Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a power generation group double-layer game bidding method and a power generation group double-layer game bidding system in a power spot market environment, wherein a power generation group double-layer bidding model is constructed to simplify modeling behaviors; and converting the double-layer bidding model into a single-layer model by applying a KKT condition and a binary extension method so as to reduce the calculation difficulty.
Therefore, the invention adopts the following technical scheme: a power generation group double-layer game bidding method in a power spot market environment comprises the following steps: firstly, constructing a double-layer bidding model of a power generation group participating in a day-ahead main energy market and a day-ahead standby service market, wherein electric energy and standby service quotation of the power generation group are linear 'electricity price-electricity quantity' curves, secondly, converting the double-layer bidding model into a single-layer model through a KKT condition and a binary expansion method, converting a bidding strategy of the power generation group into a discrete strategy set, and eliminating a nonlinear term existing in the single-layer model, so that the single-layer model is solved through a general commercial solver and a global optimal solution is obtained; and finally, solving a game equilibrium quotation result of the power generation group based on a quotation adjustment mechanism.
The other technical scheme adopted by the invention is as follows: a power generation group double-layer game bidding method in a power spot market environment comprises the following steps:
1) acquiring basic clearing rules and day-ahead clearing boundary conditions of the power system;
2) constructing a double-layer bidding model of a power generation group participating in a day-ahead main energy market and a day-ahead standby service market, wherein an upper layer model reflects the bidding behavior of the power generation group, a lower layer model reflects the clearing process of a spot market, and random optimization is applied to the uncertainty of the output and the load of new energy of a real-time market;
3) based on a KKT condition and a binary extension method, the double-layer bidding model is converted into a single-layer model, nonlinear terms in the single-layer model are eliminated, and meanwhile, a power generation group bidding strategy is converted into a discrete strategy set.
Further, the double-layer bidding model of the power generation group is as follows:
1) an upper layer model: bidding model of power generation group
The objective function of the upper model is expressed as:
in the formula, k
e,i,k
r,iExpressing the quotation coefficient as a decision variable; t represents the time-sharing period number of the market in the day ahead; pi
ωThe omega scene occurrence probability; n is a radical of
jRepresenting the total number of the thermal generator sets in the power generation group j;
respectively representing the normal bid price and the medium bid electric quantity of the unit i at the moment t of the day-ahead electric energy market;
respectively representing the bid price and the bid electric energy adjustment quantity of the unit i at the real-time market t moment under an omega scene;
the standby bid price and the intermediate scalar quantity of the unit i at the moment t of the day-ahead standby service market are obtained by a lower-layer model; c
iThe total cost of the standard-out power in the unit i is represented and simplified into a quadratic form:
in the formula, a
m,i、b
m,i、c
m,iRespectively a quadratic term cost coefficient, a primary term cost coefficient and a constant term cost coefficient;
the total bid amount of the unit i in the spot market under the omega scene is represented as:
decision variables of the upper layer model are quotation strategies of each unit in the generator set in the day-ahead electric energy market and the day-ahead standby service market; the linear quotation curve of the unit in the electric energy market is based on marginal cost, and the standard quotation curve of experience history is referred to in the future standby service market and is expressed as follows:
λe,m,i(P)=ke,m,i(am,iP+bm,i),
λr,m,i(r)=kr,m,i(αm,ir+βm,i),
wherein λ ise,m,i,λr,m,iRespectively showing the quotation curves of the unit i in the day-ahead electric energy market and the day-ahead standby service market, P, r respectively showing the reported electric quantity of the day-ahead electric energy market and the reported capacity a of the day-ahead standby service marketm,i,bm,i,αm,i,βm,iIs a constant related to cost, ke,m,i,kr,m,iAll represent the quotation coefficient, are decision variables, and satisfy the following constraints:
in the formula (I), the compound is shown in the specification,
are each k
e,m,i,k
r,m,iAn upper limit value of (d);
2) the lower layer model: spot market clearing model
The objective function of the underlying model is to minimize the spot market electricity purchase cost, expressed as:
wherein, the electricity purchasing cost comprises three items of the current electricity purchasing cost, the standby service calling cost and the real-time electricity purchasing cost, M represents the total number of the spot market electricity generation group,
deciding a variable set for a lower layer;
further, the day-ahead spot market constraints include:
1) node power balance constraints
In the formula i
b,j
bRepresents a network node;
respectively indicate the time t is connected to the point i
bTotal bid power quantity of thermal power generating units in node market day ahead, output power of new energy source unit and load predicted value, K
LRepresents a collection of lines in a network topology,
representing the line trend of the market at the day-ahead,
is a lagrange multiplier;
2) line flow equality constraints
In the formula (I), the compound is shown in the specification,
for node i at time t
bAnd j
bThe phase angle of (a) is,
is a node i
bAnd j
bThe impedance of the line between the two lines,
is a lagrange multiplier;
3) line capacity constraint
In the formula (I), the compound is shown in the specification,
is a node i
bAnd j
bThe maximum transmission capacity of the inter-line,
is a lagrange multiplier;
4) unit output constraint
In the formula (I), the compound is shown in the specification,
respectively the maximum and minimum output values of the generator sets i under the generating group m,
is a lagrange multiplier;
5) spare capacity constraint
In the formula (I), the compound is shown in the specification,
minimum spare capacity for time period t; mu.s
r,tIs the Lagrange coefficient;
6) power justification rate constraints
In the formula (I), the compound is shown in the specification,
the maximum up-regulation and down-regulation output force of the unit i in unit time,
is a lagrange multiplier;
7) phase angle constraint
In the formula (I), the compound is shown in the specification,
is a lagrange multiplier;
further, real-time market constraints include:
1) node power balance constraints
In the formula (I), the compound is shown in the specification,
is a node i
bProcessing the difference value between the output value corresponding to the time interval t of the thermal power generating unit under the scene omega and the scalar in the market before the day;
is a node i
bThe difference value between the output value corresponding to the time interval t of the new energy unit under the scene omega and the scalar in the market before the day;
is a node i
bDifference values of load values corresponding to the time period t under the scene omega and the market predicted values in the day before;
is a node i
bThe difference value between the line power flow corresponding to the time period t under the scene omega and the line power flow of the market at the day before;
is the Lagrange coefficient;
2) line flow equality constraints
In the formula (I), the compound is shown in the specification,
are respectively node i
b,j
bThe difference value between the phase angle corresponding to the time period t in the scene omega and the day-ahead market phase angle;
is a lagrange multiplier;
3) line capacity constraint
In the formula (I), the compound is shown in the specification,
is a node i
b,j
bMaximum capacity of inter-line transmission;
is the Lagrange coefficient;
4) unit output constraint
In the formula (I), the compound is shown in the specification,
is a lagrange multiplier;
5) phase angle constraint
In the formula (I), the compound is shown in the specification,
is a lagrange multiplier;
further, the specific process of solving the double-layer bidding model is as follows:
firstly, expressing a lower layer model problem by using an equivalent KKT condition, linearizing a complementary relaxation condition based on a large M rule, explicitly unifying upper and lower layer variables, changing a double-layer bidding model into a single-layer model, wherein a binary variable product term still exists in the model, linearizing the binary product term by using a binary expansion method, and simultaneously converting a power generation group bidding strategy into a discrete strategy set, so that the subsequent game equilibrium solution is conveniently solved.
Further, when the underlying model is analyzed alone, the variable ke,m,i,kr,m,iA decision variable of an upper layer model is regarded as a constant; therefore, the lower model problem is a quadratic convex optimization problem containing linear constraint, and the decision variables are continuous variables, so that the application requirements of the KKT condition are met.
Further, the lower model equivalent KKT condition is expressed as:
wherein, the complementary relaxation conditions are expressed in a form that x ^ y is more than or equal to 0 and less than or equal to 0, and a large enough constant M and a 0-1 variable u are introduced for linearization treatment;
at the moment, the double-layer bidding model is converted into a single-layer model, but the binary variable product term still exists in the single-layer model
And k
r,m,ir
m,i,tConsidering k, consider
e,m,i,k
r,m,iThe binary extension method is applied to solve the two problems.
Further, the games among the power generation groups will eventually reach a nash equilibrium state, and in this state, no market subject can obtain greater profit by relying on self decision, that is, the games among the power generation groups can finally reach a nash equilibrium state, that is, no market subject can obtain greater profit by relying on self decision
Wherein i represents the subject of interest, r
iAny bidding strategy on behalf of the subject is,
on behalf of the subject's bid balancing strategy,
representing a competitive bidding balancing strategy, wherein the formation of a Nash balancing solution depends on a quotation adjusting mechanism in the market;
BAE based round-by-round offers in a spot market environment comprises the following steps:
the first step, before the market is cleared in the day, the power generation group reports an initial quotation curve and unit technical parameters to a dispatching and trading center; the dispatching and trading center discloses quotation information of all the power generation groups, user-side report information and other related network and unit parameters, and starts a first bidding round;
secondly, each power generation group obtains a new round of quotation and reports the quotation to a dispatching and trading center within a certain time limit according to an initial quotation curve of a competitor and a double-layer bidding model;
thirdly, the dispatching transaction center obtains a new round of quotation result, if the quotation of each power generation group is the same as that of the previous round, the game is ended, and the quotation result of the round is a Nash balance result; if the power generation group adjustment strategy still exists, the dispatching transaction center continuously discloses the quotation condition of the round to the whole society, and carries out bidding of the next round until all the power generation groups do not change the quotation strategy.
The invention adopts another technical scheme that: a power generation group double-layer game bidding system in a power spot market environment comprises an information acquisition module, a bidding analysis module, a model conversion and solving module, a quoted price adjusting module and a data output module;
the information acquisition module acquires the boundary condition information of market clearing;
the bidding analysis module forms a power generation group bidding model according to the boundary conditions;
the model conversion and solution module simplifies and solves the double-layer model;
the quotation adjusting module supports the power generation group to quotation round by round until a balanced solution is achieved;
and the data output module formats and outputs the bidding strategy and the market balance point information of the power generation group.
The invention has the following beneficial effects:
1. the invention models the bidding behaviors of the power generation group participating in the main energy and standby market on the basis of a double-layer bidding model. The double-layer bidding model clearly shows the bidding behaviors of the power generation group and the clearing process of the spot market, and can simplify the modeling process of the market behaviors of the power generation group.
2. The invention converts the double-layer bidding model into a single-layer model by a KKT condition and a binary expansion method and linearizes the nonlinear term. Compared with a double-layer model and a nonlinear problem, the single-layer model and the linear problem avoid the problem of infinite quotation strategies, are convenient to solve by using a commercial solver, and greatly reduce the calculation difficulty.
3. The invention can realize the automation of the processes of quotation information acquisition, bid analysis, model solution and data output by using the technology of an integrated modular system, thereby saving a large amount of labor cost.
4. The method has automation and generalizability, can be used for migrating and applying to the current market simulation analysis of the power grid company, and can greatly improve the analysis and utilization efficiency of data.
Detailed Description
The invention is further described with reference to the drawings and the detailed description.
Example 1
The embodiment provides a power generation group double-layer game bidding method in a power spot market environment, as shown in fig. 1, the steps are as follows:
firstly, obtaining basic information such as basic clearing rules and day-ahead clearing boundary conditions of a power system, and constructing a double-layer bidding model of a power generation group participating in a day-ahead main energy market and a day-ahead standby service market, wherein electric energy and standby service quotations of the power generation group are linear power price-electric quantity curves; secondly, converting a double-layer bidding model into a single-layer model through a KKT condition and a binary extension method, converting a power generation group bidding strategy into a discrete strategy set, and eliminating a nonlinear term existing in the single-layer model, so that the single-layer model is solved through a general commercial solver to obtain a global optimal solution; and finally, solving and outputting a game balanced quotation result of the power generation group based on a quotation adjusting mechanism.
The double-layer bidding model of the power generation group comprises the following steps:
1) an upper layer model: bidding model of power generation group
The objective function of the upper model is expressed as:
in the formula, k
e,i,k
r,iExpressing the quotation coefficient as a decision variable; t represents the time-sharing period number of the market in the day ahead; pi
ωThe omega scene occurrence probability; n is a radical of
jRepresenting the total number of the thermal generator sets in the power generation group j;
respectively representing the normal bid price and the medium bid electric quantity of the unit i at the moment t of the day-ahead electric energy market;
respectively representing the bid price and the bid electric energy adjustment quantity of the unit i at the real-time market t moment under an omega scene;
the standby bid price and the intermediate scalar quantity of the unit i at the moment t of the day-ahead standby service market are obtained by a lower-layer model; c
iThe total cost of the standard-out power in the unit i is represented and simplified into a quadratic form:
in the formula, a
m,i、b
m,i、c
m,iRespectively a quadratic term cost coefficient, a primary term cost coefficient and a constant term cost coefficient;
the total bid amount of the unit i in the spot market under the omega scene is represented as:
decision variables of the upper layer model are quotation strategies of each unit in the generator set in the day-ahead electric energy market and the day-ahead standby service market; the linear quotation curve of the unit in the day-ahead electric energy market is based on marginal cost, and the standard quotation curve of experience history is referred to in the day-ahead standby service market and is expressed as follows:
λe,m,i(P)=ke,m,i(am,iP+bm,i),
λr,m,i(r)=kr,m,i(αm,ir+βm,i),
wherein λ ise,m,i,λr,m,iRespectively showing the quotation curves of the unit i in the day-ahead electric energy market and the standby service market, wherein P, r respectively represents the reported electric quantity of the day-ahead electric energy market and the reported capacity a of the standby service marketm,i,bm,i,αm,i,βm,iIs a constant related to cost, ke,m,i,kr,m,iAll represent the quotation coefficient, are decision variables, and satisfy the following constraints:
in the formula (I), the compound is shown in the specification,
are each k
e,m,i,k
r,m,iAn upper limit value of (d);
2) the lower layer model: spot market clearing model
The objective function of the underlying model is to minimize the spot market electricity purchase cost, expressed as:
wherein, the electricity purchasing cost comprises three items of the current electricity purchasing cost, the standby service calling cost and the real-time electricity purchasing cost, M represents the total number of the spot market electricity generation group,
the set of variables is decided for the lower layer.
Further, the day-ahead spot market constraints include:
1) node power balance constraints
In the formula i
b,j
bRepresents a network node;
respectively indicate the time t is connected to the point i
bTotal bid power quantity of thermal power generating units in node market day ahead, output power of new energy source unit and load predicted value, K
LRepresents a collection of lines in a network topology,
representing the line trend of the market at the day-ahead,
is a lagrange multiplier;
2) line flow equality constraints
In the formula (I), the compound is shown in the specification,
for node i at time t
bAnd j
bThe phase angle of (a) is,
is a node i
bAnd j
bThe impedance of the line between the two lines,
is a lagrange multiplier;
3) line capacity constraint
In the formula (I), the compound is shown in the specification,
is a node i
bAnd j
bThe maximum transmission capacity of the inter-line,
is a lagrange multiplier;
4) unit output constraint
In the formula (I), the compound is shown in the specification,
respectively the maximum and minimum output values of the generator sets i under the generating group m,
is lagrange multiplicationA seed;
5) spare capacity constraint
In the formula (I), the compound is shown in the specification,
minimum spare capacity for time period t; mu.s
r,tIs the Lagrange coefficient;
6) power justification rate constraints
In the formula (I), the compound is shown in the specification,
the maximum up-regulation and down-regulation output force of the unit i in unit time,
is a lagrange multiplier;
7) phase angle constraint
In the formula (I), the compound is shown in the specification,
is a lagrange multiplier.
Further, real-time market constraints include:
1) node power balance constraints
In the formula (I), the compound is shown in the specification,
is a node i
bProcessing the difference value between the output value corresponding to the time interval t of the thermal power generating unit under the scene omega and the scalar in the market before the day;
is a node i
bThe difference value between the output value corresponding to the time interval t of the new energy unit under the scene omega and the scalar in the market before the day;
is a node i
bDifference values of load values corresponding to the time period t under the scene omega and the market predicted values in the day before;
is a node i
bThe difference value between the line power flow corresponding to the time period t under the scene omega and the line power flow of the market at the day before;
is the Lagrange coefficient;
2) line flow equality constraints
In the formula (I), the compound is shown in the specification,
are respectively node i
b,j
bThe difference value between the phase angle corresponding to the time period t in the scene omega and the day-ahead market phase angle;
is a lagrange multiplier;
3) line capacity constraint
In the formula (I), the compound is shown in the specification,
is a node i
b,j
bMaximum capacity of inter-line transmission;
is the Lagrange coefficient;
4) unit output constraint
In the formula (I), the compound is shown in the specification,
is a lagrange multiplier;
5) phase angle constraint
In the formula (I), the compound is shown in the specification,
is a lagrange multiplier.
Further, the specific process of solving the double-layer bidding model is as follows:
firstly, expressing a lower layer model problem by using an equivalent KKT condition, linearizing a complementary relaxation condition based on a large M rule, explicitly unifying upper and lower layer variables, changing a double-layer bidding model into a single-layer model, wherein a binary variable product term still exists in the model, linearizing the binary product term by using a binary expansion method, and simultaneously converting a power generation group bidding strategy into a discrete strategy set, so that the subsequent game equilibrium solution is conveniently solved.
Further, when the underlying model is analyzed alone, the variable ke,m,i,kr,m,iA decision variable of an upper layer model is regarded as a constant; therefore, the lower model problem is a quadratic convex optimization problem containing linear constraint, and the decision variables are continuous variables which meet the requirement of the KKT conditionAnd (5) using the product.
Further, the lower model equivalent KKT condition is expressed as:
wherein, the complementary relaxation conditions are expressed in a form that x ^ y is more than or equal to 0 and less than or equal to 0, and a large enough constant M and a 0-1 variable u are introduced for linearization treatment;
at the moment, the double-layer model is converted into a single-layer model, but a binary variable product term still exists in the single-layer model
And k
r,m,ir
m,i,tConsider thatTo k
e,m,i,k
r,m,iThe binary extension method is applied to solve the two problems.
The core idea of binary extension method is to use discrete value set to replace the value range of original continuous variable approximately
For example, first, a continuous variable k is set
e,m,iIs equally divided into
Segment, then k
e,m,iIs approximated as one comprising
A set of discrete points.
Wherein k is
1And
respectively represents k
e,m,iThe infimum and supremum of the original value domain. Each discrete point in the value range is associated with an n
kBit binary number corresponding to the discrete point represented by the binary number
1Is a distance of
x
ji,kThe j-th bit of the binary number is represented,
representing a decimal number corresponding to a binary number. Δ k represents the equally spaced intervals of successive value ranges.
Then
The term can be approximatedIs shown as
0≤yji,k≤Gkxji,k
At this time, the upper layer objective function still has binary variable product terms
And
the lower layer clearing model can know the price of the power at the day-ahead market clearing node
Wherein i
b={(m,i)∈Φ
gThe node where the ith unit in the power generation group m is located is represented by the same method; real-time market clearing node electricity price
Spare service calling electricity price
Consistent for all nodes in the network.
In view of
And r
i,tAnd the infimum and infimum are existed, so that the nonlinear terms existed in the upper layer objective function are linearized by adopting a binary extension method. To be provided with
By way of example, by
The discretization process of (a) may result in relaxed constraint expressions.
0≤yji,t,p≤Gpxji,t,p
Wherein p is
1·Is composed of
Infimum,. DELTA.p is the dividing interval, G
pIs a sufficiently large constant. Of the above only objective function
Is processed to restrict
Still a continuous variable. Therefore, the two results which should be the same inevitably have a difference in decision making, and in order to minimize the difference between the two results and ensure the accuracy of the linearization process, constraints are required to be added so that the error of the decision making result of the power consumption in the continuous type and the discrete type is kept at an extremely low level. The error of the two can be limited within half of the equal division interval, and n is properly increased on the premise of considering the solving speed
pThe error can be further reduced to an acceptable range by the value of (2).
And after the double-layer bidding model is converted into the single-layer model, all nonlinear terms are subjected to linearization treatment, and the single-layer mixed integer linear programming model can be directly solved through a common commercial solver.
When bidding decision is made by the power generation groups, the bidding of competitors needs to be referred to, and competitive game relations exist among the power generation groups. Although a small alliance consisting of a plurality of units is formed inside the power generation group, each unit belongs to a beneficial agent, the problem of benefit distribution does not need to be considered, and the structure of the small alliance is determined and cannot be changed. Thus, in essence, gaming between power generation groups will eventually reach nash equilibrium. In this state, no market entity can rely solely on its own decision to gain greater revenue. Namely, it is
Wherein i represents the subject of interest, r
iOn behalf of the subject arbitrary bidding strategy, r
i *On behalf of the subject's bid balancing strategy,
representing a competitive bid balancing strategy. The formation of nash equilibrium solutions relies on the mechanisms of price adjustment in the market. BAE based round-by-round offers in a spot market environment comprises the following steps:
firstly, before the market is cleared in the day, the power generation group reports an initial quotation curve, unit technical parameters and the like to a dispatching and trading center. And the dispatching and trading center discloses the quotation information of all the power generation groups, the report information of the user side and other related network and unit parameters, and starts the first bidding.
And secondly, each power generation group obtains a new round of quotation and reports the quotation to a dispatching and trading center within a certain time limit according to an initial quotation curve of a competitor and a double-layer bidding model.
And thirdly, the dispatching and trading center obtains a new round of quotation result, and if the quotation of each power generation group is the same as that of the previous round. The game is over and the quoted result in the round is the Nash equilibrium result. If the power generation group adjustment strategy still exists, the dispatching transaction center continuously discloses the quotation condition of the round to the whole society, and carries out bidding of the next round until all the power generation groups do not change the quotation strategy.
Example 2
The embodiment provides a power generation group double-layer game bidding system in a power spot market environment, which comprises an information acquisition module, a bidding analysis module, a model conversion and solving module, a quotation adjustment module and a data output module, as shown in fig. 2.
The information acquisition module acquires the boundary condition information of market clearing; the bidding analysis module forms a power generation group bidding model according to the boundary conditions; the model conversion and solution module simplifies and solves the double-layer bidding model; the quotation adjusting module supports the power generation group to quotation round by round until a balanced solution is achieved; and the data output module formats and outputs the bidding strategy and the market balance point information of the power generation group.
For the part not described in detail in example 2, reference is made to example 1.
Application example
The invention discloses an application of a power generation group double-layer game bidding method in a spot market environment, which comprises the following steps:
1) an IEEE118 node standard system is adopted as a detection system, a reference load and an unoptimizable unit output curve in a typical day is shown in figure 3, and parameters of various types of units on the power generation side are shown in tables 1 and 2.
TABLE 1 optimizable train parameters
TABLE 2 non-optimizable train parameters
2) And constructing a double-layer bidding model of the power generation group participating in the main energy market and the standby service market, simplifying the double-layer model based on a KKT condition and a binary extension method, and finally solving the model through a general commercial solver.
3) And solving a game equilibrium quotation result of the power generation group based on a quotation adjusting mechanism. The results are shown in FIGS. 4-5 and Table 3.
TABLE 3 Game Balanced Bidding result of Power Generation group
Compared with the prior art, the method has the advantages that a double-layer game bidding strategy model of the power generation group in the spot market environment is constructed, the bidding behaviors of the power generation group participating in the main energy market and the standby service market at the same time are analyzed, the mutual restriction relationship between the power generation low-price bidding behavior and the spot market discharge is considered, and a double-layer model simplification and solving method is provided; and designing a quotation adjusting mechanism to obtain the game balanced quotation of the power generation group. The invention makes up the defects of the prior patent on the research of the field of bidding behaviors of members in the spot market, creatively constructs a power generation group bidding model and provides a double-layer bidding model solving method containing a large number of nonlinear terms.