Power system peak regulation demand and influence factor evaluation method considering external power transmission access

Abstract

The invention provides a power system peak regulation demand and influence factor evaluation method considering external power transmission access. The method comprises the following steps: acquiring power system parameters; constructing a unit commitment scheduling optimization model considering the peak regulation demand slack amount, and determining constraint conditions of the unit commitment scheduling optimization model; solving the unit combination scheduling optimization model to obtain the peak regulation demand relaxation amount of the power system; constructing a peak regulation demand influence factor sample set, inputting the peak regulation demand influence factor sample set and the peak regulation demand slack amount into an XGBoost algorithm model for processing, and determining the importance of each influence factor; the importance of the peak regulation demand influence factors is arranged in a descending order, the slack amount of the peak regulation demand of the power system can be accurately evaluated under the condition of a large amount of external power transmission, and the factors under the current slack amount can be determined as the main influence factors.

Description

Technical Field

[0001] The present invention relates to a power evaluation method, and in particular to a method for evaluating peak load regulation demand and influencing factors of a power system considering access of external power transmission. Background Art

[0002] Using inter-regional power transmission as a means to achieve wide-area optimal allocation of power resources has become an important feature of modern power systems. Due to the large differences in power distribution in my country, cross-regional and cross-provincial balance of power is a basic national condition of my country. With the access of large-scale external power transmission, the net load of the regional power grid can be expressed as follows:

[0003] P pure =P load -P new -P out , where Ppure represents the net load power value of the power system; Pload represents the original load power of the power system; Pnew represents the predicted output of new energy in the power system; and Pout represents the external power connected to the power system. From the above formula, it can be seen that: in the case of a large number of new energy sources and external power connected, the accuracy of the external power curve, the uncertainty of new energy, and the large peak-to-valley difference of the load itself will affect the net load curve of the power system, which may exceed the adjustment range of the system flexibility resources, thereby causing safety and economic problems in the operation of the power system, as shown in the attached figure of the specification Figure 1 As shown in the figure, the net load demand of the power system exceeds the adjustment range of flexible resources, and the excess part (the shaded dashed part in the figure) is the peak-shaving demand of the system. When the net load is greater than the upper limit of the adjustment of flexible resources, the power system is in short supply, and there will be additional costs such as load shedding or safety issues such as line power overload; and when the net load is lower than the lower limit of the adjustment of flexible resources, the power system is in oversupply, which will cause additional costs such as wind and solar power abandonment. Accurately quantifying the value of this "mismatch" can accurately quantify the peak-shaving demand of the power system under different external power transmission, new energy and load conditions.

[0004] There are many factors that cause the mismatch between the peak-shaving demand and capacity of the power system. For example, the peak-to-valley difference rate, maximum load, minimum load, climbing rate and other curve characteristics of the load curve, new energy forecast curve or external power transmission curve, as well as the power value, may all be factors that cause the peak-shaving demand of the power system. Therefore, how to accurately quantify the relaxation of the peak-shaving demand of the power system has become a technical difficulty. Moreover, it is also a difficulty to determine which main factors cause the size of the relaxation.

[0005] Therefore, in order to solve the above technical problems, it is urgent to propose a new technical means. Summary of the invention

[0006] In view of this, the purpose of the present invention is to provide a method for evaluating the peak-shaving demand and influencing factors of a power system taking into account the access of external power transmission, which can accurately evaluate the slack of the peak-shaving demand of the power system under the condition of a large amount of external power transmission, and can determine which factors are the main influencing factors under the current slack, thereby providing accurate data support for power dispatching and control.

[0007] The present invention provides a method for evaluating the peak load regulation demand and influencing factors of a power system considering the access of external power transmission, comprising the following steps:

[0008] S1. Obtain power system parameters;

[0009] S2. Construct a unit combination scheduling optimization model that takes into account the relaxation of peak demand and determine the constraints of the unit combination scheduling optimization model;

[0010] S3. Solve the unit combination dispatch optimization model to obtain the peak load demand relaxation of the power system;

[0011] S4. Construct a sample set of factors affecting peak demand, and input the sample set of factors affecting peak demand and the slack amount of peak demand into the XGBoost algorithm model for processing to determine the importance of each factor;

[0012] S5. Arrange the factors affecting peak demand in descending order of importance.

[0013] Furthermore, the objective function of the unit commitment scheduling optimization model is:

[0014]

[0015] Where: P g,t is the output of unit g at time t, is the startup cost of unit g, represents the shutdown cost of unit g, C g represents the output cost of unit g, G represents the set of units, T represents the set of scheduling times, ε p is the output relaxation variable of unit g, ε e is the ramp slack variable of unit g, M is the penalty coefficient of the slack variable;

[0016] The constraints of the unit combination scheduling optimization model include relaxed unit output constraints, relaxed ramp constraints, unit start and stop time limit constraints, start and stop cost constraints, and power balance constraints.

[0017] Furthermore, the relaxed unit output constraint is:

[0018]

[0019] Among them: Ug,t Indicates the start and stop status of unit g at time t; P g,max Indicates the upper limit of the output of unit g, P g,min Indicates the lower limit of the output of unit g, represents the lower limit slack variable of unit g at time t, Represents the upper limit slack variable of unit g at time t.

[0020] Furthermore, the relaxed climbing constraint is:

[0021]

[0022] η g,t =U g,t -U g,t-1 ,g∈G,t∈T, (1)

[0023] Where: ΔP g,max.up and ΔP g,max.down They represent the ramp-up rate and ramp-down rate of unit g respectively;

[0024] In ΔP g,max.up Adding uphill slack variables And in ΔP g,max.down Add downhill slack variables to Modify formula (1) as follows:

[0025]

[0026] Introducing auxiliary variables and auxiliary variables Rewrite (2) as:

[0027]

[0028] η g,t =U g,t -U g,t-1 ,g∈G,t∈T,

[0029] and:

[0030]

[0031] Among them: H is a large constant.

[0032] Furthermore, the unit start and stop time limit constraint is:

[0033]

[0034] Where: T g,U and T g,D Indicates the minimum start and stop time of unit g; t g,t,onand t g,t,off They respectively represent the continuous startup and shutdown time of unit g at time t.

[0035] Furthermore, the start-stop cost constraint is:

[0036]

[0037] Where: K g and J g Represents the startup and shutdown costs of unit g.

[0038] Furthermore, the power balance constraint is:

[0039] ∑ g∈G,t∈T P g,t =∑ d∈D,t∈T V d,t ,

[0040] Where: V d,t represents the load value of node d at time t; D is the load set.

[0041] Furthermore, the sample set of factors affecting peak demand includes the first type of factors and the second type of factors, among which:

[0042] The first type of influencing factor eigenvector for:

[0043]

[0044] in: and Represents the eigenvectors Subset, as well as They represent the power values ​​of load, new energy and external power at time t, respectively. T Indicates the total number of scheduling time periods;

[0045] The second type of influencing factor eigenvector for:

[0046]

[0047] in: and express A subset of pe ,x va ,x ga and x ra They respectively represent the peak value, valley value, peak-to-valley difference rate and maximum climbing rate of the load, new energy and external power transmission curves.

[0048] Further, the importance of the influencing factors is determined as follows:

[0049] k m =N m / N M ,

[0050] Where: k m Indicates the importance of feature m, N m N is the number of splits used for feature m during the XGBoost fitting process; M is the total number of splits during the XGBoost fitting process.

[0051] Furthermore, the unit commitment scheduling optimization model is solved using the GUROBI solver.

[0052] Beneficial effects of the present invention: Through the present invention, the slack of the peak-shaving demand of the power system can be accurately evaluated in the case of a large amount of external power transmission, and the factors that are the main influencing factors under the current slack can be determined, thereby providing accurate data support for power dispatching and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0054] Figure 1 Schematic diagram of the mismatch between power system peak load demand and capacity.

[0055] Figure 2 It is a schematic diagram of the process of the present invention.

[0056] Figure 3 This is a curve diagram of power transmission in the present invention.

[0057] Figure 4 Schematic diagram of data-driven data fitting for XGBoost.

[0058] Figure 5 It is a quantitative diagram of the importance of factors affecting peak load demand when the input feature is curve power.

[0059] Figure 6 It is a quantitative diagram of the importance of factors affecting peak demand when the input feature is a curve feature. DETAILED DESCRIPTION

[0060] The present invention is further described in detail below:

[0061] The present invention provides a method for evaluating the peak load regulation demand and influencing factors of a power system considering the access of external power transmission, comprising the following steps:

[0062] S1. Obtain power system parameters;

[0063] S2. Construct a unit combination scheduling optimization model that takes into account the relaxation of peak demand and determine the constraints of the unit combination scheduling optimization model;

[0064] S3. Solve the unit combination dispatch optimization model to obtain the peak load demand relaxation of the power system;

[0065] S4. Construct a sample set of factors affecting peak demand, and input the sample set of factors affecting peak demand and the slack amount of peak demand into the XGBoost algorithm model for processing to determine the importance of each factor;

[0066] S5. Arrange the factors affecting the peak-shaving demand in descending order of importance. Through the above method, the relaxation of the peak-shaving demand of the power system can be accurately evaluated in the case of a large amount of external power transmission, and the factors that are the main influencing factors under the current relaxation can be determined, thereby providing accurate data support for power dispatching and control.

[0067] In this embodiment, the objective function of the unit commitment scheduling optimization model is:

[0068]

[0069] Where: P g,t is the output of unit g at time t, is the startup cost of unit g, represents the shutdown cost of unit g, C g represents the output cost of unit g, G represents the set of units, T represents the set of scheduling times, ε p are the output slack variables of unit g, i.e., the two slack variables in the output constraint, ε e are the climbing slack variables of unit g, i.e., the two slack variables in the climbing constraint, and M is the penalty coefficient of the slack variable;

[0070] The constraints of the unit combination scheduling optimization model include relaxed unit output constraints, relaxed ramp constraints, unit start and stop time limit constraints, start and stop cost constraints, and power balance constraints.

[0071] Specifically: the relaxed unit output constraint is:

[0072]

[0073] Among them: U g,t Indicates the start and stop status of unit g at time t; P g,max Indicates the upper limit of the output of unit g, P g,min Indicates the lower limit of the output of unit g, represents the lower limit slack variable of unit g at time t, Represents the upper limit slack variable of unit g at time t.

[0074] The relaxed climbing constraint is:

[0075]

[0076] η g,t =U g,t -U g,t-1 ,g∈G,t∈T, (1)

[0077] Where: ΔP g,max.up and ΔP g,max.down They represent the ramp-up rate and ramp-down rate of unit g respectively;

[0078] In ΔP g,max.up Adding uphill slack variables And in ΔP g,max.down Add downhill slack variables to Modify formula (1) as follows:

[0079]

[0080] Introducing auxiliary variables and auxiliary variables The process of introducing auxiliary variables is to linearize equation (2) to facilitate subsequent solution. Equation (2) can be rewritten as:

[0081]

[0082] η g,t =U g,t -U g,t-1 ,g∈G,t∈T,

[0083] and:

[0084]

[0085] Among them: H is a large constant, that is, a very large constant that is set.

[0086] The start and stop time constraints of the unit are:

[0087]

[0088] Where: T g,U and T g,D Indicates the minimum start and stop time of unit g; t g,t,on and t g,t,off They respectively represent the continuous startup and shutdown time of unit g at time t.

[0089] The start-stop cost constraint is:

[0090]

[0091] Where: K g and J g Represents the startup and shutdown costs of unit g.

[0092] The power balance constraint is:

[0093] ∑ g∈G,t∈T P g,t =∑ d∈D,t∈T V d,t ,

[0094] Where: V d,t represents the load value of node d at time t; D is the load set. When solving, the GUROBI solver is used for solving. Of course, other solvers can also be used for solving.

[0095] In this embodiment, the sample set of factors affecting peak demand includes first-category factors and second-category factors, where:

[0096] The first type of influencing factor eigenvector for:

[0097]

[0098] in: and Represents the eigenvectors Subset, as well as They represent the power values ​​of load, new energy and external power at time t, respectively. T Indicates the total number of scheduling time periods;

[0099] The second type of influencing factor eigenvector for:

[0100]

[0101] in: and express A subset of pe ,x va ,x ga and x ra They respectively represent the peak value, valley value, peak-to-valley difference rate and maximum climbing rate of the load, new energy and external power transmission curves.

[0102] The importance of the influencing factors is determined as follows:

[0103] k m =N m / N M ,

[0104] Where: k m Indicates the importance of feature m, and its value range is (0,1); N m N is the number of splits used for feature m during the XGBoost fitting process; M is the total number of splits during the XGBoost fitting process.

[0105] Among them: The XGBoost algorithm model is an existing model that estimates the target value by building a series of decision trees and assigning quantitative weights to each leaf node. Specifically:

[0106] Its prediction function is shown in formula (3):

[0107]

[0108] Among them, N t is the number of decision trees in the XGBoost model; X i is the i-th input sample; f nt represents the predicted value of the nt-th tree; represents the predicted value of the i-th sample; Γ is the set of decision trees. The objective function of the algorithm is shown as follows:

[0109]

[0110] Where: K is the number of samples; l is the loss function; y i represents the true value of the i-th sample; Ω is the regularization term, which is used to prevent the model from overfitting; γ is the regularization term coefficient; T is the number of leaf nodes in each tree; λ is the penalty coefficient; and w is the weight of each leaf.

[0111] Since the tree model generated by each iteration of the XGBoost model is obtained by fitting the prediction residual of the model in the previous iteration, the prediction function after t iterations is shown in formula (6), and the objective function is shown in formula (7). The loss function in formula (7) is expanded twice by Taylor to obtain formula (8), where p i ,q i are the first-order derivative and the second-order derivative of the loss function l, respectively, as shown in formula (9). After removing the constant in formula (8), formula (8) can be simplified to formula (10).

[0112]

[0113] Formula (10) is the sum of all sample loss values. Since all samples will eventually be attributed to a leaf node, all samples of the same leaf node are combined. The final objective function is shown in Formula (11), and the optimal solution is obtained by using the maximum value method.

[0114]

[0115] Through the above steps, the weight w of the leaf node can be obtained, and the optimal tree structure can be obtained by minimizing the objective function obj.

[0116] The following is a specific example to further illustrate:

[0117] The unit commitment scheduling optimization model in the present invention is referred to as the UC model, and is solved using the commercial solver GUROBI, version 9.5.1;

[0118] The present invention selects the daily load and new energy forecast power curve of Chongqing power grid from 2018 to 2022 (a total of 1460 days), and the transmission curve of Xinjiang and Chongqing is selected. The curve is as follows Figure 3 Taking the example shown in the figure, the power value of each curve and the curve features such as peak value, valley value, peak-to-valley difference rate, and maximum climbing rate are extracted respectively, and the input feature sample set of influencing factors is constructed, where the input feature sample set of each curve feature is 1460×12 dimensions, and the input feature sample set of the curve power value is 1460×72 dimensions;

[0119] The test equipment for the example is based on a computer with the model "AMD Ryzen 55600X 6-Core Processor CPU, 32GB of RAM", and the solver uses GUROBI 9.5.1;

[0120] Input feature is curve power value

[0121] The quantification results of the input and output importance fitted by XGBoost are as follows Figure 5 And as shown in Table 1.

[0122] Table 1 Quantitative analysis results of factors affecting power curve on peak load demand

[0123] Input Features Load curve power New energy curve power Power transmission curve The sum of importance ratios 99.6％ 0.3％ 0.1％

[0124] Figure 3 Table 1 shows the ranking results of the importance of influencing factors when the first type of influencing factors, namely the load curve power, is used as the input feature. Figure 3 It can be seen that in the XGBoost fitting process, the importance of load (number of splits) accounts for the largest proportion, so it has the greatest impact on peak demand. In Table 1, the importance of load accounts for 99.6%, which is much larger than the 0.3% of the new energy curve and the 0.1% of the external power curve. This is also consistent with the actual situation. In reality, the proportion of new energy in Chongqing power grid is not high, and the external power transmission is stable, which is in line with Chongqing's actual resource endowment and load demand.

[0125] In addition, it is not difficult to see that in the proportion of load importance, the time periods with high importance are mainly concentrated in 12h-18h every day. This is also in line with reality, because this is usually the peak period of load usage in Chongqing, and the probability of peak-shaving demand is very high.

[0126] 3) Input feature is curve feature

[0127] The quantification results of the input and output importance fitted by XGBoost are as follows Figure 6 As shown in Tables 2 to 5.

[0128]

[0129] Table 2 Quantitative analysis results of factors affecting load curve characteristics on peak load demand

[0130]

[0131] Table 3 Quantitative analysis results of the factors affecting the peak load demand of new energy curve characteristics

[0132]

[0133]

[0134] Table 4 Quantitative analysis results of the factors affecting the peak load demand due to the characteristics of the transmission curve

[0135] Figure 4 Tables 2-4 show the importance of various curve characteristics on peak load demand. Figure 4 It can be seen that the impact of various load characteristics on peak demand is still much higher than that of other curve characteristics. This conclusion is consistent with Figure 3 The conclusions are consistent with the facts. Among all the characteristics, the peak of the load curve accounts for 92.9%, which is much higher than other characteristics. This is also consistent with the actual situation of Chongqing power grid, that is, under high load conditions, the peak demand is much higher than other times.

[0136] Judging from the simulation results, the proposed method can accurately quantify the peak-shaving demand of the power system under large-scale external power transmission, and on this basis, the influencing factors are quantitatively analyzed and evaluated through the data-driven method of XGBoost, which is consistent with the actual situation of the actual power grid. The effectiveness of the method proposed in the present invention is verified, which is beneficial for dispatchers to judge the peak-shaving demand based on the existing power grid through the predicted power curve, which serves as a favorable reference.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A method for evaluating the peak load demand and influencing factors of a power system considering the access of external power transmission, characterized by: The following steps are involved: S1. Obtain power system parameters; S2. Construct a unit combination scheduling optimization model that takes into account the relaxation of peak demand and determine the constraints of the unit combination scheduling optimization model; S3. Solve the unit combination dispatch optimization model to obtain the peak load demand relaxation of the power system; S4. Construct a sample set of factors affecting peak demand, and input the sample set of factors affecting peak demand and the slack amount of peak demand into the XGBoost algorithm model for processing to determine the importance of each factor; S5. Arrange the factors affecting peak demand in descending order of importance.

2. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 1 is characterized by: The objective function of the unit commitment scheduling optimization model is: Where: P g,t is the output of unit g at time t, is the startup cost of unit g, represents the shutdown cost of unit g, C g represents the output cost of unit g, G represents the set of units, T represents the set of scheduling times, ε p is the output relaxation variable of unit g, ε e is the ramp slack variable of unit g, M is the penalty coefficient of the slack variable; The constraints of the unit combination scheduling optimization model include relaxed unit output constraints, relaxed ramp constraints, unit start and stop time limit constraints, start and stop cost constraints, and power balance constraints.

3. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 2 is characterized by: The relaxed unit output constraint is: Among them: U g,t Indicates the start and stop status of unit g at time t; P g,max Indicates the upper limit of the output of unit g, P g,min Indicates the lower limit of the output of unit g, represents the lower limit slack variable of unit g at time t, Represents the upper limit slack variable of unit g at time t.

4. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 2 is characterized by: The relaxed climbing constraint is: Where: ΔP g,max.up and ΔP g,max.down They represent the ramp-up rate and ramp-down rate of unit g respectively; In ΔP g,max.up Adding uphill slack variables And in ΔP g,max.down Add downhill slack variables to Modify formula (1) as follows: Introducing auxiliary variables and auxiliary variables Rewrite (2) as: η g,t =U g,t -IN g,t-1 ,g∈G,t∈T, and: Among them: H is a large constant.

5. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 2 is characterized by: The unit start and stop time limit constraints are: Where: T g,U and T g,D Indicates the minimum start and stop time of unit g; t g,t,on and t g,t,off They respectively represent the continuous startup and shutdown time of unit g at time t.

6. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 2 is characterized by: The start-stop cost constraint is: Where: K g and J g Represents the startup and shutdown costs of unit g.

7. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 2 is characterized by: The power balance constraint is: ∑ g∈G,t∈T P g,t =∑ d∈D,t∈T In d,t , Where: V d,t represents the load value of node d at time t; D is the load set.

8. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 2 is characterized by: The sample set of factors affecting peak demand includes the first type of factors and the second type of factors, among which: The first type of influencing factor eigenvector for: in: and Represents the eigenvectors Subset, as well as They represent the power values ​​of load, new energy and external power at time t, respectively. T Indicates the total number of scheduling time periods; The second type of influencing factor eigenvector for: in: and express A subset of pe ,x va ,x ga and x ra They respectively represent the peak value, valley value, peak-to-valley difference rate and maximum climbing rate of the load, new energy and external power transmission curves.

9. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 8 is characterized by: The importance of the influencing factors is determined as follows: k m =N m / N M , Where: k m Indicates the importance of feature m, N m N is the number of splits used for feature m during the XGBoost fitting process; M is the total number of splits during the XGBoost fitting process.

10. The method for evaluating the peak load demand and influencing factors of the power system considering the access of external power transmission according to claim 1, characterized in that: The unit commitment scheduling optimization model is solved using the GUROBI solver.

Images