distributed big data acquisition implementation method and system |
| CN115983430A (en) * | 2022-12-02 | 2023-04-18 | 成都市迈德物联网技术有限公司 | Method and system for managing and optimizing comprehensive energy system |
| CN115983430B (en) * | 2022-12-02 | 2023-12-29 | 成都市迈德物联网技术有限公司 | Comprehensive energy system management optimization method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111126498A (en) | Customer electricity consumption behavior characteristic analysis method based on classification analysis | |
| Sial et al. | Detecting anomalous energy consumption using contextual analysis of smart meter data | |
| US10175709B2 (en) | Consumer electric power control system and consumer electric power control method | |
| Phuangpornpitak et al. | A study of load demand forecasting models in electric power system operation and planning | |
| JP2020501491A (en) | System and method for dynamic energy storage system control | |
| KR101800286B1 (en) | Method and system for managing energy usage with using big date of energy usage | |
| CN110380444B (en) | Capacity planning method for distributed wind power orderly access to power grid under multiple scenes based on variable structure Copula | |
| CN106372762A (en) | Microgrid economic optimal operation design method with demand response included | |
| CN113255968B (en) | Commercial office building refined load prediction method based on equipment and behavior information | |
| CN112149890A (en) | Comprehensive energy load prediction method and system based on user energy label | |
| CN108346009A (en) | A kind of power generation configuration method and device based on user model self study | |
| Kim et al. | Time-series clustering and forecasting household electricity demand using smart meter data | |
| Yu et al. | Estimating power demand shaving capacity of buildings on an urban scale using extracted demand response profiles through machine learning models | |
| Zhang et al. | Generation of sub-item load profiles for public buildings based on the conditional generative adversarial network and moving average method | |
| Park et al. | Demand power forecasting with data mining method in smart grid | |
| Kimata et al. | Operation planning for heat pump in a residential building | |
| CN115829418A (en) | Power consumer load characteristic portrait construction method and system suitable for load management | |
| Golovinski et al. | Electricity consumption forecast of clusters of buildings based on recurrent neural networks | |
| Zhang et al. | Time-of-use pricing model considering wind power uncertainty | |
| Brito et al. | Forecasting of Energy Consumption: Artificial Intelligence Methods | |
| CN111915105A (en) | Method and device for predicting electricity consumption in area | |
| CN110852628A (en) | Rural medium and long term load prediction method considering development mode influence | |
| Chen et al. | Review on Smart Meter Data Clustering and Demand Response Analytics | |
| Liu et al. | Optimal guidance strategy for flexible load based on hybrid direct load control and time of use | |
| CN116089847B (en) | Distributed adjustable resource clustering method based on covariance agent |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200508 |