CN102709926A - Rotary hot spare dispatching method in construction of intelligent power grid on basis of relevance vector machine - Google Patents

Abstract

The invention discloses a rotary hot spare dispatching method in the construction of an intelligent power grid on the basis of a relevance vector machine. The invention relates to the rotary hot spare dispatching method in the construction process of the intelligent power grid on the basis of the relevance vector machine and aims to solve the problem that a rotary hot spare is difficult to set for a new energy resource power system to stabilize the power fluctuation of a wind power connected grid, wherein the new energy resources comprise scale wind power and the like. An initialization setting result is transferred into a wind power relevance vector machine forecasting system; a wind power station wind-power acquisition module acquires a measurement value of a wind power of a wind power station in real time and transfers data into the wind power relevance vector machine forecasting system after carrying out data preprocessing; the wind power relevance vector machine forecasting system receives the data and carries out forecasting on the wind power at the moment in the future so as to obtain a forecasting result, i.e. an error band of the wind power value at the moment in the future and the wind power; and the obtained forecasting value and error band are sent into a dispatching controller, wherein the forecasting value is a power generation plan of the wind power station in the future and the power range represented by the error band is rotary hot spare distributed to the wind power station. The rotary hot spare dispatching method is used in the construction of the intelligent power grid.

Description

During intelligent grid is built based on the dispatching method of the rotation stand-by heat of associated vector machine

Technical field

The present invention relates to a kind of dispatching method of the rotation stand-by heat based on the associated vector machine.

Background technology

Face fossil energy human common difficulties such as exhaustion, environmental pollution day by day, develop new forms of energy, lifting traditional energy utilance and development intelligent grids such as wind energy, become the basic Consensus and the countermeasure of countries in the world.At present, along with reaching its maturity of generations of electricity by new energy such as wind-powered electricity generation technology, the safe and efficient development and use that new forms of energy electric power such as China's scale wind-powered electricity generation are incorporated into the power networks just become one of most critical issue of current urgent need solution.Wherein, primary link need be developed relatively accurate effective wind power prediction system exactly, in the power supply scheduling that is applied in electric power system, realizes the safe and efficient utilization to new forms of energy electric power.

Though China's wind-resources rich because new forms of energy such as wind have characteristics such as randomness, strong fluctuation and uncertainty, makes the active power of output of wind-powered electricity generation unit have certain uncertainty; Simultaneously, because wind energy has different Regional Distribution characteristics, therefore, the generated output of wind energy turbine set has certain region distribution character.So; Even if the single-machine capacity of present wind turbine generator has developed into the MW class level; But existing wind power prediction technology still is difficult to satisfy demands such as the efficient operation of power system security; Cause day balance of electric power and ener and power supply arrangement behind the scale wind-electricity integration very difficult, the traffic control of new forms of energy electric power system faces huge test.According to incompletely statistics, wind-powered electricity generation online purchase electric weight was 222.54 hundred million kilowatt hours to June in 2010 1, and not purchasing electric weight is 27.76 hundred million kilowatt hours, and loss wind energy ratio is 11.1%.And along with increasing substantially of wind-powered electricity generation installed capacity, scale new forms of energy electric power is dissolved the problem and the contradiction that face will be more outstanding.Another important restraining factors that China's large-scale wind power is incorporated into the power networks are that power supply architecture property is particularly thorny, the deficient compared with developed countries controllability power supply that can stabilize new forms of energy electric power random fluctuation characteristic.With 2010 be example, in total capacity of installed generator of China, thermoelectricity accounts for 73.44%, water power accounts for 22.18%, nuclear power accounts for 1.12%, combustion gas and fuel oil capacity of installed generator proportion are then less than 0.3%.Therefore, it is very low to respond combustion gas and the fuel oil generating proportion of fluctuation power supplys such as wind-powered electricity generation fast; And mostly China's water power is the plant without storage, has obvious seasonal characteristics, and the retaining in the reservoir also need be satisfied needs such as flood control and field irrigation except generating, and the hydroelectric peak ability also is restricted.And,, in electrical network, also have certain Regional Distribution characteristic at power configuration capacity such as thermoelectricity, water power, fuel engine power generation and wind-powered electricity generations.For the safety that realizes new forms of energy electric power such as scale wind-powered electricity generation is incorporated into the power networks, need utilize other scalable power supplys of new forms of energy power supply periphery such as wind-powered electricity generation to carry out the development and use pattern of multi-energy complementation, uncertain with the random fluctuation of stabilizing new forms of energy electric power such as wind-powered electricity generation.Therefore; Predicting Technique level in view of present new forms of energy electric power such as wind-powered electricity generation; Must cause abandoning on a large scale wind problem and the setting of redundant rotation stand-by heat according to routine development and use pattern; The overall grid efficiency of transmission is low, even influences problems such as system safety stable operation, causes the situation that development cost is high, utilance is low.Simultaneously, along with the continuous increase of wind-powered electricity generation installed capacity, above-mentioned contradiction will more highlight, and will cause paying future bigger cost to solve difficulty and the problem that the scale wind-electricity integration is caused.

Summary of the invention

The purpose of this invention is to provide a kind of intelligent grid build in based on the dispatching method of the rotation stand-by heat of associated vector machine, contain new forms of energy electric power system such as scale wind-powered electricity generation with solution and be difficult to be provided with the rotation stand-by heat to stabilize wind-electricity integration power fluctuation problem.

The present invention adopts the artificial intelligence Forecasting Methodology of associated vector machine; Based on historical data to the model training; Obtain following predicted value and the corresponding probability distribution of predicted value constantly simultaneously; Provide the wind variable power scope (predicated error band) under certain fiducial probability, confirm reasonably rotation stand-by heat to instruct electrical network.Thereby this method is expected also that spinning reserve setting problem rationally provides a kind of new approaches in the new forms of energy electric power systems such as scale wind-powered electricity generation in order to contain in the following intelligent grid construction.

The present invention addresses the above problem the technical scheme of taking to be:

One, initialization; According to local climate condition and geographic factor; Carry out local wind-resources assessment; According to the wind-resources assessment result computing controller is carried out the initialization setting, the parameter of required setting is the distribution form that predicts the outcome, and said distribution form is set at normal distribution; Said initialization is provided with the result is delivered in the wind power associated vector machine prognoses system, initialization finishes;

Two, transfer of data; Wind energy turbine set wind power collection module is gathered the measured value of wind energy turbine set wind power in real time, after the data preliminary treatment, data passes is given in the wind power associated vector machine prognoses system;

Three, the wind power in the moment in future is predicted; Wind power associated vector machine prognoses system 3 receives the data that passed over by wind energy turbine set wind power collection module; Wind performance number according to current time; Wind power to the moment in future predicts that prediction result is the wind performance number in the following moment and the error band of wind power; The implementation of the error band of wind performance number and wind power is following:

Set

Form by wind power historical data, in the expression training set n historical sample arranged, wherein each historical sample x _i10 dimensions are arranged, form x by the wind performance number in 10 continuous moment _i=[P _I1, P _I2... P _I10]; t _iBe sample x _iNext wind performance number constantly, i.e. t _i=P _I11For this model, promptly model is input as x _i, model is output as t _i, promptly according to the wind performance number in current 10 moment, predicting next wind performance number constantly, n is total number of samples;

More than inciting somebody to action

With

As the training of training sample set pair associated vector machine model,

For by t _iThe set of forming is formed t by n sample _iWith x _iCorresponding one by one;

Three (one), the training process of associated vector machine model is:

1) selected kernel function K is a gaussian kernel function,

X wherein ₁, x ₂Be the vector of two equal in length, δ sets up on their own for the nuclear width;

2) predictive equation does $t(x;w)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{w}_{i}K(x,x_i)+w_0---\left(1\right)$

W=[w wherein ₀, w ₁... W _n] expression weight coefficient vector;

3) establish model predication value and satisfy the normal distribution form, be i.e. p (t _i)=N (t _i| t (xi; W), σ ²) (2)

Wherein p representes distribution function, and N representes normal distribution; Above-mentioned formula (2) expression predicts the outcome and is (the x of t as a result with formula (1) _i; W) be average, σ ²Normal distribution form for variance;

4) need definite parameter to be merely the variances sigma of weight coefficient vector w and normal distribution in formula (1) and the formula (2) ², it is 0 that associated vector machine method hypothesis weight coefficient vector w satisfies average, variance does

Normal distribution, the form of formulate is promptly

$p\left(w_i\right|{\mathrm{\α}}_i)=N\left(w_i\right|0,{\mathrm{\α}}_i^{-1})$

α=[α ₀,α ₁，......,α _n] ^T

$p\left(w\right|\mathrm{\&alpha;})=\underset{i=0}{\overset{n}{\mathrm{\&Pi;}}}\frac{{\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="HDA00001819471200021.png"\; state="\{\{state\}\}">i</figure-callout>}}}{\sqrt{2\mathrm{\&pi;}}}\exp (-\frac{{\mathrm{\&alpha;}}_{\mathrm{<figure-callout\; id="2"\; label="i\; w\; i"\; filenames="HDA00001819471200021.png"\; state="\{\{state\}\}">i</figure-callout>}}{w_i}^{}{{}_{}}^2}{2})$

Parameter through needing to confirm after the above-mentioned variable replacement is merely vectorial α=[α ₀, α ₁... α _n] and variances sigma ²

5) use numerical solution to find the solution α and σ ², provide α and σ earlier ²The conjecture value, upgrade according to following formula then:

${\mathrm{\&alpha;}}_i^{\mathrm{NEW}}=\frac{{\mathrm{\&gamma;}}_i}{{{\mathrm{\&mu;}}_i}^2}$

${\left({\mathrm{\&sigma;}}^2\right)}^{\mathrm{NEW}}=\frac{{\left|\right|t-\mathrm{\&Phi;\&mu;}\left|\right|}^2}{n-\underset{i=0}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&gamma;}}_i}$

γ ₁=1-α _iB(i,i)

γ wherein _iBe intermediate variable, Φ is the matrix that constitutes with kernel function, and concrete form is as follows:

$\mathrm{\&Phi;}=\left[\begin{array}{ccccc}1& K(x_1,x_1)& K(x_1,x_2)& ...& K(x_1,x_n)\\ 1& K(x_2,x_1)& K(x_2,x_2)& ...& K(x_2,x_n)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ 1& K(x_n,x_1)& K(x_n,x_2)& ...& K(x_n,x_n)\end{array}\right]$

Matrix A is for being the square formation of diagonal entry with vectorial α, matrix B=(σ ^-2Φ ^TΦ+A) ^-1,

(i i) is the i item element on the diagonal among the B, vectorial μ=σ to B ^-2B Φ ^TWhat t, T represented is that matrix is carried out matrix transpose operation, after abundant renewal, and most α _iThe meeting convergence is infinitely great, the w that meaning is promptly corresponding _i Be 0; Other α _iCan stablize the convergence finite value, corresponding with it x _iJust be called associated vector, obtaining α and σ through above-mentioned iterative process ²After, the training process of model finishes;

Three (two), get into the prediction link; Given input x _*Be the wind power sequence (the wind performance number that comprises ten moment) of current time, according to formula (1) and formula (2) model output wind power prediction value t _*With and the probability distribution of predicted value, obtained predicted value and error band thus;

Four, be sent in the scheduling controller by the predicted value and the error band that obtain in the wind power associated vector machine prognoses system; Predicted value is following generation schedule of wind energy turbine set; The power bracket of error band representative is the rotation stand-by heat that wind energy turbine set is equipped with, and scheduling controller will control the rotation stand-by heat and generate electricity according to instruction and supply user's use.

The invention has the beneficial effects as follows:

The present invention will provide wind power prediction result comparatively accurately, when providing the deterministic forecast result, can provide probabilistic and predict the outcome, and provide wind variable power scope and predicated error band under certain fiducial probability.To foundation be provided to dispatching of power netwoks like this, the rotation stand-by heat in the electrical network rationally will be set, satisfy the needs of the safe and highly efficient operation of scale wind-electricity integration, satisfy the target that intelligent grid is built.

Innovative point of the present invention is in particular in the following aspects:

1, first the intelligent algorithm of associated vector machine is incorporated in the power prediction system of new forms of energy power supplys such as wind-powered electricity generation.

2, be directed against the development and utilization situation of current China wind power prediction system; Propose to use associated vector machine method when obtaining following predicted value (being the certainty composition of power fluctuation) constantly and the corresponding probability distribution of predicted value; Provide the wind variable power scope (being the bandwidth of uncertain error) under certain fiducial probability; So that provide effective information as the foundation that is provided with and stand-by heat is rotated in scheduling to electrical network; In the hope of realizing the safe and highly efficient operation of scale wind-electricity integration, reach the purpose of the reasonably optimizing utilization of new forms of energy maximum using and traditional resource.

3, can effectively utilize the prior informations such as local climate, power supply type and capacity of wind energy turbine set, the convenient different resource situation that adapts to different regions forms optimum scheduling method.

4, the present invention method that can introduce probabilistic prediction and the scheduling of rotation stand-by heat of prior information be injected into the new forms of energy power supply be incorporated into the power networks power fluctuation the uncertainty prediction and stabilize, for realizing in China's intelligent grid construction that from now on the safe and highly efficient operation that uncertain at random power supplys such as large-scale wind power, photoelectricity are incorporated into the power networks provides a kind of new approaches.

Advantage of the present invention is: 1. introduce the prior informations such as local environment and weather of wind energy turbine set based on the artificial intelligence approach of associated vector machine, can provide the deterministic forecast result of relative degree of precision; 2. when providing the deterministic forecast result, can also provide probabilistic and predict the outcome, provide the wind variable power scope (being the predicated error band) under certain fiducial probability; Simultaneously, the controllability power supply type and the capacity of wind energy turbine set periphery introduced in combination again, so that setting is provided and dispatches the foundation of rotating stand-by heat to electrical network; 3. predicted value and uncertain boundary estimate accurately; Not only can realize reducing significantly the risk that wind energy turbine set is abandoned wind; Also avoided existing in the electrical network problem of a large amount of rotation stand-by heats simultaneously; The fail safe that reduces the cost of wind-electricity integration significantly and improve electrical network realizes the safe and efficient utilization of scale wind-electricity integration; 4. realized to introduce during intelligent grid that the artificial intelligence prediction of probabilistic of prior information and method that the rotation stand-by heat is provided with scheduling be injected into China builds, for the safe and highly efficient operation that is incorporated into the power networks at uncertain at random power supplys such as realizing large-scale wind power, photoelectricity from now on provides a kind of new way.

Description of drawings

Fig. 1 carries out the wind power prediction sketch map for using method of the present invention; Fig. 2 carries out wind power prediction result's curve chart for using method of the present invention, and abscissa express time among the figure, ordinate are represented wind power; Fig. 3 is based on the dispatching method schematic diagram of the rotation stand-by heat of associated vector machine during intelligent grid of the present invention is built.

Embodiment

Embodiment one: combine Fig. 1 to Fig. 3 to explain, based on the dispatching method of the rotation stand-by heat of associated vector machine, said method was achieved in that during the intelligent grid of this execution mode was built

One, initialization; According to local climate condition and geographic factor; Carry out local wind-resources assessment; According to the wind-resources assessment result computing controller 3 is carried out the initialization setting, the parameter of required setting is the distribution form that predicts the outcome, and said distribution form is set at normal distribution; Said initialization is provided with the result is delivered in the wind power associated vector machine prognoses system 3, initialization finishes;

Two, transfer of data; Wind energy turbine set wind power collection module 1 is gathered the measured value of wind energy turbine set wind power in real time, after the data preliminary treatment, data passes is given in the wind power associated vector machine prognoses system 3;

Three, the wind power in the moment in future is predicted; Wind power associated vector machine prognoses system 3 receives the data that passed over by wind energy turbine set wind power collection module 1; Wind performance number according to current time; Wind power to the moment in future predicts that prediction result is the wind performance number in the following moment and the error band of wind power; The implementation of the error band of wind performance number and wind power is following:

Set

Form by wind power historical data, in the expression training set n historical sample arranged, wherein each historical sample x _i10 dimensions are arranged, form x by the wind performance number in 10 continuous moment _i=[P _I1, P _I2... P _I10]; t _iBe sample x _iNext wind performance number constantly, i.e. t _i=P _I11For this model, promptly model is input as x _i, model is output as t _i, promptly according to the wind performance number in current 10 moment, predicting next wind performance number constantly, n is total number of samples;

More than inciting somebody to action

With

As the training of training sample set pair associated vector machine model,

For by t _iThe set of forming is formed t by n sample _iWith x _iCorresponding one by one;

Three (one), the training process of associated vector machine model is:

1) selected kernel function K is a gaussian kernel function,

X wherein ₁, x ₂Be the vector of two equal in length, δ sets up on their own for the nuclear width;

2) predictive equation does $t(x;w)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_{i}K(x,x_i)+w_0---\left(1\right)$

W=[w wherein ₀, w ₁... W _n] expression weight coefficient vector; w _oBe the undetermined parameter in the formula (1);

3) establish model predication value and satisfy the normal distribution form, be i.e. p (t _i)=N (t _i| t (x _i; W), σ ²) (2)

Wherein p representes distribution function, and N representes normal distribution; Above-mentioned formula (2) expression predicts the outcome and is (the x of t as a result with formula (1) _i; W) be average, σ ²Normal distribution form for variance;

4) need definite parameter to be merely the variances sigma of weight coefficient vector w and normal distribution in formula (1) and the formula (2) ², it is 0 that associated vector machine method hypothesis weight coefficient vector w satisfies average, variance does Normal distribution, the form of formulate is promptly

$p\left(w_i\right|{\mathrm{\&alpha;}}_i)=N\left(w_i\right|0,{\mathrm{\&alpha;}}_i^{-1})$

α=[α ₀,α ₁,......,α _n] ^T

$p\left(w\right|\mathrm{\&alpha;})=\underset{i=0}{\overset{n}{\mathrm{\&Pi;}}}\frac{{\mathrm{\&alpha;}}_i}{\sqrt{2\mathrm{\&pi;}}}\exp (-\frac{{\mathrm{\&alpha;}}_i{w_{\mathrm{<figure-callout\; id="2"\; label="i"\; filenames="HDA00001819471200021.png"\; state="\{\{state\}\}">i</figure-callout>}}}^2}{2})$

Parameter through needing to confirm after the above-mentioned variable replacement is merely vectorial α=[α ₀, α ₁... α _n] and variances sigma ²

6) use numerical solution to find the solution α and σ ², provide α and σ earlier ²The conjecture value, upgrade according to following formula then:

${\mathrm{\&alpha;}}_i^{\mathrm{NEW}}=\frac{{\mathrm{\&gamma;}}_i}{{{\mathrm{\&mu;}}_i}^2}$

${\left({\mathrm{\&sigma;}}^2\right)}^{\mathrm{NEW}}=\frac{{\left|\right|t-\mathrm{\&Phi;\&mu;}\left|\right|}^2}{n-\underset{i=0}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&gamma;}}_i}$

γ ₁=1-α _iB(i,i)

Wherein γ i is an intermediate variable, and Φ is the matrix that constitutes with kernel function, and concrete form is as follows:

$\mathrm{\&Phi;}=\left[\begin{array}{ccccc}1& K(x_1,x_1)& K(x_1,x_2)& ...& K(x_1,x_n)\\ 1& K(x_2,x_1)& K(x_2,x_2)& ...& K(x_2,x_n)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ 1& K(x_n,x_1)& K(x_n,x_2)& ...& K(x_n,x_n)\end{array}\right]$

Matrix A is for being the square formation of diagonal entry with vectorial α, matrix B=(σ ^-2Φ ^TΦ+A) ^-1,

(i i) is the i item element on the diagonal among the B, vectorial μ=σ to B ^-2B Φ ^TWhat t, T represented is that matrix is carried out matrix transpose operation, after abundant renewal, and most α _iThe meeting convergence is infinitely great, the w that meaning is promptly corresponding _iBe 0; Other α _iCan stablize the convergence finite value, corresponding with it x _iJust be called associated vector, obtaining α and σ through above-mentioned iterative process ²After, the training process of model finishes;

Three (two), get into the prediction link; Given input x _*Be the wind power sequence (the wind performance number that comprises ten moment) of current time, according to formula (1) and formula (2) model output wind power prediction value t _*With and the probability distribution of predicted value, obtained predicted value and error band thus;

With certain wind energy turbine set data instance, use said method to predict that the result is as shown in Figure 3, wherein solid line is represented true wind power curve, the wind power curve that dotted line representative prediction obtains, the error band that dotted line representative prediction obtains.

Four, be sent in the scheduling controller 4 by the predicted value and the error band that obtain in the wind power associated vector machine prognoses system 3; Predicted value is following generation schedule of wind energy turbine set; The power bracket of error band representative is the rotation stand-by heat 5 that wind energy turbine set is equipped with, and scheduling controller 4 will control rotation stand-by heat 5 and generate electricity according to instruction and supply user's use.

Rotation stand-by heat 5 is meant when electrical network needs subsequent use the exerting oneself that can employ immediately at any time.The formulation of control strategy at first will be considered according to the prior information that wind energy turbine set periphery power supply is arranged; Be that which stand-by power supply (fired power generating unit the wind energy turbine set periphery has; Hydropower Unit, gas turbine unit, energy storage device); When the use that is rotated subsequent use 5, preferentially select the fast power supplys of response speed such as gas turbine unit, use other power supplys simultaneously to satisfy the requirement of reserve level (being power demand).

The fluctuation power supply of mains side is less in the electric power system of traditional energy, mainly is that the load center side wave is moving bigger, and therefore, the power output of controllable electric power that can be through the control mains side in traditional electrical network is kept the overall operation balance and the optimization of electrical network; And current greatly developing owing to new forms of energy electric power such as wind power generations; Being incorporated into the power networks of new forms of energy such as scale wind-powered electricity generation also can be increasing to the influence of electrical network; Existing problem is to have certain fluctuation in the mains side power output in the current electrical network; And still there is fluctuation in load side; Therefore, to how accurate and effective is predicted the problem of the fluctuation of following wind power as the dispatching of power netwoks foundation, we have proposed to utilize the method for associated vector machine to predict the uncertainty of wind-electricity integration power fluctuation in the scale wind field; So that the rotation stand-by heat 5 in the electrical network rationally to be set, and then be expected to solve the power prediction problem when being incorporated into the power networks such as other fluctuation type power supply such as photoelectricity in the following intelligent grid construction.

The present invention is directed to the new forms of energy electric power system that contains the scale wind-powered electricity generation and be difficult to the random fluctuation uncertainty that rational management rotation stand-by heat 5 is stabilized wind-electricity integration power; The present invention proposes to be applied to based on the wind power probability forecasting method of associated vector machine in the intelligent grid construction; The uncertainty that predicts the outcome is estimated to be that the predicated error band offers electrical network, as the foundation that reasonable rotation stand-by heat is set.The present invention is based on the thought of associated vector machine (RVM); The probability distribution that provides certainty wind power prediction result simultaneously and predict the outcome; Provide the scope of the wind power random fluctuation under certain confidential interval; As the dispatching of power netwoks foundation, be beneficial to the efficient operation of the whole electrical network in wind power networking back, reach and build the purpose that intelligent grid is made rational use of resources.

The present invention is directed to the new forms of energy electric power system that contains the scale wind-powered electricity generation and be difficult to be provided with and dispatch rational rotation stand-by heat to stabilize the uncertain problem of wind-electricity integration power fluctuation; Adopt the artificial intelligence Forecasting Methodology of associated vector machine; To the model training, obtain following predicted value and the corresponding probability distribution of predicted value constantly based on historical data simultaneously, as shown in Figure 1; P* is the predicted power curve that obtains; Δ is the wind variable power scope under certain fiducial probability, and p is the distribution function of predicted value, can rationally be provided with according to prior informations such as the local environment of wind energy turbine set and weathers.Based on above result, combine prior informations such as peripheral controllability power supply type of wind energy turbine set and capacity again, can instruct electrical network to confirm suitable rotation stand-by heat, be illustrated in fig. 2 shown below.Thereby this method is expected for the reasonable setting and the scheduling problem of spinning reserve in the new forms of energy electric power system that contains the scale wind-powered electricity generation a kind of new approaches to be provided.

Claims (1)

1. based on the dispatching method of the rotation stand-by heat of associated vector machine, it is characterized in that: said method was achieved in that during an intelligent grid was built

One, initialization; According to local climate condition and geographic factor; Carry out local wind-resources assessment; According to the wind-resources assessment result computing controller (3) is carried out the initialization setting, the parameter of required setting is the distribution form that predicts the outcome, and said distribution form is set at normal distribution; Said initialization is provided with the result is delivered in the wind power associated vector machine prognoses system (3), initialization finishes;

Two, transfer of data; Wind energy turbine set wind power collection module (1) is gathered the measured value of wind energy turbine set wind power in real time, after the data preliminary treatment, data passes is given in the wind power associated vector machine prognoses system (3);

Three, the wind power in the moment in future is predicted; Wind power associated vector machine prognoses system (3) receives the data that passed over by wind energy turbine set wind power collection module (1); Wind performance number according to current time; Wind power to the moment in future predicts that prediction result is the wind performance number in the following moment and the error band of wind power; The implementation of the error band of wind performance number and wind power is following:

Set

Form by wind power historical data, in the expression training set n historical sample arranged, wherein each historical sample x _i10 dimensions are arranged, form x by the wind performance number in 10 continuous moment _i=[P _I1, P _I2... P _I10]; t _iBe sample x _iNext wind performance number constantly, i.e. t _i=P _I11For this model, promptly model is input as x _i, model is output as t _i, promptly according to the wind performance number in current 10 moment, predicting next wind performance number constantly, n is total number of samples;

More than inciting somebody to action With

As the training of training sample set pair associated vector machine model,

For by t _iThe set of forming is formed t by n sample _iWith x _iCorresponding one by one;

Three (one), the training process of associated vector machine model is:

1) selected kernel function K is a gaussian kernel function, X wherein ₁, x ₂Be the vector of two equal in length, δ sets up on their own for the nuclear width;

2) predictive equation does $t(x;w)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{w}_{i}K(x,x_i)+w_0---\left(1\right)$

W=[w wherein ₀, w ₁... W _n] expression weight coefficient vector;

3) establish model predication value and satisfy the normal distribution form, be i.e. p (t _i)=N (t _i| t (xi; W), σ ²) (2)

Wherein p representes distribution function, and N representes normal distribution; Above-mentioned formula (2) expression predicts the outcome and is (the x of t as a result with formula (1) _i; W) be average, σ ²Normal distribution form for variance;

4) need definite parameter to be merely the variances sigma of weight coefficient vector w and normal distribution in formula (1) and the formula (2) ², it is 0 that associated vector machine method hypothesis weight coefficient vector w satisfies average, variance does

Normal distribution, the form of formulate is promptly

$p\left(w_i\right|{\mathrm{\&alpha;}}_i)=N\left(w_i\right|0,{\mathrm{\&alpha;}}_i^{-1})$

α=[α ₀,α ₁,......,α _n] ^T

$p\left(w\right|\mathrm{\&alpha;})=\underset{i=0}{\overset{n}{\mathrm{\&Pi;}}}\frac{{\mathrm{\&alpha;}}_i}{\sqrt{2\mathrm{\&pi;}}}\exp (-\frac{{\mathrm{\&alpha;}}_i{w_i}^2}{2})$

Parameter through needing to confirm after the above-mentioned variable replacement is merely vectorial α=[α ₀, α ₁... α _n] and variances sigma ²

5) use numerical solution to find the solution α and σ ², provide α and σ earlier ²The conjecture value, upgrade according to following formula then:

${\mathrm{\&alpha;}}_i^{\mathrm{NEW}}=\frac{{\mathrm{\&gamma;}}_i}{{{\mathrm{\&mu;}}_i}^2}$

${\left({\mathrm{\&sigma;}}^2\right)}^{\mathrm{NEW}}=\frac{{\left|\right|t-\mathrm{\&Phi;\&mu;}\left|\right|}^2}{n-\underset{i=0}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&gamma;}}_i}$

γ ₁=1-α _iB(i,i)

Wherein γ i is an intermediate variable, and Φ is the matrix that constitutes with kernel function, and concrete form is as follows:

$\mathrm{\&Phi;}=\left[\begin{array}{ccccc}1& K(x_1,x_1)& K(x_1,x_2)& ...& K(x_1,x_n)\\ 1& K(x_2,x_1)& K(x_2,x_2)& ...& K(x_2,x_n)\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ 1& K(x_n,x_1)& K(x_n,x_2)& ...& K(x_n,x_n)\end{array}\right]$

Matrix A is for being the square formation of diagonal entry with vectorial α, matrix B=(σ ^-2Φ ^TΦ+A) ^-1, (i i) is the i item element on the diagonal among the B, vectorial μ=σ to B ^-2B Φ ^TWhat t, T represented is that matrix is carried out matrix transpose operation, after abundant renewal, and most α _iThe meeting convergence is infinitely great, the w that meaning is promptly corresponding _iBe 0; Other α _iCan stablize the convergence finite value, corresponding with it x _iJust be called associated vector, obtaining α and σ through above-mentioned iterative process ²After, the training process of model finishes;

Three (two), get into the prediction link; Given input x _*Be the wind power sequence (the wind performance number that comprises ten moment) of current time, according to formula (1) and formula (2) model output wind power prediction value t _*With and the probability distribution of predicted value, obtained predicted value and error band thus;

Four, be sent in the scheduling controller (4) by predicted value that obtains in the wind power associated vector machine prognoses system (3) and error band; Predicted value is following generation schedule of wind energy turbine set; The power bracket of error band representative is the rotation stand-by heat (5) that wind energy turbine set is equipped with, and scheduling controller (4) will control rotation stand-by heat (5) and generate electricity according to instruction and supply user's use.

Images