CN113065278B - Frequent pattern mining-based prediction method for wind power small-occurrence event statistical characteristic model in rich period - Google Patents
Frequent pattern mining-based prediction method for wind power small-occurrence event statistical characteristic model in rich period Download PDFInfo
- Publication number
- CN113065278B CN113065278B CN202110276871.6A CN202110276871A CN113065278B CN 113065278 B CN113065278 B CN 113065278B CN 202110276871 A CN202110276871 A CN 202110276871A CN 113065278 B CN113065278 B CN 113065278B
- Authority
- CN
- China
- Prior art keywords
- wind power
- occurrence
- mode
- rich period
- power small
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/06—Wind turbines or wind farms
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/28—The renewable source being wind energy
-
- 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
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E10/00—Energy generation through renewable energy sources
- Y02E10/70—Wind energy
- Y02E10/76—Power conversion electric or electronic aspects
Abstract
The invention relates to a prediction method of a wind power small occurrence event statistical characteristic model in a rich period based on frequent pattern mining, which comprises the steps of firstly extracting corresponding duration time periods and interval time periods when wind power small occurrence events occur in the rich period of each wind power plant; secondly, respectively carrying out clustering analysis on the continuous time period mode and the interval time period mode based on a K-means clustering algorithm to obtain different basic modes, simultaneously obtaining meteorological features corresponding to each basic mode, and training a support vector machine classifier so as to use the basic modes to carry out event sequence recoding on the wind power sequence; and finally, mining the event sequence through an APRIORI correlation analysis algorithm to obtain the correlation between the wind power small-occurrence events in the rich period and the events, further establishing an autocorrelation statistical characteristic model of the wind power small-occurrence events in the rich period, and predicting through the model. The method deeply digs the autocorrelation characteristic among the wind power small-occurrence events in the rich period, and effectively solves the problem that the modeling by a mathematical model is difficult.
Description
Technical Field
The invention belongs to the technical field of power grid wind power fluctuation characteristic modeling, and particularly relates to a method for predicting a wind power small-occurrence event statistical characteristic model in a rich period based on frequent pattern mining.
Background
The energy is the material basis on which the human beings rely to live, is the life line of the development of the national economy and society, and the renewable energy is developed on a large scale under the large background of global climate change and energy safety. Renewable energy in China is developed rapidly, wherein the total amount of wind power integration and consumption is increased at a high speed, and according to data published by the State energy agency, the wind power integration in China is 2.17 hundred million kilowatts by 6 months in 2020. While China pays attention to the development of renewable energy, the consumption of large-scale renewable energy sources faces more challenges due to the problems of natural characteristics of resources, power system conditions and market mechanisms in China. The output of wind, light and other renewable energy sources has randomness and volatility, and the system is difficult to completely adapt to new trend requirements due to the facts that the current power system in China is lack of flexibility in adjustment, the power grid dispatching operation mode is relatively rigid and the like, and a large-scale unit is difficult to exert the advantages of energy conservation and high efficiency. With the large-scale development of renewable energy sources, the contradiction between local consumption in China is gradually highlighted, and the problems of wind abandoning, light abandoning and water abandoning are more serious in some areas.
Nowadays, a lot of problems still exist in large-scale wind power prediction and fluctuation characteristic modeling, aiming at common wind power output scene analysis, most of the wind power output scenes can only discuss the output characteristics of wind power under typical scenes, and various 'unexpected' situations which may occur under extreme conditions are ignored, for example, the wind power output during the rich period may be in a low level instead. The rich period refers to 1-5 months and 10-12 months of each year, the climate change degree is large, and the wind power level is relatively high according to the statistics data of the past year. However, due to special meteorological causes, in a certain region or a local range, the abnormal phenomena of no wind and small wind occur, so that the wind power level at the moment is far lower than the level which is required in the previous period and the current period, namely the wind power small occurrence event in the rich period. The scene is different from an extreme climbing scene, the extreme climbing scene refers to the fact that the change process of wind power is violent, generally refers to a phenomenon in a short time period, and a wind power small-occurrence event in a rich period refers to the fact that wind power is lost in a long time scale. The wind power small-occurrence event in the rich period is an unpredictable wind power long-time shortage phenomenon, and the existing scheduling plan of the system is greatly influenced. On one hand, the power shortage changes the actual output level of each unit; on the other hand, the sudden power change provides great challenges for unit maintenance and safe operation of the power system. Under the condition, deep analysis is carried out aiming at an extreme scene that a wind power small-occurrence event occurs in a rich period, and a scene analysis model of the wind power small-occurrence event in the rich period is established, so that the output characteristic of wind power is quantitatively depicted, the wind power output rule is more deeply excavated, the completeness of the model is improved, the reliability of safe operation of a power grid is enhanced, the power grid is helped to dispatch more effectively utilize wind power resources, and basic theoretical support is provided for subsequent efficient utilization of renewable energy.
Disclosure of Invention
The invention aims to provide a method for establishing a wind power small occurrence event self-correlation statistical characteristic model in the rich period based on an APRIORI correlation analysis algorithm after event sequence recoding is carried out on a wind power sequence by considering the influence of meteorological conditions on wind power output and completing statistical characteristic research of the wind power small occurrence event in the rich period by using actual wind power data.
In order to achieve the purpose, the invention adopts the technical scheme that:
a prediction method of a wind power small occurrence event statistical characteristic model in a rich period based on frequent pattern mining comprises the following specific steps:
step 2, introducing meteorological data, selecting meteorological indexes, respectively carrying out clustering analysis on the continuous time period mode and the interval time period mode based on a K-means clustering algorithm to obtain different basic modes, simultaneously obtaining meteorological features corresponding to each basic mode, and carrying out learning training on the meteorological data under each basic mode by using a support vector machine algorithm to obtain a support vector machine classifier;
and 4.1, mining and analyzing the event sequence through an APRIORI association analysis algorithm according to the event sequence after the dimensionality reduction obtained by the processing of the step 3.2 to obtain the association relation between the wind power small-occurrence events in the rich period and the events. And (3) performing frequent item set search on the mode event sequence based on the principle of APRIORI, and counting SC (mode) as the support degree, namely the frequency of occurrence of a certain mode. Because the wind power small-occurrence events at the wind-rich period which are far away from each other have small correlation, only three frequent item sets need to be considered, namely the wind power small-occurrence events at the wind-rich period and the corresponding frequent item sets at intervals are included in sequence. Therefore, the autocorrelation information of the wind power small events in the rich period can be contained from the frequent binomial set to the frequent trinomial set, the autocorrelation characteristics of the mode A and the mode B are considered from the frequent binomial set, and after the mode A occurs, the probability of the mode B occurring is shown as the following formula:
the formula establishes the autocorrelation characteristics of the mode A and the mode B, and the autocorrelation characteristics of all the modes can be calculated by a frequent item set;
4.2, setting a frequent item threshold as 2 support counts, analyzing the event sequence after dimensionality reduction based on an APRIORI algorithm, searching a frequent binomial set and a frequent trinomial set, and obtaining corresponding support counts of the frequent binomial set and the frequent trinomial set so as to calculate and obtain autocorrelation characteristics among different modes;
and 4.3, obtaining the autocorrelation characteristics among the modes based on the statistics of the steps, further establishing an autocorrelation statistical characteristic model of the wind power small-occurrence events in the rich period, wherein all the frequent binomial sets and the frequent trinomial sets can represent the autocorrelation characteristics of the wind power small-occurrence events in the rich period, and the set of incidence relations among different modes forms the autocorrelation statistical characteristic model.
and 5.1, obtaining the duration time period of the wind power small-occurrence event in the rich period and each basic mode in the interval time period according to the clustering analysis in the step 2, numbering the duration time period modes as 1-3 and the interval time period modes as 4-6, and obtaining the number of events in the mode, the average wind speed value, the average wind direction value, the average air temperature value, the average air pressure value and the average duration time value/average interval time value corresponding to each basic mode according to the step 2.
And 5.2, supposing that the duration time period and the interval time period of the wind power small-occurrence events in the rich period of the patterns 1 and 4 occur first, then the duration time periods of the wind power small-occurrence events occur together, wherein the patterns are determined to be one of the types 1, 2 and 3, since the autocorrelation characteristics and the specific probability conditions among the patterns are calculated according to the step 4.2, the probabilities of the patterns 1, 2 and 3 occurring next and the corresponding meteorological characteristic data values are utilized to calculate the sum of the probabilities of the patterns 1, 2 and 3 occurring next and the meteorological data characteristic values corresponding to the patterns, namely the meteorological data characteristic values corresponding to the patterns occurring next, specifically comprising wind speed, wind direction, air temperature and air pressure, and then the data are utilized to identify based on the predict function of the svm model, so that the specific patterns belonging to the patterns can be obtained, and finally, iteration is carried out to predict the patterns occurring later.
In the modeling method for the statistical property of the wind power small-occurrence event in the rich period based on frequent pattern mining, the implementation of the step 1 comprises the following steps:
step 1.1, selecting actual wind power data of a wind power station in one year based on a data driving mode, wherein data acquisition is performed once every 15 minutes every day, and the number of the data is 96 in total every day;
step 1.2, preprocessing actual wind power data, deleting data which are missing and have numerical values which are obviously beyond the physical significance range by adopting a deletion method, and realizing the processing of missing values and abnormal values;
step 1.3, based on actual wind power data obtained after preprocessing, screening the occurrence time and frequency of the wind power small occurrence events in the rich period by taking the wind power data which is lower than 5% of the maximum wind power in one year as a standard, and extracting corresponding duration time periods and interval time periods when the wind power small occurrence events in the rich period occur in each wind power plant;
and 1.4, respectively depicting a probability distribution diagram of duration time and interval time of the wind power small events in the rich period and a probability distribution diagram of starting time of the wind power small events in the rich period from time and space scales according to the obtained characteristic data.
In the modeling method for the statistical property of the wind power small-occurrence event in the rich period based on frequent pattern mining, the implementation of the step 2 comprises the following steps:
step 2.1, introducing meteorological data, and selecting wind speed, wind direction, air temperature and air pressure as meteorological indexes;
and 2.2, respectively carrying out clustering analysis on the continuous time period mode and the interval time period mode based on a K-means clustering algorithm to obtain different basic modes, and simultaneously obtaining meteorological features corresponding to the basic modes. The method comprises the following specific steps of respectively carrying out cluster analysis on historical meteorological data in the continuous time period mode and the interval time period mode by using a k-means algorithm, and forming each basic mode:
2.2.1, randomly selecting k meteorological data sample points from the t historical meteorological data sample points as initial clustering centers, sequentially calculating the distances from the rest sample points to the initial clustering centers, and assigning the sample points to the closest clusters so as to form initial k clusters;
step 2.2.2, respectively calculating the mean values of the sample point data in the k clusters to obtain central samples, using the k central samples as new clustering centers, recalculating the distance between each meteorological data sample point and the new clustering centers, and allocating each sample point to the cluster with the closest distance again according to the minimum distance principle;
step 2.2.3, recalculating the mean values of the k clusters, and circulating the step 2.2.2 and the step 2.2.3 until the cluster center is not changed any more;
and 2.3, learning and training the meteorological data under each basic mode by using a support vector machine algorithm to obtain a support vector machine classifier. Analyzing a support vector machine algorithm with a meteorological pattern class larger than 2, constructing k SVM sub-classifiers by adopting a one-by-one identification method, marking sample data belonging to the jth class as a positive class and marking sample data not belonging to the jth class as a negative class when constructing the jth SVM sub-classifier; during training, the discrimination function values of the sub-classifiers are respectively calculated for historical meteorological data samples, and the category corresponding to the maximum discrimination function value is selected as the category of the meteorological data samples, so that multi-classification is realized.
In the modeling method for the statistical characteristics of the wind power small-occurrence events in the rich period based on frequent pattern mining, the step 3 is realized by the following steps:
3.1, analyzing the characteristics of the wind power small-occurrence events in the rich period, finding that the power output of the wind power small-occurrence events in the rich period is extremely low, the duration of the occurrence is an important characteristic, and in addition, the interval time between two wind power small-occurrence events in the rich period is also an important attention object, so that the duration and interval time parameters of the wind power small-occurrence events in the rich period are utilized to express the information of the wind power small-occurrence events in the rich period, and according to the step 2.2, the average values of the corresponding historical meteorological data in each duration time period and interval time period mode are respectively obtained by utilizing a k-means algorithm, so that each mode has a corresponding meteorological characteristic data set, and then the meteorological characteristic data sets are subjected to cluster analysis to obtain different meteorological modes, and the duration time period and interval time period modes can be classified according to different meteorological modes, so that each basic sub-mode can be obtained;
and 3.2, in order to facilitate event coding, clustering the modes A corresponding to the wind power small-occurrence event duration time periods in the rich period to respectively obtain different modes and numbering the modes in sequence, clustering the modes B corresponding to the wind power small-occurrence event interval time periods in the rich period to respectively obtain different modes and numbering the modes in sequence, performing event sequence recoding on the wind power sequence by using each obtained basic mode, and reducing the dimension of a multidimensional power sequence into a 2-dimensional event sequence.
The invention has the beneficial effects that: the influence of a plurality of meteorological factors such as wind speed, wind direction, air temperature and air pressure on the wind power small-occurrence event in the rich period is considered, so that the analysis and modeling result is more accurate for the extreme scene of the wind power small-occurrence event in the rich period; clustering analysis is respectively carried out on the wind power small-occurrence event duration time period and interval time period patterns in the rich period based on a K-means clustering algorithm to obtain different basic patterns, so that event sequence recoding is carried out on the wind power sequence by using each basic pattern, the time sequence is converted into an event sequence, and the incidence relation between the wind power small-occurrence events in the rich period and the events is better mined; the APRIORI correlation analysis algorithm can deeply dig out the autocorrelation characteristic among the wind power small events in the rich period, and effectively solves the problem that the modeling by a mathematical model is difficult.
In practical application, the time and frequency of the wind power small-occurrence event in the rich period are screened out through historical wind power data and meteorological data, then a clustering algorithm is utilized to perform clustering analysis on the duration time and interval time period modes of the wind power small-occurrence event in the rich period to obtain each basic mode, event recoding is performed on a power sequence by using each basic mode, and autocorrelation characteristic mining is performed on the wind power small-occurrence event in the rich period based on an APRIORI algorithm, so that the wind power law output is dug more deeply, the completeness of the model is improved, the reliability of safe operation of a power grid is enhanced, and the power grid scheduling is helped to more effectively utilize wind power resources.
Drawings
FIG. 1 is a schematic diagram of the APRIORI algorithm;
FIG. 2 is a schematic flow diagram of a method according to an embodiment of the invention;
FIG. 3 (a) is a corresponding duration probability distribution diagram when a wind power small-occurrence event occurs in the Kyoto Tang region in the rich period;
FIG. 3 (b) is a corresponding interval time probability distribution diagram when a wind power small occurrence event occurs in the rich period in Jingjin Tang area;
fig. 3 (c) is a corresponding initial time probability distribution diagram when a wind power small-scale occurrence event occurs in the rich period in the jing jin tang area.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
In order to solve the technical problems in the prior art, the embodiment provides a wind-rich period wind power small-occurrence event statistical characteristic modeling method based on frequent pattern mining, and the wind power statistical characteristic of the wind-rich period wind power small-occurrence event is modeled to obtain an autocorrelation characteristic model, so that deep analysis is performed on the extreme scene, the completeness of the model is improved, the reliability of safe operation of a power grid is enhanced, and the power grid is dispatched to more effectively utilize wind power resources.
In the embodiment, the wind power small occurrence statistical characteristics in the rich period are not analyzed from the wind power perspective alone, but the influence of meteorological conditions (wind speed, wind direction, air temperature and air pressure) on the wind power fluctuation characteristics is considered. The method comprises the steps of firstly completing statistical characteristic research of a wind power small-occurrence event in a rich period by utilizing actual wind power data to obtain a continuous time period mode and an interval time period mode of the wind power small-occurrence event in the rich period, then analyzing the continuous time period mode and the interval time period mode by utilizing a clustering algorithm by introducing historical meteorological data to obtain different basic modes, then performing event sequence recoding on a wind power sequence, and further establishing an autocorrelation statistical characteristic model of the wind power small-occurrence event in the rich period based on an APRIORI correlation analysis algorithm.
The modeling method for the wind power small occurrence events in the rich period based on frequent pattern mining comprises the steps of screening out the occurrence time and frequency of the wind power small occurrence events in the rich period by utilizing actual wind power data and time mark information based on a data driving mode, and extracting corresponding duration time periods and interval time periods when the wind power small occurrence events in the rich period occur in each wind power plant; on the basis, meteorological data are introduced, clustering analysis is respectively carried out on the continuous time period mode and the interval time period mode based on a K-means clustering algorithm to obtain different basic modes, meteorological features corresponding to the basic modes are obtained at the same time, and a support vector machine classifier is trained; using a basic mode to recode the event sequence of the wind power sequence; after the event sequence after dimensionality reduction is obtained, mining the event sequence through an APRIORI correlation analysis algorithm to obtain the correlation between the wind power small-occurrence events in the rich period and the events, further establishing an autocorrelation statistical characteristic model of the wind power small-occurrence events in the rich period, and completing prediction of the wind power small-occurrence events in the rich period.
The embodiment is realized by the following technical scheme, as shown in fig. 2, the method comprises the following steps:
s1, based on a data driving mode, screening out the occurrence time and frequency of the wind power small-occurrence events in the rich period by using actual wind power data and time mark information, extracting corresponding duration time periods and interval time periods when the wind power small-occurrence events in the rich period occur in each wind power plant, and performing statistical characteristic analysis.
In S1, preprocessing actual wind power data, deleting data which are missing and have numerical values which obviously exceed the physical significance range by adopting a deletion method, and realizing missing value and abnormal value processing; based on actual wind power data obtained after preprocessing, screening the occurrence time and frequency of the wind power small events in the rich period by taking the wind power data lower than 5% of the maximum wind power in one year as a standard, and extracting corresponding duration time periods and interval time periods when the wind power small events in the rich period occur in each wind power plant; according to the characteristic data obtained by statistics, a wind power small event duration and interval time probability distribution diagram in the rich period and a wind power small event initial time probability distribution diagram in the rich period are respectively described on the time scale and the space scale, so that the statistical analysis of the wind power small events in the rich period is realized. Statistical analysis is performed on the sample data, and probability distribution graphs of duration, interval time and starting time corresponding to the occurrence of the wind power small-occurrence event in the rich period in the kyujin Tang area are obtained and are shown in fig. 3 (a), 3 (b) and 3 (c).
S2, meteorological data are introduced, meteorological indexes are selected, clustering analysis is carried out on the modes of the continuous time periods and the interval time periods respectively on the basis of a K-means clustering algorithm to obtain different basic modes, meteorological features corresponding to the basic modes are obtained at the same time, learning training is carried out on the meteorological data in the basic modes by using a support vector machine algorithm, and a support vector machine classifier is obtained. Considering that the output of wind speed and wind direction on wind power has direct correlation, and meanwhile, the air pressure and temperature of the area where the fan is located can cause air convection to a certain extent, so that the wind speed, the wind direction, the air temperature and the air pressure are finally selected as meteorological data indexes of cluster analysis of all modes under the condition of considering the wind power small occurrence event in the rich period. For sample data, taking the wind power small-occurrence events of the triparena, the sunshine river and the cooperative three wind power plants in the rich period as an example, after K-means clustering is carried out on the continuous time period mode and the interval time period mode respectively, the obtained basic modes and the meteorological characteristics thereof are shown in tables 1 and 2.
TABLE 1
| Mode(s) | Number of events in a pattern | Mean wind speed | Average wind direction | Mean air temperature | Mean air pressure | Average duration (hours) |
| 1 | 31 | 2.97 | 125.60 | 11.62 | 599.50 | 9.55 |
| 2 | 17 | 3.40 | 238.62 | -9.81 | 586.85 | 11.34 |
| 3 | 20 | 3.66 | 211.04 | 19.49 | 649.62 | 9.38 |
TABLE 2
| Mode(s) | Number of events in a pattern | Mean wind speed | Mean wind direction | Average air temperature | Mean air pressure | Average duration (hours) |
| 4 | 32 | 7.06 | 243.94 | 4.56 | 595.79 | 109.45 |
| 5 | 28 | 4.65 | 162.76 | 18.53 | 633.46 | 39.05 |
| 6 | 7 | 8.22 | 256.28 | -4.65 | 597.34 | 486.18 |
The method comprises the following specific steps of respectively carrying out cluster analysis on historical meteorological data in the continuous time period mode and the interval time period mode by using a k-means algorithm, and forming each basic mode:
s2.1, randomly selecting k meteorological data sample points from t historical meteorological data sample points as initial clustering centers, then sequentially calculating the distances from the rest sample points to the initial clustering centers, and assigning the sample points to the cluster with the closest distance so as to form initial k clusters;
s2.2, respectively calculating the mean value of the sample point data in the k clusters to obtain a center sample, using the k center samples as new clustering centers, recalculating the distance between each meteorological data sample point and the new clustering centers, and allocating each sample point to the cluster with the closest distance again according to the minimum distance principle;
and S2.3, recalculating the mean value (center sample) of the k clusters, and circulating S2.2 and S2.3 until the cluster center is not changed any more.
And S3, according to each basic mode obtained through clustering analysis in the S2, event sequence recoding is carried out on the wind power sequence by using the basic mode, and dimension reduction processing on the wind power sequence is realized.
In S3, analyzing the statistical characteristics of the wind power small events in the rich period to know that the duration and interval time parameters of the wind power small events in the rich period can express the wind power small event information in the time sequence, and obtaining each basic mode according to the statistical clustering analysis in S2; in order to facilitate event coding, clustering the modes A corresponding to the wind power small-occurrence event duration time period in the rich period respectively to obtain different modes and numbering the modes in sequence, clustering the modes B corresponding to the wind power small-occurrence event interval time period in the rich period respectively to obtain different modes and numbering the modes in sequence, carrying out event sequence recoding on the wind power sequence by using the obtained basic modes, and reducing the dimension of a multidimensional power sequence into a 2-dimensional event sequence.
And S4, mining the event sequence according to the dimension-reduced event sequence processed in the S3 through an APRIORI correlation analysis algorithm to obtain the correlation between the wind power small-occurrence events in the rich period, further establishing an autocorrelation statistical characteristic model of the wind power small-occurrence events in the rich period, and completing prediction of the wind power small-occurrence events in the rich period.
In S4, frequent item set search is performed on the pattern event sequence based on APRIORI principle, and SC (mode) is taken as a support count, that is, the frequency of occurrence of a certain pattern. Because the relevance of the wind power small-occurrence events in the rich period which are far away from each other is small, only three frequent item sets need to be considered, namely the wind power small-occurrence events in the rich period and the corresponding frequent item sets in the interval are sequentially included. Therefore, the autocorrelation information of the wind power small-occurrence events in the rich period can be contained from the frequent binomial set to the frequent trinomial set, wherein autocorrelation characteristics of all modes can be calculated by the frequent binomial set; setting a frequent item threshold as 2 support counts, analyzing the event sequence after dimensionality reduction based on an APRIORI algorithm, searching a frequent binomial set and a frequent trinomial set, and obtaining corresponding support counts of the frequent binomial set and the frequent trinomial set so as to calculate and obtain autocorrelation characteristics among different modes; and counting based on the steps to obtain the autocorrelation characteristics among all the modes, further establishing an autocorrelation statistical characteristic model of the wind power small-occurrence event in the rich period, and completing the prediction of the wind power small-occurrence event in the rich period on the basis.
S5, acquiring actual wind power data and meteorological data of the deer source, the cooperation and the sunshine river wind power station in 2018 years, processing the obtained association relation between each basic mode according to the step 4.2, and using the obtained frequent binomial set and the frequent trinomial set to represent the autocorrelation characteristic of the wind power small-occurrence event in the rich period, and performing example analysis based on the characteristic to predict the future wind power small-occurrence event in the rich period, wherein the method specifically comprises the following steps of:
s5.1, obtaining the basic modes of the duration time period and the interval time period of the wind power small-occurrence event in the rich period according to the clustering analysis in the step 2, numbering the modes of the duration time period as 1-3 and the modes of the interval time period as 4-6, and obtaining the number of the events in the modes, the average wind speed value, the average wind direction value, the average air temperature value, the average air pressure value and the average duration time value/average interval time value corresponding to each basic mode according to the step 2.
S5.2, supposing that the duration time period and the interval time period of the wind power small-occurrence events in the rich period of the modes 1 and 4 occur first, then the duration time period of the wind power small-occurrence events occurs together, the modes are determined to be one of the types 1, 2 and 3, since the autocorrelation characteristics and the specific probability conditions among the modes are calculated according to the step 4, the probabilities of the modes 1, 2 and 3 occurring next and the corresponding meteorological characteristic data values are utilized to calculate the sum of the probabilities of the modes 1, 2 and 3 occurring and the meteorological data characteristic values corresponding to the modes, namely the meteorological data characteristic values corresponding to the modes occurring next, specifically comprising wind speed, wind direction, air temperature and air pressure, then the data are utilized to identify based on the predict function of the svm model, so that the specific modes belonging to the modes can be obtained, and finally, iteration is carried out to predict the modes occurring later.
Specifically, in this embodiment, based on each basic pattern after clustering, APRIORI correlation analysis is performed on an event sequence in consideration of a frequent binomial set, a frequent binomial set and a support count thereof are searched, 13 frequent binomial sets are searched in total, and a support count sum of the frequent binomial sets is 133. Assume that the probability of pattern 1 occurring first, followed by patterns 4, 5, and 6 is 53.33%, 46.67%, 0, respectively. Thus, the autocorrelation characteristics of the mode 1 and the modes 4, 5, and 6 are modeled. Similarly, in consideration of the frequent three-item set, APRIORI relevance analysis is performed on the event sequence, the frequent three-item set and the support degree thereof are searched, and after the duration time period and the interval time period of the wind power small-occurrence events in the rich period of the previous occurrence modes 1 and 4, the probabilities of occurrence of the duration time period modes 1, 2 and 3 are respectively 61%, 0 and 39% in sequence. Based on the obtained autocorrelation statistical characteristics, assuming that the modes 2 and 4 occur first, and according to the obtained autocorrelation characteristic rule of the frequent item set and the prediction method, predicting that the modes which can occur next are 2, 4, 1 and 5 in sequence.
It should be understood that parts of the specification not set forth in detail are well within the prior art.
Although specific embodiments of the present invention have been described above with reference to the accompanying drawings, it will be appreciated by those skilled in the art that these embodiments are merely illustrative, and that various changes or modifications may be made therein without departing from the principles and spirit of the invention. The scope of the invention is only limited by the appended claims.
Claims (4)
1. A prediction method of a wind power small-occurrence event statistical characteristic model in a rich period based on frequent pattern mining is characterized by comprising the following specific steps:
step 1, based on a data driving mode, screening out the occurrence time and frequency of a wind power small occurrence event in a rich period by using actual wind power data and time mark information, extracting corresponding duration time periods and interval time periods when the wind power small occurrence event occurs in each wind power plant, and performing characteristic analysis;
step 2, introducing meteorological data, selecting meteorological indexes, respectively carrying out clustering analysis on the continuous time period mode and the interval time period mode based on a K-means clustering algorithm to obtain different basic modes, simultaneously obtaining meteorological features corresponding to each basic mode, and carrying out learning training on the meteorological data under each basic mode by using a support vector machine algorithm to obtain a support vector machine classifier;
step 3, according to each basic mode obtained through clustering analysis in the step 2, event sequence recoding is carried out on the wind power sequence by using the basic mode, and dimension reduction processing of the wind power sequence is realized;
step 4, mining the event sequence through an APRIORI correlation analysis algorithm according to the event sequence after dimensionality reduction processed in the step 3 to obtain the correlation between the wind power small-occurrence events in the rich period and the events, further establishing an autocorrelation statistical characteristic model of the wind power small-occurrence events in the rich period, and completing prediction of the wind power small-occurrence events in the rich period, wherein the method specifically comprises the following steps of:
step 4.1, mining and analyzing the event sequence by an APRIORI correlation analysis algorithm according to the event sequence after dimensionality reduction obtained by processing in the step 3.2 to obtain the correlation between the wind power small-occurrence event and the event in the rich period; performing frequent item set search on a mode event sequence based on an APRIORI principle, and counting SC (mode) as a support degree, namely the frequency of occurrence of a certain mode; because the relevance of the wind power small-occurrence events at the rich period which are far away from each other is small, only three frequent item sets need to be considered, namely the wind power small-occurrence events at the rich period and the corresponding frequent item sets at intervals are sequentially contained; therefore, the autocorrelation information of the wind power small-occurrence event in the rich period can be contained from the frequent binomial set to the frequent trinomial set, the autocorrelation characteristics of the mode A and the mode B are considered from the frequent binomial set, and after the mode A occurs, the probability of the mode B occurring is as follows:
the formula establishes the autocorrelation characteristics of the mode A and the mode B, and the autocorrelation characteristics of all the modes can be calculated by a frequent item set;
4.2, setting a frequent item threshold as 2 support counts, analyzing the event sequence after dimensionality reduction based on an APRIORI algorithm, searching a frequent binomial set and a frequent trinomial set, and obtaining corresponding support counts of the frequent binomial set and the frequent trinomial set so as to calculate and obtain autocorrelation characteristics among different modes;
4.3, obtaining the self-correlation characteristics among the modes based on the statistics of the steps, further establishing a self-correlation statistical characteristic model of the wind power small-occurrence events in the rich period, wherein all the frequent binomial sets and the frequent trinomial sets can represent the self-correlation characteristics of the wind power small-occurrence events in the rich period, and the set of incidence relations among different modes forms the self-correlation statistical characteristic model;
step 5, acquiring actual wind power data and meteorological data of the deer sources, the cooperation and the sun river wind power station in 2018 years, processing the association relation between each basic mode according to the step 4.2 to obtain a frequent binomial set and a frequent trinomial set, using the obtained frequent binomial set and the frequent trinomial set to represent the autocorrelation characteristic of the wind power small-occurrence event in the wind-rich period, performing example analysis based on the characteristic, and predicting the future wind power small-occurrence event in the wind-rich period, wherein the prediction method specifically comprises the following steps:
step 5.1, obtaining the basic modes of the duration time period and the interval time period of the wind power small occurrence event in the rich period according to the clustering analysis in the step 2, numbering the modes of the duration time period as 1-3 and the modes of the interval time period as 4-6, and obtaining the number of the events in the modes, the average wind speed value, the average wind direction value, the average gas temperature value, the average gas pressure value and the average duration time value/average interval time value corresponding to each basic mode according to the step 2;
and 5.2, supposing that the duration time period and the interval time period of the wind power small-occurrence events in the rich period of the patterns 1 and 4 occur first, then the duration time periods of the wind power small-occurrence events occur together, wherein the patterns are determined to be one of the types 1, 2 and 3, since the autocorrelation characteristics and the specific probability conditions among the patterns are calculated according to the step 4.2, the probabilities of the patterns 1, 2 and 3 occurring next and the corresponding meteorological characteristic data values are utilized to calculate the sum of the probabilities of the patterns 1, 2 and 3 occurring next and the meteorological data characteristic values corresponding to the patterns, namely the meteorological data characteristic values corresponding to the patterns occurring next, specifically comprising wind speed, wind direction, air temperature and air pressure, and then the data are utilized to identify based on the predict function of the svm model, so that the specific patterns belonging to the patterns can be obtained, and finally, iteration is carried out to predict the patterns occurring later.
2. The frequent pattern mining based modeling method for wind power small-occurrence events in the rich period according to claim 1, characterized in that the implementation of step 1 comprises the following steps:
step 1.1, selecting actual wind power data of a wind power station in one year based on a data driving mode, wherein data acquisition is performed once every 15 minutes every day, and the number of the data is 96 in total every day;
step 1.2, preprocessing actual wind power data, and deleting data which are missing and have numerical values which are obviously beyond the physical significance range by adopting a deletion method to realize the processing of missing values and abnormal values;
step 1.3, based on actual wind power data obtained after preprocessing, screening the occurrence time and frequency of the wind power small occurrence events in the rich period by taking the wind power data which is lower than 5% of the maximum wind power in one year as a standard, and extracting corresponding duration time periods and interval time periods when the wind power small occurrence events in the rich period occur in each wind power plant;
and 1.4, respectively depicting the duration time and interval time probability distribution map of the wind power small events in the rich period on the time scale and the space scale according to the obtained characteristic data, and depicting the starting time probability distribution map of the wind power small events in the rich period.
3. The frequent pattern mining based modeling method for wind power small-occurrence events in the rich period according to claim 2, characterized in that the implementation of step 2 comprises the following steps:
step 2.1, introducing meteorological data, and selecting wind speed, wind direction, air temperature and air pressure as meteorological indexes;
2.2, respectively carrying out clustering analysis on the continuous time period mode and the interval time period mode based on a K-means clustering algorithm to obtain different basic modes, and simultaneously obtaining meteorological features corresponding to the basic modes; the method comprises the following specific steps of respectively clustering and analyzing historical meteorological data in a continuous time period mode and an interval time period mode by using a k-means algorithm to form each basic mode:
2.2.1, randomly selecting k meteorological data sample points from the t historical meteorological data sample points as initial clustering centers, sequentially calculating the distances from the rest sample points to the initial clustering centers, and assigning the sample points to the closest clusters so as to form initial k clusters;
step 2.2.2, respectively calculating the mean value of the sample point data in the k clusters to obtain a central sample, taking the k central samples as new clustering centers, recalculating the distance between each meteorological data sample point and the new clustering centers, and distributing each sample point to the cluster with the closest distance again according to the minimum distance principle;
step 2.2.3, the mean value of the k clusters is recalculated, and the step 2.2.2 and the step 2.2.3 are circulated until the cluster center is not changed any more;
step 2.3, learning and training meteorological data under each basic mode by using a support vector machine algorithm to obtain a support vector machine classifier; analyzing a support vector machine algorithm with a meteorological pattern class of more than 2, constructing k SVM sub-classifiers by adopting a one-by-one identification method, and when constructing the jth SVM sub-classifier, marking sample data belonging to the jth class as a positive class and marking sample data not belonging to the jth class as a negative class; during training, the discrimination function values of the sub-classifiers are respectively calculated for historical meteorological data samples, and the category corresponding to the maximum discrimination function value is selected as the category of the meteorological data samples, so that multi-classification is realized.
4. The frequent pattern mining based modeling method for wind power small-occurrence events during the rich period as claimed in claim 2, characterized in that the implementation of step 3 comprises the following steps:
3.1, analyzing the characteristics of the wind power small-occurrence events in the rich period, finding that the power output of the wind power small-occurrence events in the rich period is extremely low, the duration of the occurrence is an important characteristic, and in addition, the interval time between two wind power small-occurrence events in the rich period is also an important attention object, so that the duration and interval time parameters of the wind power small-occurrence events in the rich period are utilized to express the information of the wind power small-occurrence events in the rich period, and according to the step 2.2, the average values of the corresponding historical meteorological data in each duration time period and interval time period mode are respectively obtained by utilizing a k-means algorithm, so that each mode has a corresponding meteorological characteristic data set, and then the meteorological characteristic data sets are subjected to cluster analysis to obtain different meteorological modes, and the duration time period and interval time period modes can be classified according to different meteorological modes, so that each basic sub-mode can be obtained;
and 3.2, in order to facilitate event coding, clustering the modes A corresponding to the wind power small-occurrence event duration time period in the rich period, respectively obtaining different modes and numbering in sequence, clustering the modes B corresponding to the wind power small-occurrence event interval time period in the rich period, respectively obtaining different modes and numbering in sequence, performing event sequence recoding on the wind power sequence by using each obtained basic mode, and reducing the dimension of a multidimensional power sequence into a 2-dimensional event sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110276871.6A CN113065278B (en) | 2021-03-15 | 2021-03-15 | Frequent pattern mining-based prediction method for wind power small-occurrence event statistical characteristic model in rich period |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110276871.6A CN113065278B (en) | 2021-03-15 | 2021-03-15 | Frequent pattern mining-based prediction method for wind power small-occurrence event statistical characteristic model in rich period |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113065278A CN113065278A (en) | 2021-07-02 |
| CN113065278B true CN113065278B (en) | 2023-01-31 |
Family
ID=76560860
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110276871.6A Active CN113065278B (en) | 2021-03-15 | 2021-03-15 | Frequent pattern mining-based prediction method for wind power small-occurrence event statistical characteristic model in rich period |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113065278B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113496262B (en) * | 2021-09-08 | 2022-02-08 | 山东大学 | Data-driven active power distribution network abnormal state sensing method and system |
| CN115204322B (en) * | 2022-09-16 | 2022-11-22 | 成都新希望金融信息有限公司 | Behavior link abnormity identification method and device |
| CN116384598A (en) * | 2023-06-06 | 2023-07-04 | 中国电力科学研究院有限公司 | Wind power process prediction method and device based on time sequence event coding |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106447086B (en) * | 2016-09-07 | 2019-09-24 | 中国农业大学 | One kind being based on the pretreated wind power combination forecasting method of wind farm data |
| CN107909212B (en) * | 2017-11-20 | 2021-12-17 | 武汉大学 | Climbing event prediction method based on mesoscale wind power development process serialization |
| CN111008504B (en) * | 2019-12-18 | 2022-03-15 | 武汉大学 | Wind power prediction error modeling method based on meteorological pattern recognition |
-
2021
- 2021-03-15 CN CN202110276871.6A patent/CN113065278B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113065278A (en) | 2021-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113065278B (en) | Frequent pattern mining-based prediction method for wind power small-occurrence event statistical characteristic model in rich period | |
| US11043808B2 (en) | Method for identifying pattern of load cycle | |
| CN107766990B (en) | Method for predicting power generation power of photovoltaic power station | |
| US20230136352A1 (en) | Method and system for predicting a day-ahead wind power of wind farms | |
| CN112381181B (en) | Dynamic detection method for building energy consumption abnormity | |
| CN104598565B (en) | A kind of K mean value large-scale data clustering methods based on stochastic gradient descent algorithm | |
| CN107748940B (en) | Power-saving potential quantitative prediction method | |
| CN110781068B (en) | Data center cross-layer energy consumption prediction method based on isomorphic decomposition method | |
| CN115829105A (en) | Photovoltaic power prediction method based on historical data feature search | |
| CN108596227B (en) | Mining method for dominant influence factors of electricity consumption behaviors of users | |
| CN111859283A (en) | Scene generation method considering multi-energy charge-time sequence and correlation | |
| CN105631465A (en) | Density peak-based high-efficiency hierarchical clustering method | |
| Zhang et al. | A Density-Center-Based Automatic Clustering Algorithm for IoT Data Analysis | |
| CN114676931B (en) | Electric quantity prediction system based on data center technology | |
| CN116050666A (en) | Photovoltaic power generation power prediction method for irradiation characteristic clustering | |
| CN111768066B (en) | Park electric heating load coupling relation analysis method and device based on fusion characteristics | |
| CN111382763A (en) | Photovoltaic cluster division method and system | |
| CN113159437B (en) | Method for predicting short-term photovoltaic power generation output power | |
| Tian et al. | Research on clustering based meteorological data mining methods | |
| CN110659681A (en) | Time sequence data prediction system and method based on pattern recognition | |
| CN117094453B (en) | Scheduling optimization system and method for virtual power plant | |
| CN113792778B (en) | Wind power plant partition and output correlation coefficient calculation method and system thereof | |
| CN110932265B (en) | Static voltage stability evaluation method and system | |
| CN115618258B (en) | Method and system for extracting key operation modes of power system planning | |
| CN116845869A (en) | Clustering method, device and equipment for wind-light output typical scene set |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |