Power market power index screening method based on improved elastic network regression algorithm

Abstract

The invention relates to the technical field of power systems, in particular to a power market power index screening method based on an elastic network regression algorithm. The method comprises the following steps: step 1, acquiring market parameters, information of power generation enterprises in the market and generator set parameters; 2, establishing a market power index system of the electricity market, wherein the market power index system comprises a market structure index, an individual market power index and a collusion market power index; step 3, calculating market parameters, power generation enterprise information in the market and generator set parameters according to the market power index system of the power market established in the step 2 to obtain various market power indexes; step 4, carrying out dimensionless processing on the indexes; and 5, constructing a penalty function by taking the L1 norm and the L2 norm as penalty terms through an elastic network regression algorithm, and enabling coefficients of some indexes to become 0 in a regression process, thereby achieving an index screening effect.

Description

Technical Field

[0001] The invention relates to the technical field of power systems, and in particular to a method for screening power market power indicators based on an improved elastic network regression algorithm. Background Art

[0002] The economic definition of market power is the ability to make prices deviate from the competitive level and profit from it. According to the economic definition, market power is the ability to influence electricity prices, even if the degree of influence is only a little and the duration is very short. There are two key points in the definition: 1) The effect must be profitable; 2) The price must deviate from the competitive level. From the perspective of the existence of market power, market power can be divided into vertical market power and horizontal market power. Vertical market power refers to the participation of enterprises in multiple economic activities in related fields, using their dominant position in some fields to increase prices in related fields, thereby increasing corporate profits. Horizontal market power refers to the behavior of a member of a market transaction taking advantage of certain special conditions and advantages to obtain excess profits relative to similar competitors due to mechanisms, market conditions, etc. after the market-oriented reform of the power industry. From the perspective of the object of the exercise of market power, market power can also be divided into seller's market power and buyer's market power. Seller's market power in the power market generally refers to the ability of various power producers to use the characteristics of the market structure, their own advantages or strategic competition behavior and alliance cooperation to control or influence the market clearing price, so that it is higher than the marginal cost of power product production, so as to maintain the market monopoly position of a single manufacturer or alliance organization and obtain excess profits. The so-called buyer's market power refers to the market power exercised on the demand side, with the purpose of lowering the market price. The exerciser of buyer's market power can be an ISO under a single purchase. In order to lower the market price, the ISO can interrupt the load or reduce the output by paying a price higher than the market price; it can also be a seller under the condition of demand side participating in the bidding.

[0003] With the release of relevant supporting documents, a new round of large-scale power system reform with provinces as entities has begun. In the early stages of power market construction and power system reform, it is easy for power market members to abuse their market power, resulting in reduced power market operation efficiency and damage to the interests of some market players. The determination of market power in the power market is a complex process. At present, the determination of market power in the power market is mainly based on various evaluation indicators, but there is relatively little research on the screening method of market power indicators, and there is a lack of systematic theoretical research on the screening method of market power indicators. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method for screening electricity market power indicators based on an improved elastic network regression algorithm, which can effectively screen electricity market power indicators, provide auxiliary decision-making means for market operating organizations to monitor market power, and use the elastic network regression algorithm to solve the problem of multicollinearity. It also has more stable performance and lower computational complexity when processing high-dimensional data.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A method for screening power market power indicators based on an improved elastic network regression algorithm, the method comprising the following steps:

[0007] Step 1: Obtain market parameters, information on power generation companies in the market, and parameters of power generation units;

[0008] Step 2: Establish a market power indicator system for the electricity market, including market structure indicators, individual market power indicators, and collusion market power indicators;

[0009] Step 3: Calculate various market power indicators by using the market parameters, information of power generation enterprises in the market, and parameters of power generation units according to the market power indicator system of the power market established in step 2;

[0010] Step 4: Dimensionless processing of indicators;

[0011] Step 5: By improving the elastic network regression algorithm, taking the L1 norm and L2 norm as penalty terms, constructing a penalty function, so that the coefficients of certain indicators become 0 during the regression process, thereby achieving the effect of indicator screening.

[0012] This technical solution is further optimized. In step 1,

[0013] Market parameters include the total market size S, the total declared capacity of the market at time t Q t 、The total market demand at time t is D t , the number of historical periods T;

[0014] The information of power generation enterprises includes the total number of power generation enterprises participating in the market N, the market size S of power generation enterprise k k ;

[0015] The parameters of the generator set include the total number of generator sets participating in the market M, the declared capacity Q of generator set i at time t, and the total number of generator sets participating in the market M. i,t , the declared price P of generator set i at time t i,t 、The marginal cost MC of generator set i at time t i,t ;

[0016] This technical solution is further optimized. In step 2, the index system includes market structure index, individual market power index, and collusion market power index:

[0017] Market structure indicators include the Herfindahl-Hirschman Index HHI, the industry concentration index CR n ,

[0018]

[0019] Individual market power indicators include market share MS, residual supply coefficient RSI, required operation rate MRR, Lerner index LI, horizontal quotation rate HBR, vertical quotation rate VBR,

[0020] X ind =(MS,RSI,MRR,LI,HBR,VBR),

[0021]

[0022]

[0023] Collusion market power indicators include joint market share CMS, joint surplus supply coefficient CRSI, joint relative quotation rate CRBR, X col =(CMS,CRSI,CRBR),

[0024]

[0025] This technical solution is further optimized. In step 4,

[0026] Step 4.1: First, make the indicators consistent:

[0027] For positive indicators, the larger the indicator value, the better the evaluation result, so consistency processing is not required;

[0028] For negative indicators, the smaller the indicator value, the better the evaluation result. The specific consistency processing formula for negative indicators is:

[0029]

[0030] In the formula, is the jth index of object i after consistency processing at time t, X j,i,t is the jth index of object i before consistency processing at time t, are the maximum and minimum values ​​of the jth index respectively;

[0031] For neutral indicators, the closer the indicator value is to a specific value, the better the evaluation result is. The specific consistency processing formula for neutral indicators is:

[0032]

[0033] In the formula, is the optimal target value of the jth indicator;

[0034] For interval indicators, when the indicator value is in a specific interval, the evaluation result is the best. The specific consistency processing formula for interval indicators is:

[0035]

[0036] Where [a, b] is the optimal interval value of the j-th indicator.

[0037] The indicators after uniform processing are

[0038] Step 4.2, then standardize the indicators. The specific processing method is as follows:

[0039]

[0040] In the formula, is the jth index of object i after standardized processing at time t, are the maximum and minimum values ​​of the jth indicator after consistency processing, and c and d are constants.

[0041] The index after standardization is

[0042]

[0043] This technical solution is further optimized. In step 5, the improved elastic network regression algorithm combines the regularization methods of Lasso regression and ridge regression, and the three parameters λ, ρ and β are used to j To control the size of the penalty term, so that the coefficients of some indicators become 0 during the regression process, thereby achieving the effect of indicator screening. The specific steps are as follows:

[0044] Step 5.1: According to the indicator system, select individual market power and collusion market power as the dependent variable Y = (Y ind ,Y col ), select 9 evaluation indicators in the indicator system as independent variables The Lagrangian equation for improved elastic net regression is as follows:

[0045]

[0046] In the formula, is the average value of the jth index, Y i,t is the market power of object i at time t, is the average value of market power.j The reason is: when the absolute value of the Pearson correlation coefficient between an indicator and market power is larger, it means that the impact of the indicator on market power is more significant, then the relevant information of the indicator should be retained and a smaller penalty should be given in order to better fit the error value.

[0047] Step 5.2: In order to obtain more accurate regression results, select K-fold cross validation to determine the penalty parameters λ and ρ;

[0048] Step 5.3: After obtaining the best λ and ρ, perform regression operation and the final weight coefficient w of each indicator * Calculated by using the coordinate descent method;

[0049] Step 5.4: In order to test the degree of collinearity between the selected indicators, the variance inflation factor is introduced to test whether the indicators pass the collinearity test. The specific calculation formula is as follows:

[0050]

[0051] Where VIF j is the variance inflation factor of the jth indicator, is the correlation coefficient between the jth indicator and other selected indicators.

[0052] This technical solution is further optimized, and the final weight coefficient w of each indicator * It is calculated by using the coordinate descent method. The specific steps are as follows:

[0053] Step 5.3.1. Initialize weight coefficients

[0054] Step 5.3.2, traverse all weight coefficients, take one of the weight coefficients as a variable in turn, and fix the other weight coefficients to the results of the previous calculation, and find the optimal solution under the current conditions when there is only one weight coefficient variable. At the kth iteration, the method for updating the weight coefficient is as follows:

[0055]

[0056] Steps 5.3.3 and 5.3.2 are a complete iteration, and the change in all weight coefficients Δw = |w k -w k-1 When |<0.01 or the maximum number of iterations K is reached, the iteration ends.

[0057] Different from the prior art, the above technical solution has the following beneficial effects:

[0058] 1. The present invention can realize the effective screening of power market power indicators and provide auxiliary decision-making means for market operation institutions to monitor market power;

[0059] 2. The present invention uses an improved elastic network regression algorithm for index screening, solves the multicollinearity problem by introducing L1 and L2 regularization, and avoids the occurrence of model estimation distortion and inaccurate model parameter estimation.

[0060] 3. The present invention adopts an improved elastic network regression algorithm, which is more stable when processing high-dimensional data, less susceptible to feature selection, and has lower computational complexity, making it more efficient when processing large data sets. DETAILED DESCRIPTION

[0061] In order to explain the technical content, structural features, achieved objectives and effects of the technical solution in detail, the following is a detailed description in conjunction with specific embodiments.

[0062] A method for screening power market power indicators based on an improved elastic network regression algorithm, the method comprising the following steps:

[0063] Step 1: Obtain market parameters, information on power generation companies in the market, and parameters of power generation units.

[0064] Market parameters include the total market size S, the total declared capacity of the market at time t Q t 、The total market demand at time t is D t , the number of historical periods T;

[0065] The information of power generation enterprises includes the total number of power generation enterprises participating in the market N, the market size S of power generation enterprise k k ;

[0066] The parameters of the generator set include the total number of generator sets participating in the market M, the declared capacity Q of generator set i at time t, and the total number of generator sets participating in the market M. i,t , the declared price P of generator set i at time t i,t 、The marginal cost MC of generator set i at time t i,t .

[0067] Step 2: Establish a market power indicator system for the electricity market, including market structure indicators, individual market power indicators, and collusion market power indicators.

[0068] Market structure indicators include the Herfindahl-Hirschman Index HHI, the industry concentration index CR n ,

[0069]

[0070] Individual market power indicators include market share MS, residual supply coefficient RSI, required operation rate MRR, Lerner index LI, horizontal quotation rate HBR, vertical quotation rate VBR,

[0071] Xind =(MS,RSI,MRR,LI,HBR,VBR),

[0072]

[0073]

[0074] Collusion market power indicators include joint market share CMS, joint surplus supply coefficient CRSI, joint relative quotation rate CRBR, X col =(CMS,CRSI,CRBR),

[0075]

[0076] Step 3: Calculate various market power indicators by using the market parameters, information of power generation enterprises in the market, and parameters of power generation units according to the market power indicator system of the power market established in step 2.

[0077] Step 4: Dimensionless processing of individual market power index and collusion market power index, including the following steps:

[0078] Step 4.1: First, make the indicators consistent:

[0079] For positive indicators, the larger the indicator value, the better the evaluation result, so consistency processing is not required;

[0080] For negative indicators, the smaller the indicator value, the better the evaluation result. The specific consistency processing formula for negative indicators is:

[0081]

[0082] In the formula, is the jth index of object i after consistency processing at time t, X j,i,t is the jth index of object i before consistency processing at time t, are the maximum and minimum values ​​of the jth index respectively;

[0083] For intermediate indicators, the closer the indicator value is to the specific value, the better the evaluation result is. The specific consistency processing formula for intermediate indicators is:

[0084]

[0085] In the formula, is the optimal target value of the jth indicator;

[0086] For interval indicators, when the indicator value is in a specific interval, the evaluation result is the best. The specific consistency processing formula for interval indicators is:

[0087]

[0088] Where [a, b] is the optimal interval value of the j-th indicator.

[0089] The indicators after uniform processing are

[0090] Step 4.2, then standardize the indicators. The specific processing method is as follows:

[0091]

[0092] In the formula, is the jth index of object i after standardized processing at time t, are the maximum and minimum values ​​of the jth indicator after consistency processing, and c and d are constants.

[0093] The index after standardization is

[0094]

[0095] Step 5: By improving the elastic network regression algorithm, the L1 norm and L2 norm are used as penalty terms to construct a penalty function. and The coefficients of some indicators (i.e. the screened indicators) become 0 during the regression process, thus achieving the effect of indicator screening.

[0096] In step 5, the improved elastic network regression algorithm combines the regularization methods of Lasso regression and ridge regression through three parameters λ, ρ and β j To control the size of the penalty term, and The coefficients of some indicators in the regression process become 0, thereby achieving the effect of indicator screening. The specific steps are as follows:

[0097] Step 5.1: According to the indicator system, select individual market power and collusion market power as the dependent variable Y = (Y ind ,Y col ), select 9 evaluation indicators in the indicator system as independent variables The Lagrangian equation for improved elastic net regression is as follows:

[0098]

[0099] In the formula, is the average value of the jth index, Y i,t is the market power of object i at time t, is the average value of market power. jThe reason is: when the absolute value of the Pearson correlation coefficient between an indicator and market power is larger, it means that the impact of the indicator on market power is more significant, then the relevant information of the indicator should be retained and a smaller penalty should be given in order to better fit the error value.

[0100] Step 5.2: In order to obtain more accurate regression results, select K-fold cross validation to determine the penalty parameters λ and ρ;

[0101] Step 5.3: After obtaining the best λ and ρ, perform regression operation and the final weight coefficient w of each indicator * It is calculated by using the coordinate descent method. The specific steps are as follows:

[0102] Step 5.3.1. Initialize weight coefficients

[0103] Step 5.3.2, traverse all weight coefficients, take one of the weight coefficients as a variable in turn, and fix the other weight coefficients to the results of the previous calculation, and find the optimal solution under the current conditions when there is only one weight coefficient variable. At the kth iteration, the method for updating the weight coefficient is as follows:

[0104]

[0105] Steps 5.3.3 and 5.3.2 are a complete iteration. When the change of all weight coefficients Δw=|w k -w k-1 When |<0.01 or the maximum number of iterations K is reached, the iteration ends;

[0106] Step 5.4: In order to test the degree of collinearity between the selected indicators, the variance inflation factor is introduced to test whether the indicators pass the collinearity test. The specific calculation formula is as follows:

[0107]

[0108] Where VIF j is the variance inflation factor of the jth indicator, is the correlation coefficient between the jth indicator and other selected indicators.

[0109] In order to test the sensitivity of each indicator data to individual market power and collusion market power, an improved elastic network regression was performed on the market power of the above indicators, and a 10-fold cross-validation was performed. The final regression results are shown in the following table.

[0110] Table 1 Results of improved elastic network regression and collinearity test of market power evaluation indicators

[0111]

[0112] For individual market power, the indicator VBR i,t 、CMS i,s,t , CRSI i,s,t and CRBR i,s,t The weight coefficient of is 0. After removing these indicators, the variance inflation factor of all indicators is less than 10, thus passing the collinearity test. i,t 、RSI i,t , LI i,t 、HBR i,t and VBR i,t The weight coefficient of is 0. After removing these indicators, the variance inflation factor values ​​of all indicators are less than 10, thus passing the collinearity test.

[0113] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "include..." or "comprise..." do not exclude the existence of other elements in the process, method, article or terminal device including the elements. In addition, in this article, "greater than", "less than", "exceed" and the like are understood to exclude the number itself; "above", "below", "within" and the like are understood to include the number itself.

[0114] Although the above embodiments have been described, once those skilled in the art know the basic creative concepts, they can make additional changes and modifications to these embodiments. Therefore, the above description is only an embodiment of the present invention and does not limit the patent protection scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for screening power market power indicators based on an improved elastic network regression algorithm, characterized in that: The following steps are involved: Step 1: Obtain market parameters, information on power generation companies in the market, and parameters of power generation units; Step 2: Establish a market power indicator system for the electricity market, including market structure indicators, individual market power indicators, and collusion market power indicators; Step 3: Calculate various market power indicators by using the market parameters, information of power generation enterprises in the market, and parameters of power generation units according to the market power indicator system of the power market established in step 2; Step 4: Dimensionless processing of indicators; Step 5: By improving the elastic network regression algorithm, taking the L1 norm and L2 norm as penalty terms, constructing a penalty function, so that the coefficients of certain indicators become 0 during the regression process, thereby achieving the effect of indicator screening.

2. The method for screening power market power indicators based on elastic network regression algorithm according to claim 1, characterized in that: In step 1, the market parameters include the total market size S, the total declared capacity of the market at time t Q t 、The total market demand at time t is D t , the number of historical periods T; the information of power generation enterprises includes the total number of power generation enterprises participating in the market N, the market size of power generation enterprise k S k ; Generator parameters include the total number of generators participating in the market M, the declared capacity of generator i at time t Q i,t , the declared price P of generator set i at time t i,t 、The marginal cost MC of generator set i at time t i,t .

3. The method for screening power market power indicators based on elastic network regression algorithm according to claim 2, characterized in that: The market power index system in step 2 includes market structure index, individual market power index and collusion market power index: Market structure indicators include the Herfindahl-Hirschman Index HHI, the industry concentration index CR n , Individual market power indicators include market share MS, residual supply coefficient RSI, required operation rate MRR, Lerner index LI, horizontal quotation rate HBR, vertical quotation rate VBR, X ind =(MS,RSI,MRR,LI,HBR,VBR), Collusion market power indicators include joint market share CMS, joint surplus supply coefficient CRSI, joint relative quotation rate CRBR, X col =(CMS,CRSI,CRBR), 4. The method for screening power market power indicators based on elastic network regression algorithm according to claim 3, characterized in that: The step 4 specifically includes the following steps: Step 4.1: First, make the indicators consistent: For positive indicators, the larger the indicator value, the better the evaluation result, so consistency processing is not required; For negative indicators, the smaller the indicator value, the better the evaluation result. The specific consistency processing formula for negative indicators is: In the formula, is the jth index of object i after consistency processing at time t, X j,i,t is the jth index of object i before consistency processing at time t, are the maximum and minimum values ​​of the jth index respectively; For neutral indicators, the closer the indicator value is to a specific value, the better the evaluation result is. The specific consistency processing formula for neutral indicators is: In the formula, is the optimal target value of the jth indicator; For interval indicators, when the indicator value is in a specific interval, the evaluation result is the best. The specific consistency processing formula for interval indicators is: Where [a, b] is the optimal interval value of the jth indicator, The indicators after uniform processing are Step 4.2, then standardize the indicators. The specific processing method is as follows: In the formula, is the jth index of object i after standardized processing at time t, are the maximum and minimum values ​​of the jth index after consistency processing, c and d are constants, The index after standardization is 5. The method for screening power market power indicators based on improved elastic network regression algorithm according to claim 4, characterized in that: In step 5, the elastic network regression algorithm combines the regularization methods of Lasso regression and ridge regression, and uses three parameters λ, ρ and β j To control the size of the penalty term, so that the coefficients of some indicators become 0 during the regression process, thereby achieving the effect of indicator screening. The specific steps are as follows: Step 5.1: According to the indicator system, select individual market power and collusion market power as the dependent variable Y = (Y ind ,Y col ), select 9 evaluation indicators in the indicator system as independent variables The Lagrangian equation for improved elastic net regression is as follows: In the formula, is the average value of the jth index, Y i,t is the market power of object i at time t, is the average value of market power. j The reason is: when the absolute value of the Pearson correlation coefficient between an indicator and market power is larger, it means that the impact of the indicator on market power is more significant, then the relevant information of the indicator should be retained and a smaller penalty should be given in order to better fit the error value. Step 5.2: In order to obtain more accurate regression results, select K-fold cross validation to determine the penalty parameters λ and ρ; Step 5.3: After obtaining the best λ and ρ, perform regression operation and the final weight coefficient w of each indicator * Calculated by using the coordinate descent method; Step 5.4: In order to test the degree of collinearity between the selected indicators, the variance inflation factor is introduced to test whether the indicators pass the collinearity test. The specific calculation formula is as follows: Where VIF j is the variance inflation factor of the jth indicator, is the correlation coefficient between the jth indicator and other selected indicators.

6. The method for screening power market power indicators based on elastic network regression algorithm according to claim 5, characterized in that: The final weight coefficient w of each indicator in step 5.3 * The coordinate descent method is used for calculation. The specific steps are as follows: Step 5.3.

1. Initialize weight coefficients Step 5.3.2, traverse all weight coefficients, take one of the weight coefficients as a variable in turn, and fix the other weight coefficients to the results of the previous calculation, and find the optimal solution under the current conditions when there is only one weight coefficient variable. At the kth iteration, the method for updating the weight coefficient is as follows: Steps 5.3.3 and 5.3.2 are a complete iteration. When the change of all weight coefficients Δw = |w k -w k-1 When |<0.01 or the maximum number of iterations K is reached, the iteration ends.

Images