CN1828656A - Electricity-generating terminal contest price decision information obtaining and processing method based on Markov chain - Google Patents

Abstract

The related method comprises: a. collecting the original data of power market on free-wandering state, describing the obtained power of different enterprises on bid time opposite to the Qi[k] in table, and normalizing the market share; b. taking 2D power spectrum analysis to obtain the sample subset satisfied the fluctuation request; c. with Markov chain approach algorithm, calculating the Markov chain; d. based on the obtained Markov chain, explicit expressing information.

Description

Obtain and disposal route based on markovian electricity-generating terminal contest price decision information

Technical field

The present invention is a kind of information acquisition and disposal route that is applied to power industry, belongs to the technical field of soft measurement of information (signal) and processing.

Background technology

Electricity market research classification as shown in Figure 1.Electric energy quotation need be based on a large amount of enterprises self cost, production information, and the current power market information.The goal in research of generating side then is how to obtain information and to be used for the quotation decision-making from electricity market.

Decision-making based on the marginal electricity price of prediction is the electricity market generating side decision-making technique that occurs the earliest, also is present the most frequently used decision-making technique.So the document about marginal electricity price research also is a most extensive class in the electricity market research.By consulting all kinds of documents and induction-arrangement, generally the research of marginal electricity price can be done following classification:

(1) based on the marginal Research on electricity price prediction of marginal electricity price self;

(2) the marginal Research on electricity price prediction of importing based on multidimensional;

(3) reconstruct chaos phase space limit Research on electricity price prediction.

Marginal Research on electricity price prediction based on marginal electricity price self is a kind of marginal Research on electricity price prediction mode that occurs the earliest, and this method is only predicted class methods of current marginal electricity price according to the historical data of marginal electricity price self.These class methods mainly contain: the time series forecasting method ^[1], to neural net method ^[1,4], blur method ^[11], the genetic algorithm method ^[12,13], and the intersection of these methods is used ^[2,3,13]

Marginal Research on electricity price prediction based on the multidimensional input is replenishing preceding a kind of method.Other input that can adopt has the market workload demand, season, and environment temperature, festivals or holidays, situation or the like had the variable of direct or indirect influence to electricity price.

Reconstruct chaos phase space limit Research on electricity price prediction is a kind of relatively independent research method, and many documents about this method are also arranged at present ^[5,6]These class methods are general makes the Lyapunov index analysis, maximum λ to the marginal electricity price data of history earlier ₁(Lyapunov index) greater than zero, the correlation dimension of its phase space attractor is a mark, and then system limit electricity price has chaotic characteristic.Carry out the phase space reconfiguration of electricity price attractor then, calculate the fractal dimension of electricity price attractor, on the basis that embeds phase space, carry out marginal Research on electricity price prediction at last.

Yet said method does not go deep into electricity market mechanism research electricity price and forms mechanism and fluctuation internal factor, so do not obtain the most essential information of reflection electricity market all based on the market process analysis.The research of existing market mechanism is progressively goed deep into, as the proposition of electricity market internal state opinion ^[7], the mechanism research of corporate behavior in the electricity market ^[8], POWER MARKET DYNAMIC ANALYSIS ^[9], preliminary electricity market stability study ^[10]This algorithm promptly under this technical background, is started with from the market internal mechanism, analyzes the market process approximate algorithm, and provides information acquisition algorithm.

The good and bad status of each bidders in market directly influences the market behavior and the market equilibrium state of each enterprise in the electricity market.But these information all are the trade secrets of enterprise, can not directly obtain from the electricity market process.The electricity power market equilibrium state needs the migration of electricity market process much to go on foot (even infinite multistep) could to arrive.And under the situation that can't obtain rival's details, very difficult again the parsing based on the electricity market model calculates equilibrium state.As can be seen, go for market equilibrium state and market price bidding person's quality relation, only the free migration by the electricity market process is infeasible.If can find a kind of computation model that can obtain the electricity power market equilibrium state according to limited step migration sample, will successfully solve an above-mentioned difficult problem.

According to above argumentation, as can be seen, it is not enough only using direct data from the electricity market process to be used to make a strategic decision.The good and bad status of each bidders in market directly influences the market behavior and the market equilibrium state of each enterprise in the electricity market, but these information all are the trade secrets of enterprise, can not directly obtain from the electricity market process.The electricity power market equilibrium state needs electricity market process migration multistep (even infinite multistep) to arrive, and under the situation that can't obtain rival's details, is difficult to based on electricity market model analytical Calculation equilibrium state again.As can be seen, go for market equilibrium state and market price bidding person's quality relation, only the free migration by the electricity market process is unpractiaca.If can find a kind ofly can obtain the computation model of equilibrium state, will successfully solve an above-mentioned difficult problem from limited step migration sample.

Summary of the invention

Technical matters: the purpose of this invention is to provide a kind of electricity-generating terminal contest price decision information and obtain and disposal route based on Markov (Markov) chain, this method can obtain the electricity power market equilibrium state by limited step migration sample, and proposition information obtains rule, obtains the good and bad information in market based on market equilibrium state and Markov chain.

Technical scheme: of the present invention based on markovian electricity-generating terminal contest price decision information obtain and disposal route from limited electricity market sample, approach the electricity market process by the Markov chain, and obtain the market characteristics information of electricity power market equilibrium state, the good and bad relation in market.

Based on markovian electricity-generating terminal contest price decision information obtain and disposal route from limited electricity market sample, approach the electricity market process by the Markov chain, and obtain electricity power market equilibrium state, the good and bad market characteristics information that concerns in market, information is obtained and disposal route is:

A. gather the electricity market raw data---the market distributed data under the free migration state in market, describing respectively bids constantly is that each free migration goes on foot the electric weight that each electricity power enterprise obtains, Q in the corresponding symbol table in the corresponding symbol table _i[k], normalized then is the market share of each enterprise,

B. two-dimensional power spectrum analysis is met the sample subclass that undulatory property requires;

C. calculate the Markov chain with Markov chain approximate algorithm;

D. based on the Markov chain that obtains, the dominance expressing information that is: is obtained rule according to information, obtains information.

Calculating Markov chain method with Markov chain approximate algorithm is:

A. set up Markov chain probability transition matrix: this matrix obtains employing " Markov chain approximate algorithm ";

B. initialization;

C., the algorithm circulation is set uses I=1, represent the call number Index=1 of good and bad index, corresponding array is E[Index], the sequence number of its array content representation enterprise;

If P d. _Ii=1, P _Ij=0, expression enterprise enters the attraction attitude, and it is Index E[Index that the I of enterprise is denoted as good and bad index]=I;

E.Index is from adding 1, for the evaluation of next enterprise is prepared;

F.I is from adding 1, for the evaluation of next enterprise is prepared;

G.M represents the quantity of enterprise, if I＜=M, expression also has enterprise to be evaluated, the evaluation that enters next enterprise;

If I＞M h., this estimates end, attracts the evaluation of attitude enterprise to finish, and enters the evaluation of non-attraction attitude enterprise;

J., I=1 is set, enters circulation;

K. calculate Represent that the market share of other enterprise flows into the probability of the i of enterprise;

If l. Inferior big, represent except that the strong enterprises that estimated,

Maximum enterprise, then the I of enterprise being denoted as good and bad index is Index E[Index]=I;

M.Index is from adding 1, for the evaluation of next enterprise is prepared;

N.I is from adding 1, for the evaluation of next enterprise is prepared;

If o. I＜=M represents to also have enterprise to be evaluated, the evaluation that enters next enterprise;

If I＞M p., this estimates end, and the evaluation of all enterprises finishes.

Two-dimensional power spectrum analysis and sample subclass are chosen, and flow process is:

A., undulatory property threshold value P is set _y, the meaning of this threshold value is the undulatory property upper limit of sample subclass to be chosen, if sample subclass two-dimensional power spectrum to be chosen is greater than threshold value P _y, then do not satisfy undulatory property, if sample subclass two-dimensional power spectrum to be chosen is less than threshold value P _y, then satisfy undulatory property;

B. be provided with and investigate step number N, N represents the capacity of sample subclass;

C. initialization;

D. read sample M, establishing its capacity is M, and this sample data is M step market distributed data under the free migration state of describing the current research electricity market;

If e. M＞N represents this sample size greater than sample subclass capacity to be chosen, enter filtering and sample subclass and select flow process, otherwise withdraw from;

F., I=1 is set, begins to investigate, enter circulation, the undulatory property of verification sample subclass from first sample of sample set;

G. be initial with I, choose N sample continuously as sample subclass to be investigated;

H. according to the formula of formula (4) definition, calculate the two-dimensional power spectrum P of sample subclass;

J. if P＜P _Min, show that the undulatory property of sample subclass of current investigation is littler than the undulatory property of the sample subclass of having investigated, with P _MinReplace with this two-dimensional power spectrum P, call number I _Min=I;

K. if I＜=(M-N), show to still have the sample subclass of not investigating in the sample set that I continues verification from adding;

L. otherwise show that all sample subclass all investigate, withdraw from cycle calculations.Compare P _MinWith P _y, if P _Min＜P _y, there is the sample subclass that satisfies the undulatory property requirement in the expression sample set.This sample subclass is that call number is from I _MinN sample of beginning.

" market process based on the Markov chain is approached " algorithm is at first made analysis of spectrum from limited electricity market sample, obtains to satisfy the sample subclass of undulatory property, estimates (least square method) according to the Markov chain, can obtain to approach the Markov chain of this electricity market process.The Markov chain progressively develops and then can converge to the stationary point that characterizes the electricity power market equilibrium state.Its process as shown in Figure 3.

The Markov chain that actual electric power market process is approached has comprised a lot of market informations, and wherein some information is very directly perceived, as the equilibrium state of electricity market; Also there are some information not directly perceived, need some extra algorithms of design or rule to obtain.The present invention has designed two information and has obtained the probability transfer matrix acquisition enterprise cost good and bad information of rule from the Markov chain.This part algorithm block diagram as shown in Figure 4.

" based on the decision making algorithm of the pre-structure of market price bidding " algorithm is based on the information that obtains, and the pre-structure electricity market process of bidding obtains marginal electricity price.The block diagram of this method as shown in Figure 5.Need to prove that the emphasis of the invention of this paper is to propose a kind of information getting method of decision service that is.On this basis, can select multiple decision-making technique.

P _Ij---the unit of probability transfer matrix, the expression from state i (i=1,2 ..., n) transfer to state j (j=1,2 ..., probability n)

Q _i[k]---unit i (i=1,2 ..., n) k (k=1,2 ...) constantly generated energy

Δ Q _i[k, k+1]---unit i (i=1,2 ..., n) k+1 (k=1,2 ...) constantly generated energy (market share) increment

The market clearance electricity price (marginal electricity price) of λ--electric energy transaction

λ [k]---k (k=1,2 ...) marginal electricity price of the moment

λ ⁰---the marginal electricity price under the balanced attitude

λ '---the marginal electricity price under the lack of balance attitude

_i---cause electricity price fluctuation factor i (i=1,2 ..., n)

P (ω _x, ω _y)---two-dimensional power spectrum

P _y---the threshold value of two-dimensional power spectrum

P _Min---the two-dimensional power spectrum of the sample subclass of undulatory property minimum in the sample set.

Beneficial effect:

Solve electricity market generating side market characteristics information and obtain problem

Electricity market generating side is made up of electricity power enterprise, and electricity power enterprise is the difficult market information, particularly market characteristics information of obtaining in the electricity market process, and this comes a difficult problem to enterprise's valence band of sending a telegram.And this paper proposes based on markovian electricity market information acquisition and disposal route from limited electricity market sample, approaches the electricity market process by the Markov chain, and obtains market characteristics information such as electricity power market equilibrium state, the good and bad relation in market.Preferably resolve an above-mentioned difficult problem.

Theoretical advantage based on the marginal Research on electricity price prediction of the pre-structure in market

Electricity market under the good condition should operate under the equilibrium state, or to equilibrium state migration in an orderly manner.But because some objective factors, market can present undulatory property, even can operate under other state away from balanced attitude.In this state, be the electricity price that pre-structure electricity market under the mode of bidding by white box cannot be obtained as marginal electricity price.This just needs a kind of method, can set up the contact between marginal electricity price under the marginal electricity price and lack of balance attitude under the balanced attitude.

Still treat this problem from the angle of state, market makes the marginal electricity price λ ' under the current state depart from marginal electricity price λ under the balanced attitude because of undulatory property ⁰If, thisly depart from itself and have regularity, then can set up λ ⁰Mapping relations to λ '.

λ′＝f(λ ⁰) (1)

Further, if can set up the factor that causes undulatory property and the electricity price funtcional relationship between departing from, then can expanded type (1):

λ′＝f(λ ⁰， ₁， ₂，…， _n) (2)

In the formula, _i(i=1,2 ..., n) expression causes the factor of undulatory property.

In the actual physics process, all be the factor that causes market fluctuation as market supply and demand ratio, season, temperature, unplanned stoppage in transit etc.Some have also considered these influence factors based on the Forecasting Methodology of marginal electricity price historical data.This prediction mode can be expressed as:

λ′[k+1]＝f(λ′[0，k]， ₁， ₂，…， _n) (3)

K+1 represents next constantly in the formula, and λ ' [0, k] represents marginal electricity price historical data.The just autoregression between the electricity price data of setting up between this, _i(i=1,2 ..., n) just play correcting action.And that the middle description of formula (2) is _i(i=1,2 ..., n) to making electricity price depart from equilibrium value λ ⁰Inherent physical mappings relation.

Description of drawings

Fig. 1 is the synoptic diagram of electricity market research classification,

Fig. 2 is that the electricity market information of Markov chain is obtained and offered and forms strategic process figure,

Fig. 3 is based on the market process approximate algorithm block diagram of Markov chain,

Fig. 4 is the information acquisition algorithm block diagram,

Fig. 5 is each generating merchant's in per step of Markov chain a market share synoptic diagram,

Fig. 6 is a two-dimensional power spectrum filtering algorithm process flow diagram,

Fig. 7 is that the unit and the state of market of probability transfer matrix shifts synoptic diagram.

Embodiment

Of the present invention based on markovian electricity-generating terminal contest price decision information obtain and disposal route from limited electricity market sample, approach the electricity market process by the Markov chain, and obtain the market characteristics information of the good and bad relation of electricity power market equilibrium state, market.

Information is obtained and disposal route is:

A. gather the electricity market raw data---the market distributed data under the free migration state in market, the electric weight that (each free migration step) each electricity power enterprise that respectively bids constantly obtains is described, Q in the corresponding symbol table in the corresponding symbol table _i[k], normalized then is the market share of each enterprise.These data are directly obtained from electricity market (trade center);

B. two-dimensional power spectrum analysis is met the sample subclass that undulatory property requires;

C. calculate the Markov chain with Markov chain approximate algorithm;

D. based on the Markov chain that obtains, the dominance expressing information that is: is obtained rule according to information, obtains information.

The electricity market data are steadily analyzed and spectral analysis method

The market migration raw data that obtains in the electricity market process has certain randomness, by the two-dimensional power spectrum analysis, then can choose and satisfy data interval that undulatory property requires as calculating sample.

Two-dimensional power spectrum filtering algorithm process flow diagram as shown in Figure 6.Among Fig. 6, P _yThe undulatory property threshold value that expression is provided with, N represents the sample subclass capacity desiring to choose, M represents sample size.Calculate through two-dimensional power spectrum, if there is interval (the corresponding P of interval undulatory property minimum _Min) less than threshold value P _y, then select for use this section interval as the sample subclass of calculating the Markov chain.

Two-dimensional power spectrum definition (being the Fourier transform of autocorrelation function)

$P({\mathrm{\ω}}_x,{\mathrm{\ω}}_y)=\underset{{\mathrm{\τ}}_y=-\∞}{\overset{+\∞}{\mathrm{\Σ}}}\underset{{\mathrm{\τ}}_x=-\∞}{\overset{+\∞}{\mathrm{\Σ}}}R({\mathrm{\τ}}_x,{\mathrm{\τ}}_y)e^{-j({\mathrm{\ω}}_x{\mathrm{\ψ}}_x+{\mathrm{\ω}}_y{\mathrm{\τ}}_y)}$

$=\underset{{\mathrm{\τ}}_y=-\∞}{\overset{+\∞}{\mathrm{\Σ}}}\underset{{\mathrm{\τ}}_x=-\∞}{\overset{+\∞}{\mathrm{\Σ}}}z(x,y)z(x+{\mathrm{\τ}}_x,y+{\mathrm{\τ}}_y)e^{-j({\mathrm{\ω}}_x{\mathrm{\ψ}}_x+{\mathrm{\ω}}_y{\mathrm{\τ}}_y)}---\left(4\right)$

$={\left|Z({\mathrm{\ω}}_x,{\mathrm{\ω}}_y)\right|}^2$

In the formula, at numeric field | ω _x|≤π, | ω _y| the Fourier transform of≤π.Z (ω _x, ω _y) be z (x, Fourier transform y).

Two-dimensional power spectrum analysis and sample subclass are chosen flow process:

A., undulatory property threshold value P is set _y, the meaning of this threshold value is the undulatory property upper limit of sample subclass to be chosen, if sample subclass two-dimensional power spectrum to be chosen is greater than threshold value P _y, then do not satisfy undulatory property, if sample subclass two-dimensional power spectrum to be chosen is less than threshold value P _y, then satisfy undulatory property;

B. be provided with and investigate step number N, N represents the capacity of sample subclass;

C. initialization;

D. read sample M, establishing its capacity is M, and this sample data is M step market distributed data under the free migration state of describing the current research electricity market;

If e. M＞N represents this sample size greater than sample subclass capacity to be chosen, enter filtering and sample subclass and select flow process, otherwise withdraw from;

F., I=1 is set, begins to investigate, enter circulation, the undulatory property of verification sample subclass from first sample of sample set;

G. be initial with I, choose N sample continuously as sample subclass to be investigated;

H. according to the formula of formula (4) definition, calculate the two-dimensional power spectrum P of sample subclass;

J. if P＜P _Min, show that the undulatory property of sample subclass of current investigation is littler than the undulatory property of the sample subclass of having investigated, with P _MinReplace with this two-dimensional power spectrum P, call number I _Min=I;

K. if I＜=(M-N), show to still have the sample subclass of not investigating in the sample set that I continues verification from adding;

L. otherwise show that all sample subclass all investigate, withdraw from cycle calculations.Compare P _MinWith P _y, if P _Min＜P _y, there is the sample subclass that satisfies the undulatory property requirement in the expression sample set.This sample subclass is that call number is from I _MinN sample of beginning.

The Markov approximate algorithm

On the basis of the sample subclass that " the electricity market data are steadily analyzed and the analysis of spectrum algorithm " obtains, calculate based on least square method.Sequences y is represented the sample subclass, and the least-squares estimation algorithm is as follows:

$\hat{P}=Y^{-1}M$

In the formula, $Y={\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="A200610039506.900191.png"\; state="\{\{state\}\}">X</figure-callout>}}_1^{\mathrm{\&tau;}}{\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="A200610039506.900191.png"\; state="\{\{state\}\}">X</figure-callout>}}_1,M={\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="A200610039506.900191.png"\; state="\{\{state\}\}">X</figure-callout>}}_1^{\mathrm{\&tau;}}{X}_2.$

${\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="A200610039506.900191.png"\; state="\{\{state\}\}">X</figure-callout>}}_1=\left[\begin{array}{cccc}{y}_1\left(0\right)& y_2\left(0\right)& \&CenterDot;\&CenterDot;\&CenterDot;& y_m\left(0\right)\\ y_1\left(1\right)& y_2\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& y_m\left(1\right)\\ \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;\\ y_1(n-1)& y_2(n-1)& \&CenterDot;\&CenterDot;\&CenterDot;& y_m(n-1)\end{array}\right]---\left(5\right)$

$X_2=\left[\begin{array}{cccc}{y}_1\left(1\right)& y_2\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& y_m\left(1\right)\\ y_1\left(2\right)& y_2\left(2\right)& \&CenterDot;\&CenterDot;\&CenterDot;& y_m\left(2\right)\\ \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;\\ y_1\left(n\right)& y_2\left(n\right)& \&CenterDot;\&CenterDot;\&CenterDot;& y_m\left(n\right)\end{array}\right]---\left(6\right)$

Corresponding this paper, y in the formula _i(k) market share in i the k of electricity power enterprise step of expression, Q in the corresponding symbol table _i[k].

The market information acquisition algorithm

Approach actual electric power market process the Markov chain reaction market essence, comprise the market essential information, but these information are abstract or recessive express, so the rule that proposition information is obtained is expressed information dominance.Calculation process as shown in Figure 8.Among Fig. 8, will calculate good Markov chain, at first obtain and enter the enterprise that attracts attitude in the market, and decide its good and bad index according to the order of sequence according to rule 1 as input; According to rule 2, analyze the unit of Markov chain probability transition matrix then, draw the good and bad index of non-attraction attitude enterprise.P among the figure _IjThe unit of expression Markov chain, E[Index] the good and bad index of Index enterprise of expression.

In conjunction with electric weight, can further investigate the unit and the state of market relations of metastasis of probability transfer matrix, as shown in Figure 7.Among the figure, Q _i[k] (i=1,2 ..., n; K=0,1,2 ...) represent that generating merchant i is in the k market share constantly.

Definition Δ Q _i[k, k+1]=Q _i[k+1]-Q _i[k]

For generating is discussed i at k+1 market share increment constantly.Then have

$\mathrm{\&Delta;}{Q}_i[k,k+1]=\underset{j=1}{\overset{n}{\mathrm{\&Sigma;}}}{p}_{\mathrm{ji}}\&CenterDot;Q_j\left[k\right]-Q_i\left[k\right]---\left(7\right)$

Because it is irrelevant that the plateau of Markov chain and original state are chosen, assumed initial state S ₀Be the state (its mathematical meaning only is discussed) that each generating merchant market share equates, i.e. a Q ₁=Q ₂=...=Q _n

After then passing through the migration of Markov chain multistep, always It is big that big generating merchant's the market share becomes, Little generating merchant's the market share diminishes.This is explanation just, from p _IjMarket flows to as can be seen, Big generating merchant's advantage is big, corresponding Little generating merchant advantage is little.

The rule that is defined as follows at this transition matrix:

[attract attitude definition] for rival i (i=1,2 ..., n), if p _Ii=1, p _Ij=0 (j=1 ... n, j ≠ i), then enter the attraction attitude, corresponding physical state is completely sent out for rival i.

[market information is obtained rule 1] enters and attracts the rival of attitude more excellent than the rival of any non-attraction attitude in the matrix.

[market information is obtained rule 2] can the unit from transition matrix obtain for the quality relation that does not enter between the generating merchant who attracts attitude.For rival j, p _Ij(j=1,2 ..., meaning n) is the probability that the market share of rival i flows to rival j, Big rival's advantage is big.

The idiographic flow that information is obtained is:

A. set up Markov chain probability transition matrix: this matrix obtains employing " Markov chain approximate algorithm ";

B. initialization;

C., the algorithm circulation is set uses I=1, represent the call number Index=1 of good and bad index, corresponding array is E[Index], the sequence number of its array content representation enterprise;

If P d. _Ii=1, P _Ij=0, expression enterprise enters the attraction attitude, and the good and bad index of the I of enterprise is denoted as Index, corresponding E[Index]=I;

E.Index is from adding 1, for the evaluation of next enterprise is prepared;

F.I is from adding 1, for the evaluation of next enterprise is prepared;

G.M represents the quantity of enterprise, if I＜=M, expression also has enterprise to be evaluated, the evaluation that enters next enterprise;

If I＞M h., this estimates end, attracts the evaluation of attitude enterprise to finish, and enters the evaluation of non-attraction attitude enterprise;

J., I=1 is set, enters circulation;

K. calculate

Represent that the market share of other enterprise flows into the probability of the i of enterprise;

If l. Inferior big, represent except that the strong enterprises that estimated,

Maximum enterprise, then the I of enterprise being denoted as good and bad index is Index E[Index]=I;

M.Index is from adding 1, for the evaluation of next enterprise is prepared;

N.I is from adding 1, for the evaluation of next enterprise is prepared;

If o. I＜=M represents to also have enterprise to be evaluated, the evaluation that enters next enterprise;

If I＞M p., this estimates end, and the evaluation of all enterprises finishes.

Claims (3)

1. one kind is obtained and disposal route based on markovian electricity-generating terminal contest price decision information, it is characterized in that from limited electricity market sample, approach the electricity market process by the Markov chain, and obtain electricity power market equilibrium state, the good and bad market characteristics information that concerns in market, information is obtained and disposal route is:

A. gather the electricity market raw data---the market distributed data under the free migration state in market, describing respectively bids constantly is that each free migration goes on foot the electric weight that each electricity power enterprise obtains, Q in the corresponding symbol table in the corresponding symbol table _i[k], normalized then is the market share of each enterprise,

B. two-dimensional power spectrum analysis is met the sample subclass that undulatory property requires;

C. calculate the Markov chain with Markov chain approximate algorithm;

D. based on the Markov chain that obtains, the dominance expressing information that is: is obtained rule according to information, obtains information.

2. according to claim 1 based on markovian electricity market information acquisition and disposal route, it is characterized in that calculating Markov chain method with Markov chain approximate algorithm is:

A. set up Markov chain probability transition matrix: this matrix obtains employing " Markov chain approximate algorithm ";

B. initialization;

C., the algorithm circulation is set uses I=1, represent the call number Index=1 of good and bad index, corresponding array is E[Index], the sequence number of its array content representation enterprise;

If P d. _Ii=1, P _Ij=0, expression enterprise enters the attraction attitude, and it is Index E[Index that the I of enterprise is denoted as good and bad index]=I;

E.Index is from adding 1, for the evaluation of next enterprise is prepared;

F.I is from adding 1, for the evaluation of next enterprise is prepared;

G.M represents the quantity of enterprise, if I＜=M, expression also has enterprise to be evaluated, the evaluation that enters next enterprise;

If I＞M h., this estimates end, attracts the evaluation of attitude enterprise to finish, and enters the evaluation of non-attraction attitude enterprise;

J., I=1 is set, enters circulation;

K. calculate ∑ _J=0 ^mp _IjRepresent that the market share of other enterprise flows into the probability of the i of enterprise;

If ∑ l. _J=0 ^mp _IjInferior big, represent except that the strong enterprises that estimated ∑ _J=0 ^mp _IjMaximum enterprise, then the I of enterprise being denoted as good and bad index is Index E[Index]=I;

M.Index is from adding 1, for the evaluation of next enterprise is prepared;

N.I is from adding 1, for the evaluation of next enterprise is prepared;

If o. I＜=M represents to also have enterprise to be evaluated, the evaluation that enters next enterprise;

If I＞M p., this estimates end, and the evaluation of all enterprises finishes.

3. according to claim 1 based on markovian electricity market information acquisition and disposal route, it is characterized in that two-dimensional power spectrum analysis and sample subclass choose, flow process is:

A., undulatory property threshold value P is set _y, the meaning of this threshold value is the undulatory property upper limit of sample subclass to be chosen, if sample subclass two-dimensional power spectrum to be chosen is greater than threshold value P _y, then do not satisfy undulatory property, if sample subclass two-dimensional power spectrum to be chosen is less than threshold value P _y, then satisfy undulatory property;

B. be provided with and investigate step number N, N represents the capacity of sample subclass;

C. initialization;

D. read sample M, establishing its capacity is M, and this sample data is M step market distributed data under the free migration state of describing the current research electricity market;

If e. M＞N represents this sample size greater than sample subclass capacity to be chosen, enter filtering and sample subclass and select flow process, otherwise withdraw from;

F., I=1 is set, begins to investigate, enter circulation, the undulatory property of verification sample subclass from first sample of sample set;

G. be initial with I, choose N sample continuously as sample subclass to be investigated;

H. according to the formula of formula (4) definition, calculate the two-dimensional power spectrum P of sample subclass;

J. if P＜P _Min, show that the undulatory property of sample subclass of current investigation is littler than the undulatory property of the sample subclass of having investigated, with P _MinReplace with this two-dimensional power spectrum P, call number I _Min=I;

K. if I＜=(M-N), show to still have the sample subclass of not investigating in the sample set that I continues verification from adding;

L. otherwise show that all sample subclass all investigate, withdraw from cycle calculations.Compare P _MinWith P _y, if P _Min＜P _y, there is the sample subclass that satisfies the undulatory property requirement in the expression sample set.This sample subclass is that call number is from I _MinN sample of beginning.

Images