A method for identifying a power market capacity reservation unit in a new power system

Abstract

The application relates to a novel identification method for a capacity holding unit of a power market under a new power system, which is based on an SVM algorithm and a naive Bayes algorithm, and specifically comprises the following steps: a data collection step, a data preprocessing step, a capacity holding unit feature extraction step, an identification model establishment step, a coal-fired unit behavior risk correction step, a new energy unit behavior risk correction step and a risk analysis step; through analysis on the capacity holding unit, capacity holding unit features are extracted, and spot market data is quantitatively processed; different behavior features of the capacity holding unit are analyzed through the SVM algorithm and the naive Bayes algorithm to build a risk identification model, so that the correct rate of the identification model is improved. Compared with the prior art, the application has the advantages of high accuracy in identifying the capacity holding unit and comprehensive evaluation range.

Description

Technical Field

[0001] This invention relates to a novel method for identifying generating units with capacity holding in the electricity spot market under a new power system, belonging to the technical field of identifying generating units with capacity holding in the electricity market. Background Technology

[0002] In the electricity market under the new power system, there are numerous market participants and a complex variety of traded products. Some generating units, in pursuit of higher profits, hold onto a portion of their generating capacity, resulting in higher clearing prices in the spot market. This held capacity can also participate in lower-level markets to generate additional profits. When multiple units jointly hold capacity, or when a single unit holds a significant market share, this capacity holding behavior severely undermines the spot market's ability to discover prices. Therefore, to ensure fair competition and healthy development of the market, establishing a method for identifying units holding capacity in the spot market is crucial for maintaining the safe, stable, and reliable operation of the electricity market. However, current methods for identifying capacity-holding units often rely on manual analysis of collected unit data. This is not only labor-intensive and slow, but also prone to accuracy errors due to human bias, making it unsuitable for the complex data conditions in the spot market of the new power system. Furthermore, the new electricity market includes a large number of renewable energy units. Unlike traditional coal-fired power units, renewable energy generation is affected by weather, temperature, and other uncertainties, rendering methods for identifying capacity holding in traditional coal-fired power units inapplicable to renewable energy units. In addition, coal-fired power units in the traditional electricity market mainly participate in direct electricity trading in the electricity market, while coal-fired power units in the new electricity market participate in the electricity market and also participate in the ancillary services market to provide services such as frequency regulation and peak shaving for the system. This makes the capacity retention of coal-fired power units subject to flexibility restrictions. Therefore, the capacity retention behavior of coal-fired power units can be more accurately judged from the perspective of flexibility. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a new method for identifying capacity-holding generating units in the power spot market under a power system.

[0004] The technical solution adopted in this invention is: a novel method for identifying generating units with capacity retention in the electricity spot market under a power system, the method comprising the following steps:

[0005] Step 1) Collect electricity spot market operation data through the electricity market operation data support system. The collected electricity spot market operation data includes overall market information, unit performance parameters, market transaction information, and historical credit information.

[0006] Step 2): Based on the capacity retention risk theory, and using the spot market operation data collected in Step 1), establish a set of capacity retention identification indicators. The corresponding characteristics of the capacity retention units include high clearing prices, high bids, and high-bid rates. The set of capacity retention identification indicators includes average bids, high bid ratios, changes in high bid ratios, bid success rates, changes in bid success rates, and retention ratios.

[0007] The average bid price for a generating unit is defined as the sum of the products of the bid price and the bid capacity in each effective bid segment in the spot market, divided by the total effective capacity of the two generating units. The formula for calculating the average bid price for a generating unit is as follows:

[0008]

[0009] In the formula, p is the average quoted price of unit i at time t; i，t，h This indicates the price declared by the i-th generating unit in the h-th segment at time t; q i,t，h Y represents the capacity declared by the i-th generating unit at time t in segment h; Y represents the total number of segments declared in the unit's declaration curve; y represents the starting segment of the valid bid segment.

[0010] The high bid ratio of a generator set is defined as the proportion of the number of bids that reach the high bid level out of the total number of bids submitted by the generator set in that round. The calculation formula is as follows:

[0011]

[0012] In the formula, R high，i，t The high bid ratio of unit i at time t; NUM high，i，t NUM represents the number of times unit i's bid reaches the high bid level at time t; i，t Let be the total number of bids submitted by unit i at time t.

[0013] The rate of change of the high bid ratio of generating units is defined as the ratio of the difference between the high bid ratio of generating units in the current period and the high bid ratio of generating units in past periods to the high bid ratio of generating units at past times. The calculation formula is as follows:

[0014]

[0015] In the formula, RR high，i，t R is the rate of change of the high bid ratio of unit i in period t; high，i，t The ratio of the highest bid for unit i in cycle t; R high，i，t′ Let be the high bid ratio of unit i in period t′, where t′ is the period preceding period t.

[0016] The unit's success rate is defined as the ratio of the total electricity volume won in the bid to the total electricity volume declared by the unit. The calculation formula is as follows:

[0017]

[0018] In the formula, WR i，t Let be the success rate of unit i at time t; Unit i wins the bid for the total electricity at time t; This represents the total electricity declared by unit i at time t.

[0019] The rate of change in the unit bidding success rate is defined as the ratio of the difference between the current unit bidding success rate and the past unit bidding success rate to the past unit bidding success rate. The calculation formula is as follows:

[0020]

[0021] In the formula, WRR i，t WR represents the rate of change in the unit's successful bid rate at time t within a given period. i，t WR represents the success rate of unit i at time t. i，t′ Let be the success rate of unit i at time t′, where t′ is the time before time t. The period can be one day, one week, one month, or one year.

[0022] The generator set retention ratio is defined as the ratio of the difference between the generator set's claimable power capacity and the actual claimable power capacity to the generator set's claimable power capacity, and can be calculated using the following formula:

[0023]

[0024] In the formula, W gen,i，t This represents the retention rate of unit i at time t; The electricity that unit i can declare at time t; Let t represent the declared electricity volume of unit i at time t. The higher the unit retention ratio, the greater the likelihood that power generation companies are using market forces to limit supply and raise prices.

[0025] The indicator data is homogenized, meaning all indicators are converted into indicators where larger values ​​indicate a higher probability of unit capacity retention. Then, extreme value standardization is used to normalize all homogenized indicators. Correlation coefficients are calculated for each pair of indicators in the capacity retention discrimination indicator set, and a correlation coefficient threshold is set. Indicators with the highest correlation coefficient exceeding the threshold are selected for screening, updating the indicator set, and identifying the abnormal unit set during the risk monitoring phase.

[0026] Step 3) Utilize electricity spot market operation data and monitoring results from the risk monitoring phase to obtain behavioral characteristics of the capacity retention identification indicator set, and establish a capacity retention unit behavior identification indicator system. Perform risk identification on the capacity retention identification indicator set established in Step 2). The capacity retention unit behavioral characteristics include high bids and low success rates. The capacity retention unit behavior identification indicator system includes the unit's average bid, the unit's high bid ratio, the change rate of the unit's high bid ratio, the unit's success rate, the unit's retention ratio, and the change rate of the unit's success rate.

[0027] Step 4) Based on the SVM algorithm and the Naive Bayes algorithm, establish electricity spot market capacity retention unit identification models for the high bid and low winning bid rate characteristics of capacity retention units, and preprocess the data. This includes sequentially performing data homogenization, normalization, and correlation processing on the indicators for identifying units engaging in collusion through equal bidding, and then screening the indicators. The data normalization process uses the Min-max method to process the data form of each capacity retention unit identification indicator. The indicator screening process involves using factor analysis to calculate the correlation matrix of the equal bidding collusion unit identification indicators after data homogenization, and then using this matrix for indicator screening. Specific details are as follows:

[0028] Step 41) Analyze the risk behaviors of capacity-holding units in the new power system and extract corresponding indicators:

[0029] Step 42) Based on the SVM algorithm, identify the risks associated with the high bidding characteristics of capacity-holding units. Specifically:

[0030] Step 421) Extract relevant indicators based on the characteristics of high-priced items;

[0031] Step 422) Based on the actual operation of the electricity spot market, construct a sample of capacity-holding units and a training set for the training identification model, the expression of which is:

[0032] {(x i y i |i = 1, 2, ..., 1}

[0033] In the formula, 1 represents the number of samples in the training set, and x represents the number of samples in the training set. i The index features of the sample; x i ∈R d y i ∈{1,0} represents the recognition result.

[0034] Step 423) Train and optimize the SVM recognition model using the training set. The expression is as follows:

[0035]

[0036] Subject to: yi (ω T x i +b)≥1-δ i

[0037] δ i ≥0

[0038] In the formula, ω and b are the parameters of the hyperplane formula; δ i is a slack variable; C is a parameter controlling the degree of penalty, 1 is the number of training set samples, and x is a slack variable. i The index features of the sample; x i ∈R d y i ∈{1,0} represents the recognition result.

[0039] Step 424) Using the Lagrange duality principle, analyze ω, b, and δ in the SVM identification model respectively. i Differentiate and then rearrange to:

[0040]

[0041] C≥λ i ≥0

[0042] Step 424) Find the optimal solution λ of the recognition model. * From this, we can derive ω and b in the hyperplane formula. * The parameters are as follows:

[0043] ω * x1+b * =0

[0044] Step 426) Obtain the final SVM algorithm function, specifically:

[0045] f(X) = sign(ω) *T x i +b * ),

[0046] The abnormal score S of the capacity-holding unit under the high-price characteristic was calculated. W1 .

[0047] Step 43) Based on the Naive Bayes algorithm, risk identification is performed on the low-bid-success-rate characteristics of capacity-holding units in the electricity spot market. Specifically:

[0048] Step 431) Extract relevant indicators based on the characteristics of low winning bids;

[0049] Step 432) Based on the actual operation of the electricity spot market, construct a sample of capacity-holding units and a training set for the training identification model, the expression of which is:

[0050] {(xi y i |i = 1, 2, ..., I}

[0051] In the formula, 1 represents the number of samples in the training set, and x represents the number of samples in the training set. i The index features of the sample; x i ∈R d y i ∈{1,0} represents the recognition result.

[0052] Step 433) Calculate the mean values ​​of various indicators in the training set. and variance Calculate its probability density function as follows:

[0053]

[0054] Calculate P(X=x) i |y i The value can be derived from Bayes' theorem:

[0055]

[0056] Step 434) Based on the premise that all indicators are independent of each other, obtain the final Naive Bayes algorithm function, specifically:

[0057]

[0058] Step 435) Calculate the abnormal score Sw2 of the capacity-holding unit under the low-bid characteristics.

[0059] Step 44) Calculate the anomaly score S based on the dynamic weighting method. W The calculation formula is as follows:

[0060] S w ==W1(S W1 )S W1 +W2(S W2 )S W2

[0061]

[0062] In the formula, SW1 and SW2 are the abnormal scores under the perspectives of high bids and low success rates, respectively. w1 and w2 are the weighting coefficients of the abnormal scores SW1 and SW2, respectively. SW is the comprehensive abnormal score.

[0063] Step 45) Based on the unit's anomaly score S W Perform filtering and update the abnormal unit set G. W .

[0064] Step 5): Based on the characteristics of coal-fired power units in the spot market, correct the abnormal scores of the coal-fired power units in the abnormal unit set from Step 4), and analyze and process them to obtain the set of coal-fired power units with capacity retention. The specific content is as follows:

[0065] Step 51) Based on the abnormal score S of the unit during the risk identification phase of the inspection period. w The periodic anomaly score S of the computer group pa (This article uses a monthly average anomaly score as an example.)

[0066] Step 52) Construct inference analysis indicators based on the flexibility differences of coal-fired power units, and perform fuzzy classification on the indicators; specifically:

[0067] Step 521) Select indicators that reflect the flexibility of coal-fired power units: coal-fired power unit regulation rate and coal-fired power unit response time;

[0068] Step 522) The adjustment rate of the coal-fired power unit is related to the operating conditions and adjustment direction of the unit, and directly reflects the adjustment capacity of the coal-fired power unit. The larger the adjustment rate of the coal-fired power unit, the more flexible the coal-fired power unit is, and the higher the possibility of capacity holding behavior.

[0069] Step 523) The unit regulation capacity is the stable regulation range of the unit's output power under normal operating conditions. The larger the regulation capacity, the more flexible the unit is, and the higher the possibility of capacity holding behavior.

[0070] Step 53) Combine the indicator data of coal-fired power units in terms of flexibility to calculate the flexibility correction factor and correct the abnormal scores of coal-fired power units in the risk analysis stage.

[0071] Step 54) Analyze and process the abnormal scores of the coal-fired power units after correction to obtain the set of coal-fired power units with capacity retention in the sub-line correction stage.

[0072] Step 6): Based on the characteristics of new energy generating units in the spot market, correct the abnormal scores of the new energy generating units in the abnormal unit set from Step 5), and analyze and process them to obtain the set of new energy generating units with capacity retention. The specific content is as follows:

[0073] Step 61) Based on the abnormal score S of the unit during the risk identification phase of the inspection period. w The periodic anomaly score S of the computer group pb (This article uses a monthly average anomaly score as an example.)

[0074] Step 62) Construct inference analysis indicators based on the differences in power generation uncertainty of new energy units, and perform fuzzy classification on the indicators; specifically:

[0075] Step 621) Select indicators that reflect the uncertainty of power generation by new energy units: the accuracy rate of power generation prediction for new energy units and the deviation value of transaction execution for new energy units;

[0076] Step 622) The accuracy of power generation prediction for new energy units is related to the actual output and predicted output of new energy units, and directly reflects the unit's ability to fulfill its obligations. The higher the accuracy of power generation prediction for new energy units, the lower the uncertainty of the unit and the lower the possibility of capacity holding behavior.

[0077] Step 63) Combine the indicator data of new energy units in terms of power generation uncertainty, calculate the power generation uncertainty correction factor, and correct the abnormal scores of new energy units in the risk analysis stage.

[0078] Step 64) Analyze and process the abnormal scores of the new energy units after correction to obtain the set of new energy units with capacity retention in the sub-line correction stage.

[0079] Step 7) Merge the abnormal coal-fired power units set and the abnormal new energy power units set from Step 5) and Step 6), adjust the abnormality score again based on historical performance, and screen them to finally obtain the capacity holding group set for the risk analysis stage, thereby identifying the capacity holding units in the market.

[0080] Beneficial effects of this invention:

[0081] I. This invention takes into account the complex and diverse nature of electricity market transactions. It constructs an identification model using SVM and Naive Bayes algorithms, which has significant advantages in binary classification and probability analysis of sample data, achieving high accuracy in identifying units with capacity retention.

[0082] Second, this invention utilizes deep learning methods to obtain the corresponding characteristics of capacity-holding units in the electricity spot market data itself. Compared with traditional subjective risk assessment methods, the identification results of this invention are more objective and accurate.

[0083] Third, compared with traditional capacity retention methods, this invention uses multi-perspective identification, extracts the behavioral characteristics of capacity retention units, and adopts multi-perspective comprehensive evaluation, which can greatly increase the identification effect and improve the feasibility and accuracy of the identification results;

[0084] Fourth, this invention addresses the identification of generating units with capacity retention in the power market under the new power system. Traditional identification methods only target coal-fired power units, failing to effectively highlight the differences in capacity retention between coal-fired and renewable energy units in the power market under the new power system. This paper analyzes the commonalities between coal-fired and renewable energy units; it also conducts separate analyses for each, effectively improving the accuracy of capacity retention identification for coal-fired and renewable energy units. This provides reliable data support and a solid basis for decision-making for market regulatory agencies. Attached Figure Description

[0085] Figure 1 This is a flowchart illustrating the method for identifying colluding generator units in the electricity spot market, etc., as described in the embodiment. Detailed Implementation

[0086] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.

[0087] Example 1, such as Figure 1 As shown, this invention relates to a novel method for identifying generating units with capacity retention in the spot market under a new power system. Specifically, it includes the following steps:

[0088] Step 1) Data Collection: Collect electricity spot market data through the electricity market operation data support system. The collected information includes overall market information, basic generating unit information, market transaction information, and historical credit information.

[0089] The overall market information includes the trading rules of each market, transmission and distribution prices, relevant laws and regulations, the proportion of new energy units, and the overall electricity consumption of the market.

[0090] The basic information of the generating unit includes the group to which the unit belongs, the type of the unit, the installed power generation capacity of the unit, the frequency regulation capability of the unit, the company change information, and relevant laws and regulations.

[0091] The market transaction information includes unit declaration information, unit pre-clearing transaction results information, unit historical transaction results information, market settlement prices, and fuel supply status.

[0092] The historical credit information includes the unit's historical credit status, contract performance and power deviation execution status, unit violation records and penalty records.

[0093] Step 2) Risk Monitoring Phase Steps: Establish a set of capacity retention identification indicators. The corresponding characteristics of the capacity retention units include high clearing prices, high bids, and high bidding rates. The capacity retention identification indicator system includes average bids, high bid ratios, changes in high bid ratios, bidding success rates, changes in bidding success rates, and retention ratios.

[0094] The average bid price for a generating unit is defined as the sum of the products of the bid price and the bid capacity in each effective bid segment in the spot market, divided by the total effective capacity of the two generating units. The formula for calculating the average bid price for a generating unit is as follows:

[0095]

[0096] In the formula, p is the average quoted price of unit i at time t; i，t，h This indicates the price declared by the i-th generating unit in the h-th segment at time t; q i，t，h Y represents the capacity declared by the i-th generating unit at time t in segment h; Y represents the total number of segments declared in the unit's declaration curve; y represents the starting segment of the valid bid segment.

[0097] The high bid ratio of a generator set is defined as the proportion of the number of bids that reach the high bid level out of the total number of bids submitted by the generator set in that round. The calculation formula is as follows:

[0098]

[0099] In the formula, R high，i，t The high bid ratio of unit i at time t; NUM high，i，t NUM represents the number of times unit i's bid reaches the high bid level at time t; i，t Let be the total number of bids submitted by unit i at time t.

[0100] The rate of change of the high bid ratio of generating units is defined as the ratio of the difference between the high bid ratio of generating units in the current period and the high bid ratio of generating units in past periods to the high bid ratio of generating units at past times. The calculation formula is as follows:

[0101]

[0102] In the formula, RR high,i，t The rate of change of the high bid ratio of unit i in period t; RR high，i，t The ratio of the highest bid for unit i in cycle t; R high，i，t′ Let be the high bid ratio of unit i in period t′, where t′ is the period preceding period t.

[0103] The unit's success rate is defined as the ratio of the total electricity volume won in the bid to the total electricity volume declared by the unit. The calculation formula is as follows:

[0104]

[0105] In the formula, WR i，t Let be the success rate of unit i at time t; Unit i wins the bid for the total electricity at time t; This represents the total electricity declared by unit i at time t.

[0106] The rate of change in the unit bidding success rate is defined as the ratio of the difference between the current unit bidding success rate and the past unit bidding success rate to the past unit bidding success rate. The calculation formula is as follows:

[0107]

[0108] In the formula, WRR i，t WR represents the rate of change in the unit's successful bid rate at time t within a given period. i，t WR represents the success rate of unit i at time t. i，t′ Let be the success rate of unit i at time t', where t' is the time before time t. The period can be one day, one week, one month, or one year.

[0109] The generator set retention ratio is defined as the ratio of the difference between the generator set's claimable power capacity and the actual claimable power capacity to the generator set's claimable power capacity, and can be calculated using the following formula:

[0110]

[0111] In the formula, W gen，i，t This represents the retention rate of unit i at time t; The electricity that unit i can declare at time t; Let t represent the declared electricity volume of unit i at time t. The higher the unit retention ratio, the greater the likelihood that power generation companies are using market forces to limit supply and raise prices.

[0112] Step 3) After establishing the identification indicator library, it is necessary to process the erroneous and missing data in the electricity spot market data obtained in step S1. For erroneous data, delete the erroneous data and replace it with the mean of adjacent data. For missing data, replace it with the mean of adjacent data. Next, calculate each indicator. Then, perform dimensionless processing and standardization processing on the indicators. Finally, screen the indicators to obtain standardized data.

[0113] Data preprocessing includes dimensionless data processing, data standardization, and indicator selection methods, among which:

[0114] The purpose of the dimensionless method is to solve the problem of incomparable indicator data. It uses extreme value processing to unify the data format of each indicator. The calculation formula is as follows:

[0115]

[0116] In the formula, X i The dimensionless index; x i The index before dimensionless measurement; x av x is the mean of the indicator; s The standard deviation of the indicator is given.

[0117] The purpose of the standardization method is to address the differences in the nature of various indicators by standardizing them and transforming them into positive indicators where larger values ​​are better, thus obtaining standardized indicators. An example is shown below:

[0118] If the indicator is extremely small, in order to transform it into an extremely large indicator, X... iTo carry out extremely large-scale and standardized transformation,

[0119]

[0120] In the formula, X imax With X imin These represent the maximum and minimum values ​​of the price quote curve for all generating units, respectively.

[0121] The indicator selection method performs correlation and importance analysis on the indicators. It analyzes the correlation between indicators, and selects indicators with high correlation based on their importance to the results. The calculation method for correlation analysis is shown in the following formula:

[0122]

[0123] In the formula, r ij Let X' be the correlation coefficient between the indicators of unit i and the indicators of unit j. ni These are the consistent values ​​for all evaluation indicators of unit i. X' is the average value of all evaluation indicators for unit i after standardization; nj Let be the consistent values ​​of all evaluation indicators for unit j. is the average value of all evaluation indicators for unit j after standardization; N is the number of units being evaluated.

[0124] Set a correlation threshold. Set a threshold for the correlation coefficient. Filter from the indicators corresponding to the largest correlation coefficient that exceeds the threshold. You can remove any indicator and update the indicator set.

[0125] Based on the principle of average, the thresholds for three indicators—average bid price of generating units, the ratio of high bid prices of generating units, and the rate of change of the ratio of high bid prices of generating units—are set as the average values ​​of the data for each indicator. When the data of an indicator for a generating unit exceeds its threshold, the generating unit is designated as an abnormal unit in the risk monitoring phase and placed in the abnormal unit set GW.

[0126] Step 4) Risk Analysis Stage: Based on the SVM algorithm and the Naive Bayes algorithm, establish identification models for capacity-holding units in the electricity spot market, respectively, considering the characteristics of high bids and low success rates. Specific details are as follows:

[0127] Extracting relevant indicators based on the risk behavior characteristics of capacity-holding units:

[0128] Risk identification of high- and low-bid characteristics of capacity-holding units in the electricity spot market is performed based on the SVM algorithm. Specifically:

[0129] Based on the identification indicators of high bid characteristics, and combined with the actual operation of the electricity spot market, a sample of capacity-holding units and a training set for training the identification model are constructed, the expression of which is:

[0130] {(x i y i |i = 1, 2, ..., 1}

[0131] In the formula, 1 represents the number of samples in the training set, and x represents the number of samples in the training set. i The index features of the sample; x i ∈R d y i ∈{1,0} represents the recognition result.

[0132] The SVM recognition model is trained and optimized using the training set, and its expression is as follows:

[0133]

[0134] Subject to: y i (ω T x i +b)≥1-δ i

[0135] δ i ≥0

[0136] In the formula, δ i is a slack variable; C is a parameter that controls the degree of penalty.

[0137] Using the Lagrange duality principle, we can analyze ω, b, and δ in the SVM recognition model. i Differentiate and then rearrange to:

[0138]

[0139] C≥λ i ≥0

[0140] Find the optimal solution λ * We can derive ω in the hyperplane formula. * b * Specifically:

[0141] ω * x i +b * =0

[0142] The function to obtain the final SVM algorithm is as follows:

[0143] f(X) = sign(ω) *T x i +b * )

[0144] The abnormal score Sw1 of the capacity-holding unit under the high-price characteristic was calculated.

[0145] Risk identification is performed on the low-bid-success-rate characteristics of capacity-holding units in the electricity spot market based on the Naive Bayes algorithm, specifically:

[0146] Based on the identification index of low winning bid characteristics, and combined with the actual operation of the electricity spot market, a sample of capacity-holding units and a training set for training the identification model are constructed, the expression of which is:

[0147] {(x i y i In the formula |i=1,2,...,1}, 1 is the number of samples in the training set, x i ∈R d y i ∈{1,0} represents the recognition result.

[0148] The mean μyi and variance of various indicators in the statistical training set. The calculator's probability density function is as follows:

[0149]

[0150] Calculate P(X=x) i |y i The value can be derived from Bayes' theorem:

[0151]

[0152] Assuming that all indicators are independent of each other, the final function for the Naive Bayes algorithm is obtained as follows:

[0153]

[0154] The abnormal score Sw2 of the capacity retention unit under the low bid-winning characteristics was calculated.

[0155] The outlier score SW is calculated using the dynamic weighting method, and the formula is as follows:

[0156] S W =w1(S W1 )S W1 +W2(S W2 )S W2

[0157]

[0158] The abnormal unit set GW is updated by filtering based on the abnormal unit score SW.

[0159] Step 5) Risk Correction Phase for Coal-fired Power Units: Based on the flexibility characteristic indicators of coal-fired power units, the abnormal scores of coal-fired power units in Step 3 are corrected, thereby selecting the set of abnormal coal-fired power units for the risk correction phase. The specific process includes:

[0160] Based on the anomaly score SW of the unit and the periodic anomaly score SWA of the computer group during the risk identification phase of the inspection period (this article takes a month as an example, that is, the monthly average anomaly score);

[0161] The indicators that reflect the flexibility of coal-fired power units are: unit regulation rate and unit response time.

[0162] The unit's regulation rate is related to its operating conditions and regulation direction, directly reflecting the unit's regulation capacity. A higher regulation rate indicates greater flexibility and a higher probability of capacity hold-up behavior. Based on the specific market conditions, the upper quartile, median, and lower quartile are selected as m1, m2, and m3, respectively. The fuzzy levels for index values ​​between [0, m1], (m1, m2], (m2, m3], and (m3, +∞) are small, relatively small, relatively large, and large, respectively.

[0163] The unit's regulating capacity refers to the stable adjustment range of the unit's output power under normal operating conditions. A larger regulating capacity indicates greater unit flexibility and a higher probability of capacity hold-up behavior. Based on the specific market conditions, the upper quartile, median, and lower quartile are selected as n1, n2, and n3, respectively. The fuzzy levels of the indicator values ​​between [0, n1], (n1, n2], (n2, n3], and (n3, +∞) are small, relatively small, relatively large, and large, respectively.

[0164] Based on the flexibility index data of coal-fired power units, a flexibility correction factor is calculated to correct the abnormal scores of coal-fired power units in the risk analysis phase. Flexibility indicators include unit regulation rate and unit regulation capacity. The larger the index, the greater the flexibility of the unit, and the larger the corresponding flexibility correction factor. Therefore, the relationship between the index and the correction factor is positive. The membership function is selected as follows:

[0165]

[0166] Where Z1 is the unit's flexibility correction factor; a1 and b1 are undetermined coefficients; when the deviation of the unit's electricity price is small, relatively small, relatively large, and large respectively, x1 can be 1, 2, 3, and 4 respectively; when the deviation of the unit's revenue is small, relatively small, relatively large, and large respectively, x2 can be 1, 2, 3, and 4 respectively.

[0167] The specific algorithm for screening abnormal coal-fired power units during the risk correction phase is as follows:

[0168] By utilizing the flexibility correction factor for coal-fired power units and combining the unit anomaly score SWA during the unit identification phase, the anomaly score SPA during the risk correction phase of coal-fired power units is obtained.

[0169] Input the monthly average anomaly score SWA of the unit. If the anomaly score SWA is less than Abns2, calculate the operating motivation correction factor Z1 of the unit, update SWA = Z1SWA, and go to S432; otherwise, determine that the unit is an anomaly unit and the reasoning ends.

[0170] The SWA is evaluated again. If it is greater than Abns1 and less than Abns2, the operating condition correction factor Z2 of the unit is calculated, and SWA is updated to Z2SWA. Then, the process proceeds to S433. If SWA is greater than or equal to Abns2, the unit is determined to be an abnormal unit, and the reasoning ends. If SWA is less than or equal to Abns1, the unit is considered not to be an abnormal unit, and the reasoning ends.

[0171] The SWA is evaluated again. If the SWA is greater than Abns1 and less than Abns2, the historical performance correction factor Z3 of the unit is calculated, and the SWA is updated to Z3SWA. Then, the process is repeated. If the SWA is greater than or equal to Abns2, the unit is determined to be an abnormal unit, and the reasoning ends. If the SWA is less than or equal to Abns1, the unit is considered not to be an abnormal unit, and the reasoning ends.

[0172] If the anomaly score SWA is less than Abns2, the unit is considered not to be an anomalous unit, and the reasoning ends; otherwise, the unit is determined to be an anomalous unit, and the reasoning ends.

[0173] Step 6) Risk Correction Phase for New Energy Units: Based on the uncertainty characteristic indicators of new energy units, the abnormal scores of new energy units in Step 3 are corrected, thereby selecting the set of abnormal new energy units for the risk correction phase. The specific process includes:

[0174] Based on the abnormality score Sw of the unit during the risk identification phase of the inspection period, the periodic abnormality score SpB of the computer group (this article takes a month as an example, that is, the monthly average abnormality score);

[0175] The following indicators were selected to reflect the uncertainty of power generation by new energy units: the accuracy rate of power generation forecast for new energy units and the deviation value of transaction execution for new energy units.

[0176] The accuracy of power generation forecasting for new energy units is related to the actual output and forecasted output of the new energy units, and directly reflects the unit's ability to fulfill its obligations. The higher the accuracy of power generation forecasting for new energy units, the lower the uncertainty of the units and the lower the possibility of capacity holding behavior.

[0177] By combining the indicator data of power generation uncertainty of new energy units, and similarly with the flexibility correction factor, the power generation uncertainty correction factor is calculated to correct the abnormal scores of new energy units in the risk analysis stage.

[0178] Similar to the treatment method for coal-fired power units in the correction phase, the abnormal score SPB of the new energy units after correction is analyzed and processed to obtain the set of new energy units with capacity retention in the risk correction phase.

[0179] Step 7) Merge the abnormal coal-fired power units and abnormal new energy power units from Steps 5) and 6), adjust the abnormal scores again based on historical performance, and finally identify the units with market capacity retention.

[0180] The historical performance indicator is the overall credit situation; the better the credit rating, the better the historical performance of the unit.

[0181] Credit ratings are divided into AAA, AA, A, BBB, BB, B, CCC, CC, C, and D. Among them, AAA and AA are considered good after fuzzing, A and BBB are considered relatively good after fuzzing, BB, B, and CCC are considered relatively poor after fuzzing, and CC, C, and D are considered very poor after fuzzing.

[0182] Based on the correction factor processing method, a historical performance correction factor is obtained to correct the abnormal scores of the units and set a threshold. Finally, the set of units with capacity retention in the spot market is identified and these units are identified as capacity retention units.

[0183] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and such modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for identifying capacity-holding generating units in a new type of power system, characterized in that, Includes the following steps: 1) Collect electricity spot market operation data through the electricity market operation data support system; 2) Based on the capacity retention risk theory, a set of capacity retention identification indicators is established using collected spot market operation data; 3) Utilize electricity spot market operation data and monitoring results from the risk monitoring phase to obtain behavioral characteristics of the capacity retention identification indicator set, establish a capacity retention unit behavior identification indicator system, and conduct risk identification on the capacity retention identification indicator set established in step 2). 4) Based on the SVM algorithm and the Naive Bayes algorithm, a power spot market capacity holding unit identification model is established to identify the high bid and low winning bid rate characteristics of capacity holding units, and the data is preprocessed. 5) Based on the characteristics of coal-fired power units in the electricity spot market, the abnormal scores of the coal-fired power units in the abnormal unit set in step 4) are corrected and analyzed to obtain the set of coal-fired power units with capacity retention. 6) Based on the characteristics of new energy generating units in the spot market, the abnormal scores of the new energy generating units in the abnormal unit set in step 4) are corrected and analyzed to obtain the set of new energy generating units with capacity retention. 7) Merge the abnormal coal-fired power units and abnormal new energy power units from steps 5) and 6), and adjust the abnormal scores again based on historical performance to identify the units with capacity retention in the market. Step 4) contains the following details: Step 41) Analyze the risk behaviors of capacity-holding units in the power market under the new power system and extract corresponding indicators; Step 42) Identify the risks associated with the high-bid characteristics of capacity-holding units based on the SVM algorithm; Step 421) Extract relevant indicators based on the characteristics of high-priced items; Step 422) Construct a sample and training set for capacity-holding units; Step 423) Train and optimize the SVM recognition model by introducing slack variables and penalty parameters; Step 424) Based on the Lagrange duality principle, differentiate the recognition model; Step 425) Obtain the optimal solution of the recognition model and derive the hyperplane formula parameters; Step 426) Obtain the final SVM algorithm function and calculate the anomaly score S from the high-quote perspective. W1 ; Step 43) Based on the Naive Bayes algorithm, identify the risk of low winning bid characteristics of capacity-holding units in the electricity spot market; Step 431) Extract relevant indicators based on the characteristics of low winning bids; Step 432) Construct a sample and training set for capacity-holding units; Step 433) Calculate the mean and variance of various indicators in the training set, and calculate their probability density functions; Step 434) Based on the premise that all indicators are independent of each other, obtain the final function of the Naive Bayes algorithm; Step 435) Calculate the anomaly score S of the capacity-holding unit under the low-bid characteristics. W2 ; Step 44) Calculate the anomaly score S based on the dynamic weighting method. W； Step 45) Based on the unit's anomaly score S W Perform filtering and update the abnormal unit set G. W .

2. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 1, characterized in that, In step 1), the spot market operation data includes overall market information, unit performance parameters, market transaction information, and historical credit information.

3. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 1, characterized in that, In step 2), the corresponding characteristics of the capacity retention units include high bids and low success rates. The set of capacity retention identification indicators includes the average bid of the units, the high bid ratio of the units, the change rate of the high bid ratio of the units, the success rate of the units, the change rate of the success rate of the units, and the retention ratio of the units.

4. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 1, characterized in that, In step 3), the data preprocessing includes sequentially performing data homogenization, normalization, and correlation processing on the indicators for identifying colluding bidders, and then screening the indicators.

5. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 4, characterized in that, The data normalization process uses the Min-max method to process the data format of the identification indicators of each capacity-holding unit.

6. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 4, characterized in that, The identification index screening process is as follows: For the indicators of collusive bidding groups after data consistency processing, factor analysis is used to calculate the correlation matrix, and the indicators are then screened.

7. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 6, characterized in that, The specific content of step 423) is as follows: By introducing slack variables and penalty parameters, the SVM model is optimized and trained; its expression is as follows: Subject to:y i (ω T x i +b)≥1-δ i d i ≥0 In the formula, ω and b are the parameters of the hyperplane formula; δ i is a slack variable; C is a parameter controlling the degree of penalty, l is the number of samples in the training set, and x is a slack variable. i The index features of the sample; x i ∈R d y i ∈{1,0} represents the recognition result.

8. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 6, characterized in that, The specific content of step 424 is as follows: Using the Lagrange duality principle, we can analyze ω, b, and δ in the SVM recognition model. i Differentiate, then rearrange, and solve; the expression is: C≥λ i ≥0。 9. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 6, characterized in that, The final function expression based on the SVM algorithm is as follows: f(X)=sign(ω *T x i +b * ) In the formula, ω * b * The parameter is used in the hyperplane formula.

10. The method for identifying capacity-holding generating units in the power market under a novel power system according to claim 6, characterized in that, The specific content of step 433) is as follows: Mean values ​​of various indicators in the statistical training set and variance Calculate its probability density function, its expression is: In the formula, l is the number of samples in the training set, and x i The index features of the sample; x i ∈R d y i ∈{1,0} represents the recognition result.

11. The method for identifying capacity-holding generating units in the power market under a novel power system according to claim 6, characterized in that, The expression for the final Naive Bayes algorithm function is as follows:

12. The method for identifying capacity-holding generating units in the power market under a novel power system according to claim 6, characterized in that, The specific content of step 44) is as follows: Anomaly score S calculated using dynamic weighting method W The calculation formula is as follows: S w =w1(S W1 )S W1 +w2(S W2 )S W2 In the formula, S W1 S W2 These are the outlier scores from the perspectives of high bids and low success rates, respectively. w1 and w2 are the outlier scores S. W1 S W2 Weighting coefficient, S W These are the comprehensive abnormality scores.

13. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 1, characterized in that, Specifically, step 5) is as follows: Step 51) Based on the anomaly score Sw of the unit during the risk identification phase of the observation period, and the periodic anomaly score S of the computer group... pa ; Step 52) Construct reasoning analysis indicators based on the flexibility difference characteristics of coal-fired power units, and perform fuzzy classification on the indicators; Step 53) Combine the indicator data of coal-fired power units in terms of flexibility, calculate the flexibility correction factor, and correct the abnormal scores of coal-fired power units in the risk analysis stage. Step 54) Analyze and process the abnormal scores of the coal-fired power units after correction to obtain the set of coal-fired power units with capacity retention in the sub-line correction stage.

14. The method for identifying capacity-holding generating units in the power market under the new power system according to claim 1, characterized in that, The aforementioned abnormal score correction and analysis process specifically includes: Step 61) Based on the anomaly score Sw of the unit during the risk identification phase of the observation period, and the periodic anomaly score S of the computer group... pb ; Step 62) Construct reasoning analysis indicators based on the power generation uncertainty differences of new energy units, and perform fuzzy classification on the indicators; Step 63) Combine the indicator data of new energy units in terms of power generation uncertainty, calculate the power generation uncertainty correction factor, and correct the abnormal scores of new energy units in the risk analysis stage. Step 64) Analyze and process the abnormal scores of the new energy units after correction to obtain the set of new energy units with capacity retention in the sub-line correction stage.

Images