CN113077165B - Generator set market force abuse discrimination method - Google Patents
Generator set market force abuse discrimination method Download PDFInfo
- Publication number
- CN113077165B CN113077165B CN202110404468.7A CN202110404468A CN113077165B CN 113077165 B CN113077165 B CN 113077165B CN 202110404468 A CN202110404468 A CN 202110404468A CN 113077165 B CN113077165 B CN 113077165B
- Authority
- CN
- China
- Prior art keywords
- market
- unit
- index
- force
- abuse
- 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
- 238000012850 discrimination method Methods 0.000 title description 2
- 230000002159 abnormal effect Effects 0.000 claims abstract description 37
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 16
- 238000007637 random forest analysis Methods 0.000 claims abstract description 13
- 238000012216 screening Methods 0.000 claims abstract description 12
- 230000006399 behavior Effects 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 13
- 238000010248 power generation Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 10
- 230000007246 mechanism Effects 0.000 claims description 8
- 238000007781 pre-processing Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 4
- 230000033228 biological regulation Effects 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000012854 evaluation process Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 230000005611 electricity Effects 0.000 description 8
- 238000001514 detection method Methods 0.000 description 4
- 238000012502 risk assessment Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 239000013307 optical fiber Substances 0.000 description 3
- 230000015572 biosynthetic process Effects 0.000 description 2
- 230000002860 competitive effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000002265 prevention Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- 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
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- 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
- 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Tourism & Hospitality (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a method for judging the abuse of the market force of a generator set, which comprises the following steps: 1) Collecting market overall information, unit basic information, unit market transaction information and historical credit information of an electric power market; 2) Establishing an abuse market force discrimination index set, and carrying out feature screening on each index in the set based on a random forest algorithm to determine the index adopted by the abuse market force discrimination model; 3) Establishing a judging model aiming at the abuse market force risk of the generator set based on an isolated forest algorithm, judging the abuse market force of the generator set, and obtaining a set of potential risk sets of the abuse market force; 4) An expert system is utilized to obtain a market force risk logic deduction model, and potential risk units abusing the market force are tracked; 5) And inputting the unit generation abnormal label sample determined to abuse market force into a data system. Compared with the prior art, the invention has the advantages of improving the discrimination accuracy, simplifying the processing difficulty and the like.
Description
Technical Field
The invention relates to the technical field of electric power markets, in particular to a method for judging abuse of market forces of a generator set.
Background
In recent years, along with the promotion of electric power reform, the electric power market reform in China accelerates the promotion and makes positive and effective progress, and the first 8 electric power spot market test points all start settlement test operation. However, the property structure characteristics of the power industry in China determine that the power generation side concentration degree taking province as a unit is higher, and the phenomenon of abusing market force is easy to occur, so that corresponding risks are caused. The remarkable harmful behavior damages the competitive property and effectiveness of the market, influences the optimal configuration of the power resources, reduces the operation efficiency of the power market and even influences the reform process. Therefore, the offensive behavior of the abusive market force of the market subject is identified, the risk of the market force can be effectively prevented, the fairness and the efficiency of the market are maintained, and the market loss is reduced.
The abuse of market force in the electric power market mainly refers to the risk caused by the fact that market participants make electricity prices continuously higher or frequently and greatly fluctuate by using the market force, and the abuse can exist in various electric power commodity transactions such as electric energy, spare capacity, auxiliary service and the like, and in different market participants such as power generators, electricity selling companies, electric power users and the like. That is, in the power market, the risk of abusing market forces can exist in a plurality of transaction links, and manufacturers or groups limit the sales volume of commodities to be under the complete competitive level through some means, and maintain the price above the marginal cost price, so as to strive for huge profits and bring huge risks to the power market. Because the market force of power generation enterprises at the power generation side of the electric power market in China is large, but the construction of the electric power market in China does not take relevant measures for relieving the market force, the identification and risk prevention of power generators with stronger market force are necessary.
The research of market force abuse at home and abroad at present mainly comprises the establishment of a market force abuse index system, the theoretical deduction analysis and comprehensive evaluation of the game theory of the market force abuse, and the like. With the promotion of the electric power spot market, mass data of the electric power market show unbalanced, high-dimensional and other characteristics, and the abuse of market force shows the characteristics of multiple forms, difficult prevention, difficult supervision and large influence, thereby bringing great challenges to relevant supervision authorities of the market. In addition, in the research of abusing market forces at home and abroad at present, the conventional power market force risk assessment method is usually based on supervised learning, but due to the privacy of power transaction data, a data sample is often not provided with a label, the difficulty of obtaining an abused market force abnormal unit sample is high, and detection is required under the condition of no supervision, so that the judgment of the abused market force of the unit is more prone to the problem of abnormal point detection; in addition, the prior art only remains with an assessment of the potential for abuse market forces, but does not form a closed loop management of risk.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a method for judging the abuse of the market force of a generator set, which can judge the risk of the market force more accurately in a qualitative and quantitative combined mode from the abuse of the behavior characteristics of the market force.
The aim of the invention can be achieved by the following technical scheme:
a method for judging the abuse of the market force of a generator set comprises the following steps:
step one, acquiring market overall information, unit basic information, unit market transaction information and historical credit information related to an electric power market through an electric power market operation data support system, and establishing a database of the acquired information.
The market overall information comprises market trade implementation rules, market quotation limit rules, power transmission and distribution prices, related laws and regulations and power consumption requirements of various industries; the basic unit information comprises group information, installed capacity of a power generation enterprise, generator unit type and the number of utilization hours of power generation equipment; the unit market transaction information comprises unit declaration information, unit transaction result information, market operation information, transaction settlement information, market management information and operation summary information; the historical credit information comprises historical credit conditions, transaction contract signing and performance conditions, unit violation records and punishment records.
And step two, establishing an abusive market force judging index set according to the data obtained in the step one, carrying out feature screening on each index in the abusive market force judging index set based on a random forest algorithm, and determining the index adopted by the abusive market force judging model according to the importance ranking of the returned features.
The abuse market force discrimination index set comprises a unit status index, a quotation behavior index and a winning result index. The unit status index comprises unit market share and key provider index; the quotation behavior indexes comprise unit average quotation and unit price index; and the winning result indexes comprise winning rate of the unit and high price winning rate of the unit.
The specific content of feature screening for each index in the abuse market force discrimination index set based on the random forest algorithm is as follows: calculating contribution of each index in the abuse market force discrimination index set in the random forest, averaging the contribution values, comparing the contribution values among the indexes to obtain importance scores of the indexes, and sorting and screening the indexes according to the importance scores of the indexes.
Establishing a judging model aiming at the abuse market force risk of the generator set based on an isolated forest algorithm, judging the abuse market force of the generator set participating in the market, and obtaining a set of potential risk sets of the abuse market force, wherein the specific flow comprises the following steps:
31 Using the electric power transaction data as a training data set, preprocessing the electric power transaction data based on the misuse market force discrimination index set, obtaining the calculation result of each unit in six index dimensions, forming a vector by all index results of each unit, training according to the step of building the iTree, and building an isolated forest consisting of the iTree;
32 Traversing each iTree, calculating the average height of each iTree in the isolated forest to obtain an abnormal score of each unit, judging whether the unit belongs to the abnormal unit according to the abnormal score, and completing the judgment of the abused market force generator unit.
In step 31), the specific steps of constructing the iTree include:
311 First training a sample set x= { X for power transaction data 1 ，x 2 ，…，x n }，x i ＝(x i1 ，x i2 ，…，x im ) Data preprocessing is carried out, and +.>The method comprises the steps that a plurality of sample points are used as a training sub-sample set X', and root nodes of the iTree are placed;
312 Randomly judging index set D= { D, D from abuse market force 2 ，…，d m Selecting an index d j As a cutting dimension, randomly generating a cutting point o in the dimension, wherein the cutting point is between the maximum value and the minimum value in the data of the current node in the dimension;
313 Generating a cutting surface by the cutting point o, dividing the data sample, classifying the sample points smaller than o in the dimension into a left branch of the current node, classifying the sample points larger than or equal to o into a right branch of the current node;
314 Recursive steps 312) and 313) according to the respective index dimensions) the data space is continuously cut until the following condition is met:
(a) Only one sample point in all the child nodes can not be cut continuously;
(b)iTree to a defined height
The specific content of the step 32) is as follows:
for a unit, traversing all iTree in iForest in the evaluation process, inquiring the node position of the iTree in each iTree, calculating the average depth of the unit in the iForest, and normalizing the average depth of all sample points in the iForest; the depth of each unit on the iTree is the path length of the unit from the root node of the iTree, passing through the intermediate node of the iTree, going down along each branch until reaching the leaf node, and the path length is denoted by h (i); calculating the average depth of each unit, and obtaining a value of 0 to 1, wherein the value is the abnormal score of the unit, if the abnormal score is close to 1, the probability of misusing the market force of the unit is higher, if the abnormal score is close to 0, the probability of misusing the market force of the unit is lower, and if the abnormal score of a plurality of samples is close to 0.5, the abnormal unit with no obvious misusing market force in the whole electric power market is judged.
And fourthly, acquiring a market force risk logic deduction model by utilizing an expert system, tracking a set of potential risk units for abusing the market force, and judging the set of potential risk units for abusing the market force.
Further, the market force risk logic deduction model obtained by utilizing the expert system comprises a data acquisition module, an expert knowledge base and an inference engine, wherein the data acquisition module acquires information from the electric power market operation data support system; the expert knowledge base comprises forward reasoning rules based on external expression, reverse reasoning rules based on internal mechanism, reasoning rules based on risk discrimination and reasoning rules based on historical credit; and the inference engine processes the behaviors of the generator sets according to each inference rule in the expert knowledge base, and when the inference value of a certain generator set exceeds the risk threshold set by the expert system, the generator set is judged to be in active abuse with the market-force risk set.
And fifthly, inputting the abnormal label sample generated by the unit judged to abuse market force into a data system, and realizing closed-loop management of data. Specifically:
comprehensively considering the results of the monitoring, judging and tracking of the market force risks from the second step to the fourth step, carrying out corresponding-level early warning on the market force risks with different degrees, formulating corresponding mechanisms under different early warning levels, generating risk treatment suggestions, and inputting an abused market force risk unit as an abnormal sample into the database of the first step to realize closed-loop management of an abused market force risk judging model.
Compared with the prior art, the method for judging the abuse of the market force of the generator set provided by the invention at least has the following beneficial effects:
1) In consideration of the characteristics of more unit characteristic indexes and large data volume, the invention utilizes a random forest algorithm to conduct index screening and dynamically updates index importance scores, thereby selecting more effective indexes and further improving the accuracy of distinguishing the isolated forest algorithm;
2) The traditional power market force risk assessment method is often based on supervised learning, but due to the privacy of power transaction data, a data sample is often not provided with a label, the difficulty of obtaining an abused market force abnormal unit sample is high, and detection is required to be carried out on the premise of unsupervised condition under more conditions, so that the judgment of the abused market force of the unit is more prone to the problem of abnormal point detection; aiming at the defect, the abuse market force discrimination model based on the isolated forest algorithm can calculate the possibility of abuse market force of the generator set through the behavior characteristics of the generator set under the condition that a training set is not needed, and the difficulty of obtaining an abuse market force abnormal unit sample is greatly simplified;
3) The prior art only remains with an assessment of the potential for abuse market forces, but does not form a closed loop management of risk; aiming at the defects, the invention dynamically monitors, evaluates, tracks and disposes the power market force risk, provides a market force risk logic deduction model based on an expert system, considers the logic of risk formation and the possibility of risk existence, and can remarkably improve the credibility of a risk analysis result by adopting a mixed inference engine and an expert knowledge base, thereby providing an authoritative decision basis for a market supervisor, and inputting a final discrimination result as an abnormal sample into a database to realize closed loop control of abusing the market force risk.
Drawings
Fig. 1 is a schematic flow chart of a method for discriminating the abuse of the market force of a generator set in an embodiment.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. It will be apparent that the described embodiments are some, but not all, embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
Examples
The invention relates to a method for judging the abuse of the market force of a generator set, wherein the judging object of the method is the generator set which participates in the market transaction; the method is based on a random forest algorithm, an isolated forest algorithm and an expert system, is suitable for the conditions of large data volume and few abnormal samples of the power spot market, can realize effective risk management of the power market force, and can provide effective risk early warning and decision making basis for market supervision personnel. Specifically, as shown in fig. 1, the method includes the steps of:
s1, market operation data acquisition: and acquiring data of the electric power market through the electric power market operation data support system, wherein the data comprises market overall information, unit basic information, unit market transaction information and historical credit information, and establishing a database of the acquired information. Wherein:
the market overall information comprises market trade implementation rules, market quotation limit rules, power transmission and distribution prices, related laws and regulations and power consumption requirements of various industries;
the basic unit information comprises group information, installed capacity of a power generation enterprise, type of a generator unit and utilization hours of power generation equipment;
the unit market transaction information comprises unit declaration information, unit transaction result information, market operation information, transaction settlement information, market management information and operation summary information;
the historical credit information comprises historical credit conditions, transaction contract signing and performance conditions, unit violation records and punishment records.
S2, misuse market force monitoring: and (3) establishing an abusive market force judging index set according to the electric power market data acquired in the step (S1), carrying out feature screening on each index in the set based on a random forest algorithm, and determining the index adopted by the abusive market force judging model according to the importance ranking of the returned features.
The abuse market force discrimination index set comprises a unit status index, a quotation behavior index and a winning result index, and indexes in the set are screened by utilizing an index screening model based on random forests, wherein:
the unit status class indexes comprise unit market share and key provider indexes; the unit market share is the proportion of scalar in the unit to scalar in the total market, and the calculation formula is as follows:
in the formula, SHA i Market share for unit i;and->The bid amount of the unit i and the unit j is the bid amount of the unit j; n is the number of all units participating in the market. By calculating the market share index of the generator set, the market force of the generator set can be estimated. If the unit occupies a market share that is higher than the average level of all units in the market, the generator unit has the potential to utilize the market forces to operate the market.
The key suppliers refer to generator sets which are required to be called for meeting market demands, the key supplier indexes determine the key suppliers by judging whether the sum of the capacities of other generator sets can meet load demands after deducting target generator set capacities, and the specific calculation formula is as follows:
in the formula, OPS j A key supplier index representing the aggregate i;the declaration capacity of the unit i is represented; />The declaration capacity of the unit j to be tested is represented; d represents the power required throughout the market; OPS (optical fiber System) i Less than 1 is the key provider for that period. If the key supplier index of the unit is higher than the average level of all units in the market, the unit is indicated to be more key in the market, and the unit can be focused on in subsequent evaluation.
The quotation behavior indexes comprise unit average quotation and unit price index; the average price of the unit is calculated by dividing the sum of the products of the declared electricity prices and declared capacities of each section in the effective price quotation sections of the unit in the spot market by the effective capacity, and the calculation formula of the price of the average price quotation of the unit is as follows:
in the method, in the process of the invention,average quoted price for unit i; p is p i，h Representing the h section declaration price of the unit i; q i，h Representing the h section declaration capacity of the unit i; y represents the total number of reporting segments in a unit reporting curve; y represents the initial number of valid quotation segments; n represents the total number of units currently participating in the market. By comparing the average quotations of all the units in the market, the quotation level of the units in the market can be obtained, if the average quotation of the units is higher than the average quotation of all the units in the marketHorizontally, the unit has the possibility to handle market prices. Reporting means that a market main body takes a generator set as a unit, and reporting in the spot market comprises reporting prices and corresponding reporting capacities; the declaration curve refers to a curve formed by declaration price and declaration capacity of the generator set.
The unit price index is the sum of the declaration electricity price and the declaration capacity per unit value of the unit effective quotation section. The unit price index can reflect the high price reporting behavior of the unit, and the calculation formula of the price index is as follows:
in the formula, CPI i The price index is the price index of the unit i; h is a reporting section in a unit reporting curve; y is the total number of reporting segments in a unit reporting curve; p is p i，h Reporting electricity price for the h section in the reporting curve of the unit i;reporting capacity of an h section in a reporting curve of the unit i; y is the initial number of effective quotation segments; c (C) cost Average electricity purchasing cost for the system; d is a power exponent, can take 2 to 5, and can be specifically selected according to market conditions; q (Q) avail And deducting the effective capacity after the middle-long contract electric quantity for the rated power generation capacity of the generator set. The price index can obviously reflect the relation between the capacity and the price reported by the generator set, can effectively discover the price reporting behavior and the remaining behavior of the generator set, and has good reference value. If the price index of a certain unit is higher than the average level of all units in the market, the price index can indicate that the price is too high or the possibility of controlling the market by using physical retention exists, and the index can be used as an important reference for judging the misuse market force behaviors.
The winning result indexes comprise winning rate of the unit and high price winning rate of the unit; the winning bid rate of the unit is defined as the ratio of the winning bid total electric quantity of the unit to the declared total electric quantity, and the calculation formula is as follows:
in WR i The winning rate of the unit i is the winning rate;the total electric quantity in the unit i is marked; />The total electricity quantity is declared for the unit i. The unit bid-winning rate index can directly reflect the bid-winning condition of the unit.
The unit high price bid-winning rate is defined as the proportion of the unit to the effective bid-winning electricity quantity of the high price. The high-price bid-winning rate index reflects the matching condition of bidding strategies of the generator and self strength through the comparison of the success condition and the declaration condition of the generator, and is used for evaluating the success rate of the strategies of the generator and the market force of the generator. The calculation formula is as follows:
in the method, in the process of the invention,the bid-winning rate is high price of the unit i; />The high price is reported to the unit i, and the effective bid-winning quantity is achieved; />And reporting the high-price effective reporting electric quantity for the unit i. The definition of high prices may be set by the market supervisor itself to a high bid level based on the bid distribution, such as taking the upper quartile of all bids, or predicting market price. The definition of high price may be set by the market supervisor to a high bid level by itself according to the bid distribution. The higher the bid-winning rate of the unit, the stronger the capability of the unit to control the market price and the larger the market force. At the position ofIn the actual market, the index of other units is almost 0 except for a few units in the market leading power generation enterprises.
The index screening model based on the random forest calculates how much the index makes contribution on each tree in the random forest, averages the contribution among the indexes, finally obtains index importance scores, and sorts and screens the indexes according to the index importance scores.
The index importance score is expressed by VIM, the Gini index is expressed by GI, and it is assumed that there are m indices X 1 、X 2 、X 3 ...X m Now each index X is calculated j Gini index scoring of (a)I.e., the average amount of change in node splitting uncertainty for the jth index in all of the RF decision trees. The Gini index of node m is GI m 。
Index X j The significance at node m, i.e., the Gini index variation before and after branching at node m, is:
wherein, GI l And GI r Respectively, the Gini index of two new nodes after branching.
Index X j The importance in the ith tree is:
assuming a total of n trees in RF, then
And finally, normalizing all the obtained importance scores to obtain the importance scores of the index. And sorting all indexes according to importance, and selecting the indexes ranked in the first five as an abuse market force judging index set.
S3, judging abusive market force: and establishing a judging model aiming at the abuse market force risk of the generator set based on an isolated forest algorithm, judging the abuse market force of the generator set participating in the market, and obtaining a set of potential risk sets of the abuse market force. The method specifically comprises the following steps:
s31, preprocessing the electric power transaction data based on the abuse market force discrimination index set by using a large amount of electric power transaction data acquired in the step S1 as a training data set to obtain calculation results of each unit in each index dimension, forming a vector by all index results of each unit, and training according to the step of building the iTree to build an isolated forest consisting of the iTree;
the specific steps for constructing the iTree are as follows:
1) First training a sample set x= { X for power transaction data 1 ，x 2 ，…，x n }，x i ＝(x i1 ，x i2 ，…，x im ) Data preprocessing is performed, wherein x is i For the ith unit, each unit has m index features. Randomly select +.>The individual sample points serve as a training sub-sample set X', placing the root node of the iTree.
2) Randomly judging index set D= { D, D from abuse market force 2 ，…，d m Selecting an index d j As a cut dimension, a cut point o is randomly generated in the dimension, which is between the maximum and minimum values in the data of the current node in the dimension.
3) Generating a cutting surface by using a cutting point o, and dividing a data sample: sample points less than o in this dimension are classified as left branches of the current node, and sample points greater than or equal to o are classified as right branches of the current node.
4) Recursively steps 2) and 3) according to the respective index dimensions, the data space is continuously cut. Until the following conditions are met:
(a) Only one sample point in all the child nodes can not be cut continuously;
(b) The height of the iTree reaches a limited height
S32, traversing each unit through each iTree, calculating the average height of each iTree in an isolated forest, obtaining an abnormal score of each unit, judging whether the unit belongs to an abnormal unit according to the abnormal score, and completing the judgment of the abused market force generator unit; specifically:
for a unit, all the iTree in the iForest needs to be traversed in the evaluation process, the node position of each iTree is queried, the average depth of the unit in the iForest is calculated, and the average depth of all sample points in the iForest is normalized. Each unit is located at the depth of the unit on the tree, and starts from the root node of the tree, passes through the intermediate node of the tree, and goes down along each branch until reaching the leaf node, and the path length passed in the process is denoted by h (i). The average depth of each unit is calculated to obtain a value of 0 to 1, which is the anomaly score of the unit.
The calculation formula of the anomaly score of the unit i is as follows:
where s (i, n) is the anomaly score for unit i and E [ h (i) ] is the expectation of the average depth of sample unit i in all iTrees. c (n) is used to normalize the average depth h (i) of the samples. Wherein H (k) is a harmonic order, and can be determined by the formula H (k) =ln (k) +ζ, ζ is an euler constant, and the value is 0.5772156649.
When the anomaly score s (i, n) is closer to 1, the greater the probability of abusing market forces by the unit is, the closer to 0 the probability of abusing market forces by the unit is, and if s (i, n) of most of the samples is close to 0.5, the anomaly unit which has no obvious abuse market forces in the whole electric power market is indicated.
Based on premise assumption, for a large-scale data set, a generator set abusing market force in the electric power market is usually separated earlier because of larger differences between a plurality of index dimensions and other most of generator sets, so that the average path length of the generator set in a binary search tree is shorter, the abnormal score obtained after normalization is higher, namely the possibility of abusing market force is higher, and a threshold value is set for the abnormal score according to the overall situation of the electric power market and can be set by an expert; and if the threshold value is larger than the threshold value, the abnormal unit of the abusive market force is the normal unit, and the set of potential risk units of the abusive market force is obtained, so that the judgment of the generator set of the abusive market force is realized.
S4, misuse market force analysis step: and carrying out logic deduction on a risk formation mechanism by using an expert system, collecting evidence, and carrying out tracking analysis on a potential risk unit of the abusive market force.
The risk analysis model of the expert system consists of a data acquisition module, an expert knowledge base and an inference engine, wherein: the data acquisition module acquires information from the power market operation data support system in the step S1; the content of evidence collection is the content involved in the following reasoning rules, and the function is to further track and analyze the potential risk group obtained in the step S3. The expert knowledge base comprises forward reasoning rules based on external expression, reverse reasoning rules based on internal mechanism, reasoning rules based on risk discrimination and reasoning rules based on historical credit; the inference engine analyzes the behavior of the generator set according to each inference rule in the expert knowledge base, and when the inference value of a certain generator set exceeds the risk threshold set by the expert system, the generator set is considered to be very likely to disturb the market by using the market force, and finally the abused market force risk set is judged.
The forward reasoning rules based on the appearance include:
the average bid of the market subject is above a threshold a (a is determined by an expert), the genset may misuse market forces;
the market subject's bid rate is above a threshold b (b is determined by an expert), the genset may misuse market forces;
the market subject's high bid rate is above a threshold c (c is determined by an expert), the genset may misuse market forces;
the market share of the power plant where the market subject is located is higher than a threshold d (d is determined by an expert), the power generating set may misuse market forces;
the reverse inference rules based on the intrinsic mechanism include:
if the market concentration of the market subject is higher than the threshold value e (e is determined by an expert), the generator set may misuse market forces;
the number of groups actually controlling the abnormal market subject is lower than a threshold value f (f is determined by an expert), the generator set may misuse market force; .
The reasoning rules based on risk discriminant analysis include:
the market subject has an outlier score above a threshold value g (g is determined by an expert) in the results of the isolated forest algorithm, and the motor group may misuse market forces.
The inference rules based on historical credit analysis include:
if the number of market registration losing contracts, performance losing, market transaction losing, information disclosure losing and other actions of the market main body is higher than a threshold value h (h is determined by an expert), the motor group may misuse market force.
The inference engine analyzes the behavior of the generator set according to each inference rule in the expert knowledge base, and when the inference value of a certain generator set exceeds the risk threshold set by the expert system, the generator set is considered to be very likely to disturb the market by using the market force, and finally the abused market force risk set is judged.
S5, data closed-loop management: and giving treatment suggestions to the unit for judging the abusive market force, and inputting the treatment suggestions as an abnormal label sample into a data system to realize closed-loop management of data.
The data closed-loop management step comprehensively considers the results of the steps S2 to S4 on the monitoring, distinguishing and tracking of the market force risks, performs corresponding-level early warning on the market force risks with different degrees, formulates a corresponding mechanism under different early warning levels, gives out risk treatment suggestions, and inputs an abused market force risk unit as an abnormal sample into the database of the step S1, so that the accuracy of screening features of the subsequent steps is improved, and closed-loop management of an abused market force risk distinguishing model is realized.
While the invention has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions may be made without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (1)
1. The method for judging the abuse of the market force of the generator set is characterized by comprising the following steps of:
1) The method comprises the steps of collecting market overall information, unit basic information, unit market transaction information and historical credit information related to an electric power market through an electric power market operation data support system, and establishing a database of the collected information;
2) Establishing an abusive market force judging index set according to the data obtained in the step 1), carrying out feature screening on each index in the abusive market force judging index set based on a random forest algorithm, and determining indexes adopted by an abusive market force judging model according to the importance ranking of the returned features;
3) Establishing a judging model aiming at the abuse market force risk of the generator set based on an isolated forest algorithm, judging the abuse market force of the generator set participating in the market, and obtaining a set of potential risk sets of the abuse market force;
4) An expert system is utilized to obtain a market force risk logic deduction model, a potential risk unit set of abusing market force is tracked, and the potential risk unit set of abusing market force is judged;
5) Inputting the abnormal label sample generated by the unit judged to abuse market force into a data system, and realizing closed-loop management of data;
the market overall information comprises market trade implementation rules, market quotation limit rules, power transmission and distribution prices, related laws and regulations and power consumption requirements of various industries; the basic unit information comprises group information, installed capacity of a power generation enterprise, generator unit type and the number of utilization hours of power generation equipment; the unit market transaction information comprises unit declaration information, unit transaction result information, market operation information, transaction settlement information, market management information and operation summary information; the historical credit information comprises historical credit conditions, transaction contract signing and performance conditions, unit violation records and punishment records;
the abuse market force discrimination index set comprises a unit status index, a quotation behavior index and a winning result index;
the unit status index comprises unit market share and key provider index; the quotation behavior indexes comprise unit average quotation and unit price index; the winning result indexes comprise winning rate of the unit and high price winning rate of the unit;
the specific content of feature screening for each index in the abuse market force discrimination index set based on the random forest algorithm is as follows:
calculating contribution of each index in the abuse market force discrimination index set in a random forest, averaging each contribution value, comparing the magnitudes of the contribution values among the indexes to obtain importance scores of the indexes, and sorting and screening the indexes according to the importance scores of the indexes;
the specific flow of the step 3) comprises the following steps:
31 Using the electric power transaction data as a training data set, preprocessing the electric power transaction data based on the misuse market force discrimination index set, obtaining the calculation result of each unit in six index dimensions, forming a vector by all index results of each unit, training according to the step of building the iTree, and building an isolated forest consisting of the iTree;
32 Traversing each iTree, calculating the average height of each iTree in the isolated forest to obtain an abnormal score of each unit, judging whether the unit belongs to the abnormal unit according to the abnormal score, and completing the judgment of the abused market force generator unit;
in step 31), the specific steps for constructing the iTree include:
311 First training a sample set for power transaction datax i ＝(x i1 ,x i2 ,…,x im ) Data preprocessing is carried out, and +.>The sample points are used as training subsampled set X ′ Putting the root node of the iTree;
312 Randomly judging index set D= { D, D from abuse market force 2 ,…,d m Selecting an index d j As a cutting dimension, randomly generating a cutting point o in the dimension, wherein the cutting point is between the maximum value and the minimum value in the data of the current node in the dimension;
313 Generating a cutting surface by the cutting point o, dividing the data sample, classifying the sample points smaller than o in the dimension into a left branch of the current node, classifying the sample points larger than or equal to o into a right branch of the current node;
314 Recursive steps 312) and 313) according to the respective index dimensions) the data space is continuously cut until the following condition is met:
(a) Only one sample point in all the child nodes can not be cut continuously;
(b) The height of the iTree reaches a limited height
The specific content of the step 32) is as follows:
for a unit, traversing all iTree in iForest in the evaluation process, inquiring the node position of the iTree in each iTree, calculating the average depth of the unit in the iForest, and normalizing the average depth of all sample points in the iForest; the depth of each unit on the iTree is the path length of the unit from the root node of the iTree, passing through the intermediate node of the iTree, going down along each branch until reaching the leaf node, and the path length is denoted by h (i); calculating the average depth of each unit, and obtaining a value from 0 to 1, wherein the value is an abnormal score of the unit, if the abnormal score is close to 1, the probability of misusing the market force of the unit is higher, if the abnormal score is close to 0, the probability of misusing the market force of the unit is lower, and if the abnormal score of a plurality of samples is close to 0.5, the abnormal unit without obvious misusing the market force of the whole electric power market is judged;
in the step 4), the market force risk logic deduction model obtained by utilizing the expert system comprises a data acquisition module, an expert knowledge base and an inference engine, wherein the data acquisition module acquires information from the electric power market operation data support system; the expert knowledge base comprises forward reasoning rules based on external expression, reverse reasoning rules based on internal mechanism, reasoning rules based on risk discrimination and reasoning rules based on historical credit; the inference engine processes the behaviors of the generator sets according to each inference rule in the expert knowledge base, and when the inference value of a certain generator set exceeds a risk threshold set by an expert system, the generator set is judged to be in active abuse with the market-force risk set;
the specific content of the step 5) is as follows:
comprehensively considering the results of the monitoring, judging and tracking of the market force risks in the steps 2) to 4), carrying out corresponding-level early warning on the market force risks in different degrees, formulating corresponding mechanisms under different early warning levels, generating risk treatment suggestions, and inputting an abused market force risk unit as an abnormal sample into the database in the step 1), so as to realize closed-loop management of an abused market force risk judging model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110404468.7A CN113077165B (en) | 2021-04-15 | 2021-04-15 | Generator set market force abuse discrimination method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110404468.7A CN113077165B (en) | 2021-04-15 | 2021-04-15 | Generator set market force abuse discrimination method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113077165A CN113077165A (en) | 2021-07-06 |
| CN113077165B true CN113077165B (en) | 2024-03-26 |
Family
ID=76617975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110404468.7A Active CN113077165B (en) | 2021-04-15 | 2021-04-15 | Generator set market force abuse discrimination method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113077165B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114819479A (en) * | 2022-03-11 | 2022-07-29 | 上海电力大学 | Behavior classification method based on index system, economic persistence identification method and device |
| CN114611616B (en) * | 2022-03-16 | 2023-02-07 | 吕少岚 | Unmanned aerial vehicle intelligent fault detection method and system based on integrated isolated forest |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016141739A1 (en) * | 2015-03-10 | 2016-09-15 | 华中电网有限公司 | Dynamic clustering-based method of establishing supply-and-demand early warning model in electricity market |
| CN111461774A (en) * | 2020-03-27 | 2020-07-28 | 南方电网科学研究院有限责任公司 | Power transmission network management system considering market power |
| CN111797892A (en) * | 2020-05-21 | 2020-10-20 | 国电南瑞科技股份有限公司 | Electric power market generator market force monitoring method based on random forest regression |
| CN112258341A (en) * | 2020-10-20 | 2021-01-22 | 广东电力交易中心有限责任公司 | Electric power market force risk monitoring and preventing method |
| CN112365187A (en) * | 2020-11-30 | 2021-02-12 | 上海电力大学 | Generating set market force abuse identification method based on Lasso-logit model |
| CN112434967A (en) * | 2020-12-09 | 2021-03-02 | 广东电力交易中心有限责任公司 | Electric power market online monitoring method and system for market force control |
| CN112529458A (en) * | 2020-12-22 | 2021-03-19 | 广东电力交易中心有限责任公司 | Closed-loop control method for electric power market performance risk |
-
2021
- 2021-04-15 CN CN202110404468.7A patent/CN113077165B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016141739A1 (en) * | 2015-03-10 | 2016-09-15 | 华中电网有限公司 | Dynamic clustering-based method of establishing supply-and-demand early warning model in electricity market |
| CN111461774A (en) * | 2020-03-27 | 2020-07-28 | 南方电网科学研究院有限责任公司 | Power transmission network management system considering market power |
| CN111797892A (en) * | 2020-05-21 | 2020-10-20 | 国电南瑞科技股份有限公司 | Electric power market generator market force monitoring method based on random forest regression |
| CN112258341A (en) * | 2020-10-20 | 2021-01-22 | 广东电力交易中心有限责任公司 | Electric power market force risk monitoring and preventing method |
| CN112365187A (en) * | 2020-11-30 | 2021-02-12 | 上海电力大学 | Generating set market force abuse identification method based on Lasso-logit model |
| CN112434967A (en) * | 2020-12-09 | 2021-03-02 | 广东电力交易中心有限责任公司 | Electric power market online monitoring method and system for market force control |
| CN112529458A (en) * | 2020-12-22 | 2021-03-19 | 广东电力交易中心有限责任公司 | Closed-loop control method for electric power market performance risk |
Non-Patent Citations (7)
| Title |
|---|
| 徐昊亮 ; 程紫运 ; 张海生 ; 董礼 ; 华回春 ; .基于改进支持向量机的发电企业滥用市场力违规识别.华北电力大学学报(自然科学版).2020,(04),全文. * |
| 徐昊亮 ; 程紫运 ; 张海生 ; 董礼 ; 华回春.基于改进支持向量机的发电企业滥用市场力违规识别.华北电力大学学报.2020,第47卷(第4期),全文. * |
| 李瑞庆,刘敦楠,何光宇,王治华,郭家春,赵岩,陈雪青.电力市场运营监管信息系统.电力系统自动化.2005,(14),全文. * |
| 甘倍瑜 ; 王健 ; 田琳 ; 徐立新 ; 季天瑶 ; 荆朝霞.美国电力市场违约风险量化机制分析.电网技术.2020,第44卷(第6期),全文. * |
| 董礼 ; 王胜华 ; 华回春 ; 郭海朝.中国现货电力市场中发电企业滥用市场力违规识别.中国电机工程学报.2021,第41卷(第24期),全文. * |
| 谢敬东 ; 陆池鑫 ; 鲁思薇 ; 孙波 ; 黄溪滢 ; 孙欣.基于序关系-熵权法的电力市场风险评估.中国电力.2020,第54卷(第6期),全文. * |
| 陈青 ; 杨骏伟 ; 黄远明 ; 卢恩 ; 王一.国外电力市场中市场力监测与缓解机制综述.南方电网技术.2018,第12卷(第12期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113077165A (en) | 2021-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113077165B (en) | Generator set market force abuse discrimination method | |
| CN111614491B (en) | Power monitoring system oriented safety situation assessment index selection method and system | |
| CN107506905A (en) | A kind of improved Sustainable Development of Power Grid Company integrated evaluating method | |
| JP2003535387A (en) | Rapid evaluation of asset portfolios such as financial products | |
| CN111401600A (en) | Enterprise credit risk evaluation method and system based on incidence relation | |
| CN109934371A (en) | The method that solvency risk identification and prediction are carried out to enterprise based on electricity consumption data | |
| CN107609771A (en) | A kind of supplier's value assessment method | |
| CN113177839A (en) | Credit risk assessment method, device, storage medium and equipment | |
| US8463678B2 (en) | Generating method for transaction models with indicators for option | |
| CN112258341A (en) | Electric power market force risk monitoring and preventing method | |
| CN113642922A (en) | Small and medium-sized micro enterprise credit evaluation method and device | |
| CN112785060A (en) | Lean operation and maintenance level optimization method for power distribution network | |
| Gontijo et al. | Electricity price forecasting on electricity spot market: A case study based on the Brazilian Difference Settlement Price | |
| Fu et al. | Trimming outliers using trees: winning solution of the large-scale energy anomaly detection (LEAD) competition | |
| CN114219225A (en) | Power grid investment benefit evaluation system and evaluation method based on multi-source data | |
| CN112365187A (en) | Generating set market force abuse identification method based on Lasso-logit model | |
| CN116956702A (en) | Electricity safety early warning method, medium and system | |
| CN116681333A (en) | Power consumer credit evaluation method based on energy big data | |
| CN115760400A (en) | Mining behavior detection method based on electric power data and storage medium | |
| CN115146735A (en) | User power utilization anomaly identification | |
| CN115034618A (en) | Community comprehensive energy system benefit evaluation method based on fuzzy evaluation | |
| Sun et al. | Identification of Generating Units That Abuse Market Power in Electricity Spot Market Based on AdaBoost‐DT Algorithm | |
| Zhang | Applications of the decision tree in business field | |
| Xing et al. | Research on the credit evaluation model of electricity selling company based on GA-SVM | |
| Li et al. | Research on listed companies’ credit ratings, considering classification performance and interpretability |
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 |