CN117057566A - Wind and fire source storage integrated station master-slave game optimization scheduling method - Google Patents

Abstract

The application discloses a wind and fire source storage integrated station master-slave game optimization scheduling method, which belongs to the field of wind power prediction and comprises the following steps of: step S1: according to the predicted wind speed, combining an active power output-speed characteristic curve of the wind turbine to obtain initial predicted data of wind power output; step S2: and (3) adopting a Latin hypercube sampling method, and carrying out wind power output initial scene generation by uniform layered sampling. The method solves the problem that the initial clustering center of the traditional K-means algorithm is randomly generated, avoids the problem that the algorithm falls into local optimum or iteration compensation fluctuation is large, improves the clustering accuracy, is beneficial to improving the prediction precision of uncertain wind power, enables the clustering result to be more accurate, is beneficial to the prediction of wind power output, establishes a master-slave game scheduling model for wind power storage integrated stations, and can further ensure the balance and stability of benefit distribution of each main body.

Description

Wind and fire source storage integrated station master-slave game optimization scheduling method

Technical Field

The application relates to the technical field of wind power prediction, in particular to a master-slave game optimization scheduling method for wind and fire source storage integrated stations based on an improved K-means algorithm.

Background

Traditional uncertain wind power prediction is often divided into two parts, namely generation of a wind power scene and reduction of the wind power scene. The common methods for generating the wind power scene include a self-help statistical method based on resampling, a Monte Carlo sampling method based on random sampling and a Latin hypercube sampling method based on layered sampling. And a hierarchical clustering method, an expectation maximization algorithm, a traditional K-means clustering algorithm and the like are often adopted in the reduction of wind power scenes.

The traditional optimal scheduling is generally from the aspects of economy, efficiency, flexibility, reliability and the like, and an optimal scheduling model of the system is established by taking the minimum of power generation cost, running cost and the like as a single/multiple objective function.

If wind power scenes are cut by the hierarchical clustering method, the calculation complexity is high, the method is not suitable for large-scale data sets, and the generated clustering result is difficult to adjust; if the selection of the initial center of the wind power scene is performed through the expectation maximization algorithm, the wind power scene is sensitive to the selection of the initial center, and the calculation cost is high. Whereas the conventional K-means algorithm also has disadvantages: (1) the method is sensitive to the selection of the initial clustering centers, the selection of the initial clustering centers possibly has larger influence on the final clustering result, the algorithm randomly selects the initial clustering centers, and the improper selection of the initial clustering centers easily causes the algorithm to fall into a local optimal solution or has larger iteration step fluctuation. (2) The super-parameter K of the clustering needs to be preset, different K values can bring different clustering results, and the K values are difficult to reasonably determine. Therefore, wind power prediction is performed by using a hierarchical clustering method, an expected maximum algorithm and a traditional K-means algorithm, and the prediction accuracy is relatively low.

Through retrieval, the Chinese patent with the application number of CN116191562A discloses a wind power base optimal scheduling method and a wind power base optimal scheduling device, which mention the problem that accurate scheduling of a power system cannot be realized, and mention a method for realizing optimal scheduling by adopting computer equipment and a computer program;

the Chinese patent of application number CN113128768A discloses a water, wind and fire short-term optimization scheduling method considering wind power uncertainty, which mentions that the problem of wind power random fluctuation characteristics is difficult to solve by the existing scheme.

The Chinese patent with application number CN102097828B discloses a wind power optimization scheduling method based on power prediction, and aims to improve the stability and safety problem of wind power grid connection.

The traditional power system optimization scheduling model cannot consider the requirements of all the main bodies in the system, cannot reasonably distribute benefits to all operators in the system, and cannot accurately describe the actual running condition of the system in the power market environment.

In particular, for the optimal scheduling of the wind-fire source storage integrated station, how to improve the prediction precision of the fluctuating wind power output, and meanwhile, how to solve the problem that the classical power system optimal scheduling strategy only ignores the actual problem of benefit distribution of each main body in the system, and how to establish the source storage integrated station optimal scheduling strategy based on master-slave games so that each main body in the integrated station realizes optimal scheduling of benefit balance distribution is an urgent problem to be solved in consideration of the guidance of a power grid on electricity price and competition conditions of each operator time in the power market environment.

Disclosure of Invention

The application aims to solve the defects in the prior art, and provides a master-slave game optimization scheduling method for an air-fire source storage integrated station.

In order to achieve the above purpose, the present application adopts the following technical scheme:

a wind and fire source storage integrated station master-slave game optimization scheduling method comprises the following steps:

step S1: according to the predicted wind speed, combining an active power output-speed characteristic curve of the wind turbine to obtain initial predicted data of wind power output;

step S2: generating an initial scene of wind power output by uniformly layering sampling by using a Latin hypercube sampling method;

step S3: aiming at the problems that the initial cluster center is randomly selected by the traditional K-means algorithm and is easy to fall into local optimum, and the K value of the optimum cluster number is difficult to determine, the initial cluster center selection of the K-means based on the density peak value clustering algorithm is improved, and the cluster effectiveness index is defined to find the optimum K value;

step S4: establishing a station optimization scheduling model under a Stackelberg master-slave game mode, taking a power grid company as a game leader, and taking a wind power plant, a thermal power plant and an energy storage power plant as game followers;

step S5: and solving and optimizing a scheduling strategy through a master-slave game model of the wind-fire source storage integrated station.

Further, in step S3, the method specifically includes the following steps:

s301: calculating the sample density by adopting a clustering algorithm based on rapid searching and finding density peaks, thereby determining an initial clustering center of a K-means clustering algorithm, wherein:

for wind power output dataset s= { X ₁ ，X ₂ ，…，X _n Computing data point X in its v-dimensional sample _i And X is _j European distance between Dist (X) _i ,X _j ) Expressed as:

s302: setting neighborhood cutoff distance D _np Ensuring that most samples have a neighbor number of about 1% to 2%;

s303: determining local density to represent the density of sample points in its neighborhood, using local density ρ _i The calculation method using the gaussian kernel is represented as:

s304: calculating the relative distance delta of wind power samples _i The first cluster center is defined as the locally most dense sample, i.e. the density peak sample, whose sample relative distance delta is defined _i The maximum relative distance for all samples is expressed as:

calculate the relative distance delta of other samples _i ：

S305: according to the decision value gamma _i Sequentially searching other density peak points, and determining a sample decision value gamma _i Calculating and sequencing, namely selecting the sample data point with the largest sample decision value except the first clustering center as a second density peak value, namely a second clustering center, and the like:

γ _i ＝ρ _i ·δ _i ；

s306: calculating an average value D of the distances between the single sample data and the corresponding initial cluster center O _intra To represent the compactness within the clusters, and to calculate the minimum distance D of the initial cluster center of each cluster _inter To represent the separability between different clusters, in particular:

D _inter ＝minDist(O _i ,O _j )；

s307: defining and calculating a cluster effectiveness index function T _eff ：

S308: searching the K value of the optimal cluster quantity through clustering effectiveness indexes, and calculating effectiveness index functions T respectively corresponding to the cluster quantity from 1 to K _eff The value K is the optimal clustering number K when the value is maximum, and K initial clustering centers are formed;

s309: calculating Euclidean distances between the wind power output initial scene sample obtained in the step S2 and the initial cluster center formed in the step S308, and respectively distributing the Euclidean distances to the cluster centers closest to the Euclidean distances to form K clusters;

s310: re-calculating the average value of wind power samples in K clusters to be used as a new cluster center;

s311: step S309 and step S310 are repeated until the K cluster centers in step S306 are not changed, and the K cluster centers are considered to be the final K typical scenes of wind power output.

Further, in step S4, the method specifically includes the following steps:

s401: calculating a wind power operator revenue function M _W The specific mode is as follows:

Y _W,i,t ＝μ _W P _Ws,i,t,

I _sub,W ＝i _sub,W P _Ws,i,t

wherein lambda is _W,t The wind power online electricity price is t time period; p (P) _W,i,t Delivering active power for a t-period wind power operator; lambda (lambda) _{W_B,t} 、λ _{B_W,t} The wind electricity selling price and the wind electricity purchasing price in the t-period wind electricity storage transaction are respectively; p (P) _{W_B,i,t} 、P _{B_W,i,t} The actual surplus and deficient electric energy of the wind power plant in the t period are respectively; y is Y _W,i,t The operation and maintenance expenditure for each hour is reduced for the whole life cycle of the wind turbine generator system i; p (P) _Ws,i,t, The actual wind power output is t time periods; mu (mu) _W The operation and maintenance expenditure coefficients of the wind turbine generator system are calculated; i _sub,W,i,t The method comprises the steps of subsidizing income for new energy sources of wind power grid connection; i.e _sub,W,i,t The electricity price is subsidized for the unit wind power;

s402: calculating a thermal power operator profit function M _W The specific mode is as follows:

C _{tra_H} ＝λ _H,t P _H,i,t

wherein U is _git,i,t Is the start-stop state of the thermal power unit i at the moment t, U _git,i,t =1 means that the unit i operates at time t, U _git,i,t =0 means that the unit i is shut down at time t; c (C) _Hfc,i,t The coal cost of the thermal power unit i; c (C) _H,i,TSTO The method is the start-stop cost of the thermal power unit i at the time t; c (C) _{tra_H} Revenue for thermal power generating units;the carbon emission cost of the thermal power generating unit is; k (k) _H The emission reduction coefficient of the thermal power generating unit; a, a _2i 、a _1i 、a _0i The secondary term, the primary term and the constant term coefficient of the coal cost function of the thermal power unit i; lambda (lambda) _H,t The method comprises the steps of (1) surfing electricity price for a thermal power generating unit at a moment t; p (P) _H,i,t Active power of thermal power unit i in t period, < > for thermal power unit i>The unit carbon emission treatment price of the thermal power unit is set; s is(s) _H Carbon emission for unit power generation of the thermal power generating unit;

s403: calculating energy storage operator revenue function M _B The specific mode is as follows:

s404: calculating a cost function C of the electric network _grid The specific mode is as follows:

wherein k is _G Representing the emission reduction coefficient of the power grid; c (C) _pol,G,i,t Representing the social responsibility cost of the power grid for emission reduction;

s405: setting wind power output constraint conditions, wherein the specific mode is as follows:

wherein,the minimum and maximum output of the wind turbine generator set are determined by the output characteristics of the wind turbine generator set i respectively;

s406: setting thermal power output constraint conditions, wherein the specific mode is as follows:

wherein,respectively the minimum and maximum active output of the thermal power unit i; />The climbing upper limit and the climbing lower limit of the thermal power unit i are set; TS (transport stream) _i 、TO _i The minimum shutdown time and the minimum operation time of the thermal power generating unit i are respectively;

s407: setting energy storage output constraint conditions, wherein the specific mode is as follows:

wherein U is _B,i,t,cha For the state of charge of the energy storage battery i at time t, U _B,i,t,cha =1 represents a discharge state; u (U) _B,i,t,dis U is the discharge state of the energy storage battery i at the moment t _B,i,t,dis =1 represents a discharge state;maximum charging power for the electrical energy storage i; />Maximum discharge power for the electrical energy storage i; SOC (State of Charge) _i,t The capacity of the energy storage device i is t time period; η (eta) _B,i,t,cha 、η _B,i,t,dis The charge and discharge efficiencies of the electric energy storage i are respectively;

s408: setting electric power constraint, specifically:

P _Line ≤P _Line,max

U _i,min ≤U _i ≤U _i,max

wherein P is _Line 、P _Line,max The actual active power and the maximum transmission capacity of the line are transmitted respectively for the line; u (U) _i,min U _i,max The upper and lower limits of the voltage of the node i are respectively set;

s409: calculating the balance degree of a master-slave game scheduling model of a source storage integrated station, and evaluating the balance degree of benefit distribution in the whole system by comparing the difference between the decision variable and the average value of each participant, wherein the specific mode is as follows:

x ₁ (t)＝P _W,t /Z _W

x ₂ (t)＝P _H,t /Z _H

x ₃ (t)＝P _B,t /Z _B

wherein BA (t) is the balance of the profit allocation, and the value range is [0,1 ]]；x ₁ (t)、x ₂ (t)、x ₃ (t) are the return coefficients of wind power, thermal power and energy storage at the moment t respectively; p (P) _W,t 、P _H,t 、P _B,t Respectively outputting wind power, thermal power and energy storage at the moment t; z is Z _W 、Z _H 、Z _B The wind power, the thermal power and the energy storage installed capacity are respectively.

Further, in step S5, the method specifically includes the following steps:

s501: importing the wind power data and other generator set initial data and parameters in the step S3 into a master-slave game mode integrated station scheduling model established in the step S4;

s502: given a game equalization solution initial value: randomly selecting a policy among a set of follower policies (P _W,t,0 ,P _H,t,0 ,P _B,t,0 ) As initial wind power, thermal power and energy storage decision values;

s503: follower decision: according to the online electricity price of the previous round, wind power, thermal power and energy storage of the follower are optimized and processed with the maximum benefit as targets, and a new output scheme is obtained;

s504: judging whether the optimal solution of the round is found, if the optimal solution of the round is consistent with the optimal iteration result of the previous round, indicating that the follower finds the optimal solution, otherwise, iteratively updating the decision of the follower again;

s505: leader decision: the power grid obtains the optimal solution of the leader by taking the minimum cost of the optimal solution of the wind power and the thermal power in the current iteration and the electricity price in the previous iteration as inputs;

s506: judging whether master-slave equilibrium is reached: if the result of the leading person' S turn is consistent with the result of the previous turn, the result indicates that a balance solution of Stackelberg is achieved, otherwise, the step S503 is returned;

s507: and calculating to obtain a game result and calculating the balance degree.

Compared with the prior art, the application has the beneficial effects that:

wind power output prediction is carried out based on an improved K-means clustering algorithm, and the improvement specific expression is as follows: the method solves the problem of random generation of the initial clustering center of the traditional K-means algorithm, avoids the problem that the algorithm falls into local optimum or iteration compensation fluctuation is large, improves the accuracy of clustering, and is beneficial to improving the prediction precision of uncertain wind power.

Wind power output prediction is carried out based on an improved K-means clustering algorithm, and the improvement specific expression is as follows: the method has the advantages that the clustering effectiveness index is defined to find the optimal clustering number K value, the error caused by the artificial random determination of the K value of the traditional K-means algorithm is solved, the clustering result is more accurate, and the prediction of wind power output is facilitated.

Compared with the traditional classical power system optimization scheduling model which only aims at the maximum overall economy of the system, the model fully considers the interest appeal of each main body in the source storage integrated station and further ensures the balance and stability of the benefit distribution of each main body.

Drawings

The accompanying drawings are included to provide a further understanding of the application and are incorporated in and constitute a part of this specification, illustrate the application and together with the embodiments of the application, serve to explain the application.

FIG. 1 is a flowchart of an improved K-means clustering algorithm in an embodiment of the application;

FIG. 2 is a schematic diagram of a master-slave game structure of an integrated wind and fire source storage station in an embodiment of the application;

FIG. 3 is a flowchart of a solution of an optimized scheduling model in a master-slave gaming mode of an integrated station in an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments.

Referring to fig. 1-3, the master-slave game optimization scheduling method for the wind and fire source storage integrated station comprises the following steps:

step S1: according to the predicted wind speed, combining an active power output-speed characteristic curve of the wind turbine to obtain initial predicted data of wind power output;

step S2: generating an initial scene of wind power output by uniformly layering sampling by using a Latin hypercube sampling method;

step S3: aiming at the problems that the initial cluster center is randomly selected by the traditional K-means algorithm and is easy to fall into local optimum, and the K value of the optimum cluster number is difficult to determine, the initial cluster center selection of the K-means based on the density peak value clustering algorithm is improved, and the cluster effectiveness index is defined to find the optimum K value; the flow chart of the improved K-means clustering algorithm is shown in figure 1.

Step S4: establishing a station optimization scheduling model under a Stackelberg master-slave game mode, taking a power grid company as a game leader, and taking a wind power plant, a thermal power plant and an energy storage power plant as game followers; a schematic diagram of the master-slave game structure of the wind and fire source storage integrated station is shown in fig. 2.

Step S5: solving a master-slave game model of the wind-fire source storage integrated station to optimize a scheduling strategy; the solution flow chart of the optimal scheduling model in the master-slave game mode of the integral station is shown in fig. 3.

In a specific embodiment of the present application, step S3 specifically includes the steps of:

s301: and calculating the sample density by adopting a clustering algorithm based on the rapid searching and finding density peaks, so as to determine the initial clustering center of the K-means clustering algorithm.

The method comprises the following steps: for wind power output dataset s= { X ₁ ，X ₂ ，…，X _n Computing data point X in its v-dimensional sample _i And X is _j European distance between Dist (X) _i ,X _j ) Expressed as:

s302: setting neighborhood cutoff distance D _np Ensuring that most samples have a neighbor number of about 1% to 2%;

s303: determining local density to represent the density of sample points in its neighborhood, using local density ρ _i Representing the manner of computation using a Gaussian Kernel (Gaussian Kernel):

s304: calculating the relative distance delta of wind power samples _i . The first cluster center is defined as the locally most dense sample, i.e. the density peak sample, whose sample relative distance delta is defined _i The maximum relative distance for all samples is expressed as:

calculate the relative distance delta of other samples _i ：

S305: according to the decision value gamma _i Sequentially searching other density peak points, and determining a sample decision value gamma _i Calculating and sequencing, namely selecting the sample data point with the largest sample decision value except the first clustering center as a second density peak value, namely a second clustering center, and the like:

γ _i ＝ρ _i ·δ _i ；

s306: calculating an average value D of the distances between the single sample data and the corresponding initial cluster center O _intra To represent compactness within a cluster; calculating the minimum distance D of the initial cluster center of each cluster _inter To represent the separability between different clusters:

D _inter ＝minDist(O _i ,O _j )；

s307: defining and calculating a cluster effectiveness index function T _eff ：

S308: and searching an optimal cluster quantity K value through a cluster effectiveness index. Calculating effectiveness index functions T respectively corresponding to the number of 1 to k clusters _eff The value K is the optimal clustering number K when the value is maximum, and K initial clustering centers are formed;

s309: calculating Euclidean distances between the wind power output initial scene sample obtained in the step S2 and the initial cluster center formed in the step S308, and respectively distributing the Euclidean distances to the cluster centers closest to the Euclidean distances to form K clusters;

s310: re-calculating the average value of wind power samples in K clusters to be used as a new cluster center;

s311: step S309 and step S310 are repeated until the K cluster centers in step S306 are not changed, and the K cluster centers are considered to be the final K typical scenes of wind power output.

In a specific embodiment of the present application, step S4 specifically includes the following steps;

s401: calculating a wind power operator revenue function M _W 。

Y _W,i,t ＝μ _W P _Ws,i,t,

I _sub,W ＝i _sub,W P _Ws,i,t

Wherein lambda is _W,t The wind power online electricity price is t time period; p (P) _W,i,t Delivering active power for a t-period wind power operator; lambda (lambda) _{W_B,t} 、λ _{B_W,t} The wind electricity selling price and the wind electricity purchasing price in the t-period wind electricity storage transaction are respectively; p (P) _{W_B,i,t} 、P _{B_W,i,t} The actual surplus and deficient electric energy of the wind power plant in the t period are respectively; y is Y _W,i,t The operation and maintenance expenditure for each hour is reduced for the whole life cycle of the wind turbine generator system i; p (P) _Ws,i,t, The actual wind power output is t time periods; mu (mu) _W The operation and maintenance expenditure coefficients of the wind turbine generator system are calculated; i _sub,W,i,t The method comprises the steps of subsidizing income for new energy sources of wind power grid connection; i.e _sub,W,i,t Wind power compensation for unitPasting electricity price;

s402: calculating a thermal power operator profit function M _W 。

C _{tra_H} ＝λ _H,t P _H,i,t

Wherein U is _git,i,t Is the start-stop state of the thermal power unit i at the moment t, U _git,i,t =1 means that the unit i operates at time t, U _git,i,t =0 means that the unit i is shut down at time t; c (C) _Hfc,i,t The coal cost of the thermal power unit i; c (C) _H,i,TSTO The method is the start-stop cost of the thermal power unit i at the time t; c (C) _{tra_H} Revenue for thermal power generating units;the carbon emission cost of the thermal power generating unit is; k (k) _H The emission reduction coefficient of the thermal power generating unit; a, a _2i 、a _1i 、a _0i The secondary term, the primary term and the constant term coefficient of the coal cost function of the thermal power unit i; lambda (lambda) _H,t The method comprises the steps of (1) surfing electricity price for a thermal power generating unit at a moment t; p (P) _H,i,t Active power of thermal power unit i in t period, < > for thermal power unit i>The unit carbon emission treatment price of the thermal power unit is set; s is(s) _H Carbon emission for unit power generation of the thermal power generating unit;

s403: calculating energy storage operator revenue function M _B 。

S404: calculating a cost function C of the electric network _grid ：

Wherein k is _G Representing the emission reduction coefficient of the power grid; c (C) _pol,G,i,t Representing the social responsibility cost of the power grid for emission reduction;

s405: setting wind power output constraint conditions:

wherein,the minimum and maximum output of the wind turbine generator set are determined by the output characteristics of the wind turbine generator set i respectively;

s406: setting thermal power output constraint conditions:

wherein,respectively the minimum and maximum active output of the thermal power unit i; />The climbing upper limit and the climbing lower limit of the thermal power unit i are set; TS (transport stream) _i 、TO _i The minimum shutdown time and the minimum operation time of the thermal power generating unit i are respectively;

s407: setting energy storage output constraint conditions:

wherein U is _B,i,t,cha For the state of charge of the energy storage battery i at time t, U _B,i,t,cha =1 represents a discharge state; u (U) _B,i,t,dis U is the discharge state of the energy storage battery i at the moment t _B,i,t,dis =1 represents a discharge state;maximum charging power for the electrical energy storage i; />Maximum discharge power for the electrical energy storage i; SOC (State of Charge) _i,t The capacity of the energy storage device i is t time period; η (eta) _B,i,t,cha 、η _B,i,t,dis The charge and discharge efficiencies of the electrical energy storage i are respectively.

S408: setting an electric power constraint:

P _Line ≤P _Line,max

U _i,min ≤U _i ≤U _i,max

wherein P is _Line 、P _Line,max The actual active power and the maximum transmission capacity of the line are transmitted respectively for the line; u (U) _i,min U _i,max The upper and lower limits of the voltage of the node i are respectively set;

s409: calculating the balance degree of a master-slave game scheduling model of the source storage integrated station, and evaluating the balance degree of benefit distribution in the whole system by comparing the difference between the decision variable and the average value of each participant:

x ₁ (t)＝P _W,t /Z _W

x ₂ (t)＝P _H,t /Z _H

x ₃ (t)＝P _B,t /Z _B

wherein BA (t) is the balance of the profit allocation, and the value range is [0,1 ]]；x ₁ (t)、x ₂ (t)、x ₃ (t) are the return coefficients of wind power, thermal power and energy storage at the moment t respectively; p (P) _W,t 、P _H,t 、P _B,t Respectively outputting wind power, thermal power and energy storage at the moment t; z is Z _W 、Z _H 、Z _B The wind power, the thermal power and the energy storage installed capacity are respectively.

In a specific embodiment of the present application, step S5 specifically includes the steps of:

s501: importing the wind power data and other generator set initial data and parameters in the step S3 into a master-slave game mode integrated station scheduling model established in the step S4;

s502: given a game equalization solution initial value. Randomly selecting a policy among a set of follower policies (P _W,t,0 ,P _H,t,0 ,P _B,t,0 ) As initial wind power, thermal power and energy storage decision values;

s503: follower decision. According to the online electricity price of the previous round, wind power, thermal power and energy storage of the follower are optimized and processed with the maximum benefit as targets, and a new output scheme is obtained;

s504: judging whether the optimal solution of the round is found, if the optimal solution of the round is consistent with the optimal iteration result of the previous round, indicating that the follower finds the optimal solution, otherwise, iteratively updating the decision of the follower again;

s505: the leader decides. The power grid obtains the optimal solution of the leader by taking the minimum cost of the optimal solution of the wind power and the thermal power in the current iteration and the electricity price in the previous iteration as inputs;

s506: and judging whether master-slave equilibrium is reached. If the result of the leading person' S turn is consistent with the result of the previous turn, the result indicates that a balance solution of Stackelberg is achieved, otherwise, the step S503 is returned;

s507: and calculating to obtain a game result and calculating the balance degree.

The foregoing is only a preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art, who is within the scope of the present application, should make equivalent substitutions or modifications according to the technical scheme of the present application and the inventive concept thereof, and should be covered by the scope of the present application.

Claims (4)

1. A wind and fire source storage integrated station master-slave game optimization scheduling method is characterized by comprising the following steps:

step S1: according to the predicted wind speed, combining an active power output-speed characteristic curve of the wind turbine to obtain initial predicted data of wind power output;

step S2: generating an initial scene of wind power output by uniformly layering sampling by using a Latin hypercube sampling method;

step S3: aiming at the problems that the initial cluster center is randomly selected by the traditional K-means algorithm and is easy to fall into local optimum, and the K value of the optimum cluster number is difficult to determine, the initial cluster center selection of the K-means based on the density peak value clustering algorithm is improved, and the cluster effectiveness index is defined to find the optimum K value;

step S4: establishing a station optimization scheduling model under a Stackelberg master-slave game mode, taking a power grid company as a game leader, and taking a wind power plant, a thermal power plant and an energy storage power plant as game followers;

step S5: and solving and optimizing a scheduling strategy through a master-slave game model of the wind-fire source storage integrated station.

2. The master-slave game optimization scheduling method of the wind and fire source storage integrated station according to claim 1, wherein in step S3, the method specifically comprises the following steps:

s301: calculating the sample density by adopting a clustering algorithm based on rapid searching and finding density peaks, thereby determining an initial clustering center of a K-means clustering algorithm, wherein:

for wind power output dataset s= { X ₁ ，X ₂ ，…，X _n Computing data point X in its v-dimensional sample _i And X is _j European distance between Dist (X) _i ,X _j ) Expressed as:

s302: setting neighborhood cutoff distance D _np Ensuring that most samples have a neighbor number of about 1% to 2%;

s303: determining local density to represent the density of sample points in its neighborhood, using local density ρ _i The calculation method using the gaussian kernel is represented as:

s304: calculating the relative distance delta of wind power samples _i The first cluster center is defined as the locally most dense sample, i.e. the density peak sample, whose sample relative distance delta is defined _i The maximum relative distance for all samples is expressed as:

calculate the relative distance delta of other samples _i ：

S305: according to the decision value gamma _i Sequentially searching other density peak points, and determining a sample decision value gamma _i Calculating and sequencing, namely selecting the sample data point with the largest sample decision value except the first clustering center as a second density peak value, namely a second clustering center, and the like:

γ _i ＝ρ _i ·δ _i ；

s306: calculating an average value D of the distances between the single sample data and the corresponding initial cluster center O _intra To represent the compactness within the clusters, and to calculate the minimum distance D of the initial cluster center of each cluster _inter To represent between different clustersSeparability, the specific mode is:

D _inter ＝minDist(O _i ,O _j )；

s307: defining and calculating a cluster effectiveness index function T _eff ：

S308: searching the K value of the optimal cluster quantity through clustering effectiveness indexes, and calculating effectiveness index functions T respectively corresponding to the cluster quantity from 1 to K _eff The value K is the optimal clustering number K when the value is maximum, and K initial clustering centers are formed;

s309: calculating Euclidean distances between the wind power output initial scene sample obtained in the step S2 and the initial cluster center formed in the step S308, and respectively distributing the Euclidean distances to the cluster centers closest to the Euclidean distances to form K clusters;

s310: re-calculating the average value of wind power samples in K clusters to be used as a new cluster center;

s311: step S309 and step S310 are repeated until the K cluster centers in step S306 are not changed, and the K cluster centers are considered to be the final K typical scenes of wind power output.

3. The master-slave game optimization scheduling method of the wind and fire source storage integrated station according to claim 2, wherein in step S4, the method specifically comprises the following steps:

s401: calculating a wind power operator revenue function M _W The specific mode is as follows:

Y _W,i,t ＝μ _W P _Ws,i,t ,

I _sub,W ＝i _sub,W P _Ws,i,t

wherein lambda is _W,t The wind power online electricity price is t time period; p (P) _W,i,t Delivering active power for a t-period wind power operator; lambda (lambda) _{W_B,t} 、λ _{B_W,t} The wind electricity selling price and the wind electricity purchasing price in the t-period wind electricity storage transaction are respectively; p (P) _{W_B,i,t} 、P _{B_W,i,t} The actual surplus and deficient electric energy of the wind power plant in the t period are respectively; y is Y _W,i,t The operation and maintenance expenditure for each hour is reduced for the whole life cycle of the wind turbine generator system i; p (P) _Ws,i,t, The actual wind power output is t time periods; mu (mu) _W The operation and maintenance expenditure coefficients of the wind turbine generator system are calculated; i _sub,W,i,t The method comprises the steps of subsidizing income for new energy sources of wind power grid connection; i.e _sub,W,i,t The electricity price is subsidized for the unit wind power;

s402: calculating a thermal power operator profit function M _W The specific mode is as follows:

C _{tra_H} ＝λ _H,t P _H,i,t

wherein U is _git,i,t Is the start-stop state of the thermal power unit i at the moment t, U _git,i,t =1 means that the unit i operates at time t, U _git,i,t =0 means that the unit i is shut down at time t; c (C) _Hfc,i,t The coal cost of the thermal power unit i; c (C) _H,i,TSTO The method is the start-stop cost of the thermal power unit i at the time t; c (C) _{tra_H} Revenue for thermal power generating units;the carbon emission cost of the thermal power generating unit is; k (k) _H The emission reduction coefficient of the thermal power generating unit; a, a _2i 、a _1i 、a _0i The secondary term, the primary term and the constant term coefficient of the coal cost function of the thermal power unit i; lambda (lambda) _H,t The method comprises the steps of (1) surfing electricity price for a thermal power generating unit at a moment t; p (P) _H,i,t Active power of thermal power unit i in t period, < > for thermal power unit i>The unit carbon emission treatment price of the thermal power unit is set; s is(s) _H Carbon emission for unit power generation of the thermal power generating unit;

s403: calculating energy storage operator revenue function M _B The specific mode is as follows:

s404: calculating a cost function C of the electric network _grid The specific mode is as follows:

wherein k is _G Representing the emission reduction coefficient of the power grid; c (C) _pol,G,i,t Representing the social responsibility cost of the power grid for emission reduction;

s405: setting wind power output constraint conditions, wherein the specific mode is as follows:

wherein,the minimum and maximum output of the wind turbine generator set are determined by the output characteristics of the wind turbine generator set i respectively;

s406: setting thermal power output constraint conditions, wherein the specific mode is as follows:

wherein,respectively the minimum and maximum active output of the thermal power unit i; />The climbing upper limit and the climbing lower limit of the thermal power unit i are set; TS (transport stream) _i 、TO _i The minimum shutdown time and the minimum operation time of the thermal power generating unit i are respectively;

s407: setting energy storage output constraint conditions, wherein the specific mode is as follows:

wherein U is _B,i,t,cha For the state of charge of the energy storage battery i at time t, U _B,i,t,cha =1 represents a discharge state; u (U) _B,i,t,dis U is the discharge state of the energy storage battery i at the moment t _B,i,t,dis =1 represents a discharge state;maximum charging power for the electrical energy storage i; />Maximum discharge power for the electrical energy storage i; SOC (State of Charge) _i,t The capacity of the energy storage device i is t time period; η (eta) _B,i,t,cha 、η _B,i,t,dis The charge and discharge efficiencies of the electric energy storage i are respectively;

s408: setting electric power constraint, specifically:

P _Line ≤P _Line,max

U _i,min ≤U _i ≤U _i,max

wherein P is _Line 、P _Line,max The actual active power and the maximum transmission capacity of the line are transmitted respectively for the line; u (U) _i,min U _i,max The upper and lower limits of the voltage of the node i are respectively set;

s409: calculating the balance degree of a master-slave game scheduling model of a source storage integrated station, and evaluating the balance degree of benefit distribution in the whole system by comparing the difference between the decision variable and the average value of each participant, wherein the specific mode is as follows:

x ₁ (t)＝P _W,t /Z _W

x ₂ (t)＝P _H,t /Z _H

x ₃ (t)＝P _B,t /Z _B

wherein BA (t) is the balance of the profit allocation, and the value range is [0,1 ]]；x ₁ (t)、x ₂ (t)、x ₃ (t) are the return coefficients of wind power, thermal power and energy storage at the moment t respectively; p (P) _W,t 、P _H,t 、P _B,t Respectively outputting wind power, thermal power and energy storage at the moment t; z is Z _W 、Z _H 、Z _B The wind power, the thermal power and the energy storage installed capacity are respectively.

4. The master-slave game optimization scheduling method of the wind and fire source storage integrated station according to claim 3, wherein in step S5, the method specifically comprises the following steps:

s501: importing the wind power data and other generator set initial data and parameters in the step S3 into a master-slave game mode integrated station scheduling model established in the step S4;

s502: given a game equalization solution initial value: at follower policyA strategy (P) is selected randomly in the set _W,t,0 ,P _H,t,0 ,P _B,t,0 ) As initial wind power, thermal power and energy storage decision values;

s503: follower decision: according to the online electricity price of the previous round, wind power, thermal power and energy storage of the follower are optimized and processed with the maximum benefit as targets, and a new output scheme is obtained;

s504: judging whether the optimal solution of the round is found, if the optimal solution of the round is consistent with the optimal iteration result of the previous round, indicating that the follower finds the optimal solution, otherwise, iteratively updating the decision of the follower again;

s505: leader decision: the power grid obtains the optimal solution of the leader by taking the minimum cost of the optimal solution of the wind power and the thermal power in the current iteration and the electricity price in the previous iteration as inputs;

s506: judging whether master-slave equilibrium is reached: if the result of the leading person' S turn is consistent with the result of the previous turn, the result indicates that a balance solution of Stackelberg is achieved, otherwise, the step S503 is returned;

s507: and calculating to obtain a game result and calculating the balance degree.