CN106204332A - A kind of photovoltaic plant efficiency decay Forecasting Methodology - Google Patents

Abstract

The invention discloses a kind of photovoltaic plant efficiency decay Forecasting Methodology, first choose actual measurement power station running environment parameter and output as actual measurement sample；Set up power station simulation platform, actual measurement sample environment parameter is input in emulation platform, obtains the simulation sample under respective environment parameter；Regard two groups of sample power as signal source, respectively measured power signal and simulated power signal are carried out denoising respectively, calculate the power station delivery efficiency EFF value in corresponding moment by the data after denoising；Gained EFF value sequence EMD method is decomposed, by all through EMD method decompose after EFF value save as historical series；Sample drawn point from historical series, is entered into efficiency decay forecast model, obtains predictive value.The present invention compensate for the prediction deficiency that photovoltaic plant efficiency under different open-air conditions is decayed by existing model, the beneficially Market Operation of photovoltaic plant, is conducive to realizing further the prognostic and health management technology of photovoltaic plant.

Description

A kind of photovoltaic plant efficiency decay Forecasting Methodology

Technical field

The present invention relates to a kind of photovoltaic plant efficiency decay Forecasting Methodology, belong to technical field of photovoltaic power generation.

Background technology

Along with developing rapidly of photovoltaic power generation technology, the whole nation is built or increasing in the photovoltaic plant quantity built, scale The hugest, but the problem during operation the most constantly highlights.

First, photovoltaic industry downstream race is fierce, and power station operation is undesirable.Along with photovoltaic products price continues to fall, enterprise Industry invests industry downstream eye one after another, seizes photovoltaic plant market.But, during the fierce market competition, part fortune Enterprise of battalion is eliminated unavoidably, thus causes photovoltaic plant transfer problem.Photovoltaic plant in running, its assembly, electricity Cable, inverter and other relevant spare and accessory parts can occur that efficiency declines problem, i.e. efficiency attenuation problem over time.Particularly assembly Decay, can further result in whole array and produce degradation problems under secondary efficiency because of component mismatch, make the residue of assembly Economic worth reduces.Owing to lacking effective photovoltaic plant efficiency decay appraisal procedure, during operation enterprise substitutes, light There is many difficulties in the actual transfer of overhead utility.

Secondly, photovoltaic system running status is assessed and is not considered the shadow of photovoltaic plant attenuation factor in fault diagnosis model Ring.Existing already present photovoltaic system evaluation mainly has physical method, statistical method and artificial intelligence approach with forecast model.Physics The photovoltaic effect of photovoltaic cell is equivalent to diode circuit model by method.By the modeling of this circuit model, photovoltaic electric is described The duty in pond, and then set up photovoltaic system model for system evaluation.Statistical method is by the statistics to historical data And analysis, set up and describe the forecast model of relation between meteorological factor variable and photovoltaic power generation system output power.Artificial intelligence Method, according to historical events sequence, is trained by certain learning rules, sets up prediction meteorological factor variable and exports with photovoltaic power Between relation.Above-mentioned three kinds of methods all do not consider system physical efficiency attenuation factor, with actual measurement when assessing running efficiency of system It is data when 100% that the health data of Data Comparison is supposition photovoltaic system efficiency.Owing to photovoltaic plant effect cannot be grasped Rate downward trend and time, it is difficult to it is further proposed that photovoltaic plant performance estimating method accurately and reliably, cause photovoltaic plant to be transported After row a period of time, it is impossible to continue to operate in optimum state, reduce investor income.Also it is unfavorable for realizing photovoltaic electric further The prognostic and health management stood.Therefore, it is necessary to propose the Forecasting Methodology of a kind of efficiency of plant decay, it was predicted that efficiency of plant Decline, and then realize photovoltaic plant prognostic and health management, it is ensured that photovoltaic system is safe and reliable, create greatest benefit.

Summary of the invention

For not enough present on prior art, it is an object of the invention to the decay of the efficiency to photovoltaic system and be predicted, To assess the photovoltaic system surplus value and to make system obtain maximum return, a kind of open photovoltaic plant residue EFFICIENCY PREDICTION method.

To achieve these goals, the present invention is to realize by the following technical solutions:

A kind of photovoltaic plant efficiency decay Forecasting Methodology, it is characterised in that comprise the following steps:

A. at interval of one group of power station running environment ginseng of actual measurement in a minute in good weather day 8:00AM to 4:00PM segment Number and output are as actual measurement sample.Set up power station simulation platform, actual measurement sample environment parameter is input to power station simulation and puts down In platform, obtain simulation sample.Regard two kinds of sample datas as signal source, respectively measured power signal and simulated power signal are entered Row denoising, by the EFF value (power station delivery efficiency) in the moment corresponding to simulation sample power calculation of the actual measurement sample power after denoising.

B. with week for unit of time by EFF value sequence EMD (Empirical Mode Decomposition, experience Mode decomposition) method carries out pretreatment, and obtain it and decline percentage ratio D_EFFAs chain rate correction factor, and use it for next The EFF value in week calculates.

C. gained EFF value in step b is saved as historical series.The time is reasonably selected according to actual prediction time limit length Interval, chooses EFF sample point from historical series at timed intervals, is calculated by efficiency decay forecast model, obtains final Predictive value.

In above-mentioned steps a, good weather pattern includes: fine day, irradiation weather clear to cloudy, low.

In above-mentioned steps a, measured power signal uses wavelet method with the denoising method of simulated power signal, its relevant ginseng Number is chosen as follows:

Owing to station output is relatively big by such environmental effects, actual measurement sample fluctuation per minute is relatively big, therefore takes chi Degree function is:

Wherein, ψ (τ) is scaling function, and τ is the time, and unit is s (second).

Then the translation function of ψ (τ) is { ψ (τ-k) }_k∈Z, then

Make u_k=(-1)^1-kv_1-k, then

(3) formula is converted to frequency-domain expression, i.e.It two is entered flexible translation can obtain:

Wherein, v_j,k(τ) being the translation space constituted based on ψ (τ), k is integer, j²=-1 is unit imaginary number, formula (4) being wavelet basis function, in formula, coefficient is well-determined.

In above-mentioned steps a, power station simulation platform is the analogue system built based on Matlab software Simulink module, its Topological structure is consistent with actual Power station structure and component parameter with model parameter.

In above-mentioned steps a, the computational methods of EFF are:

$EFF=\frac{P}{P_s}---\left(6\right)$

Wherein, P is measured power, P_sFor simulated power

Chain rate coefficient D in above-mentioned steps b_EFFRegeneration period be limited to per week, its computational methods are:

$D_{EFF}=\frac{{\mathrm{EFF}}_{max,w}-{\mathrm{EFF}}_{min,w}}{{\mathrm{EFF}}_{max,w}}---\left(7\right)$

Wherein, EFF_max,wIt is the maximum of EFF sequence, EFF in a week_min,wIt it is the minima of EFF sequence in a week.

Above-mentioned steps b circulates correction factor D_EFFRegeneration period be limited to per week, its modification method is:

${\mathrm{EFF}}_n=\frac{P}{D_{EFF}{P}_s}---\left(8\right)$

Wherein, EFF_nEFF value for next week.

In above-mentioned steps b, EFF sequence uses EMD method to process.

In above-mentioned steps c, efficiency attenuation model computational methods are:

Defined function X (t) represents the power station performance number in t.X (t) meets:

$X\left(t\right)=X\left(0\right)+a{\∫}_0^t\mathrm{\λ}(t;\mathrm{\θ})dt+\mathrm{\σ}B\left(\mathrm{\τ}\right(t;\mathrm{\γ}\left)\right)---\left(9\right)$

In formula: t is time variable, unit is the moon；γ is Brownian movement coefficient；θ is degeneration factor；X (0) represents that power station is pre- Survey the efficiency of initial time；λ(t；θ) represent the linear main attenuation of efficiency of plant；τ(t；γ) it is the most non-about time t Subtraction function, B (τ (t；γ)) it is non-linear Brownian movement, describes amount of degradation uncertainty on a timeline；A is coefficient of deviation, Characterize power station component quality diversity；σ is diffusion coefficient, represents the degree of fluctuation of uncertain decay.

Long according to efficiency of plant degradation period, in lifetime, year stable specific of attenuation, takes X (0)=0,τ(t；γ)=t^γ, then (8) formula is reduced to:

X (t)=at^θ+σB(t^γ) (10)

Setting the failure threshold in power station as D, even X (t) is more than D, then judge product failure.Efficiency die-away time accordingly It is represented by:

T=inf{t, X (t) >=D} (11)

Wherein, inf represents the infimum in power station attenuation sequence X (t), and T is the power station out-of-service time；

Then, on the premise of given a, the efficiency of plant attenuation probability density function of (9) formula is:

$f_T\left(t\right)\≅\frac{1}{A_T^{\′}}{g}_T\left(t\right)---\left(12\right)$

Wherein

$g_T\left(t\right)=\frac{\mathrm{\γ}}{\sqrt{2{\mathrm{\πt}}^2}}\sqrt{\frac{1}{{\mathrm{\σ}}^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}_0^2{t}^{\mathrm{\γ}}}}(D-\frac{{\mathrm{\γt}}^{\mathrm{\θ}}-{\mathrm{bt}}^{\mathrm{\θ}}}{\mathrm{\γ}}\frac{{\mathrm{\σ}}_0^2{\mathrm{Dt}}^{\mathrm{\θ}}+{\mathrm{\μ}}_0{\mathrm{\σ}}^2{t}^{\mathrm{\θ}}}{{\mathrm{\σ}}_0^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}^2{t}^{\mathrm{\γ}}})\×\exp (-\frac{{(D-{\mathrm{\μ}}_0{t}^{\mathrm{\θ}})}^2}{2({\mathrm{\σ}}_0^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}^2{t}^{\mathrm{\γ}})})---\left(13\right)$

$A_T^{\′}={\∫}_0^{\mathrm{\∞}}{g}_{T_0}\left(t\right)dt---\left(14\right)$

Wherein, g_TT () is T moment power station failure condition probability density, A'_TExpress for the full probability that power station residual life is T Formula.σ₀For the standard deviation of coefficient of deviation a, μ₀The average of coefficient of deviation a；

Make D_h=D-X (t_h), t_hFor a certain moment point, k is integer, D_hFor t_hMoment residual life, then photovoltaic plant time Carve t_hResidue efficiency phase (L_h) it is regarded as X (t_h) arrive D_hTime span, i.e.

$f_{L_h}\left(l\right)\≅\frac{1}{A_{L_h}^{\′}}{g}_{L_h}\left(l\right)---\left(15\right)$

Wherein

$\begin{array}{c}{h}_{L_h}\left(l\right)=\frac{\mathrm{\γ}{(t_h+l)}^{\mathrm{\γ}-1}}{\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})\sqrt{2\mathrm{\π}B}}\exp (-\frac{{(D_h-\mathrm{\Δ}\mathrm{\Λ}(t_h+l;\mathrm{\θ}\left)\right)}^2}{2B})\×\\ (D_h-C\frac{\mathrm{\Δ}\mathrm{\Λ}\left(t_h+l;\mathrm{\θ}\right)D_h+{\mathrm{\σ}}^2\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})}{B})\end{array}---\left(16\right)$

B=Δ Λ (t_h+l；θ)²+σ²Δτ(t_h+l；γ) (17)

$C=\mathrm{\Δ}\mathrm{\Λ}{(t_h+l;\mathrm{\θ})}^2-\frac{\mathrm{\θ}{(t_h+l)}^{\mathrm{\θ}-\mathrm{\γ}}\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})}{\mathrm{\γ}}---\left(18\right)$

$A_{L_h}^{\′}={\∫}_0^{\mathrm{\∞}}{g}_{L_h}\left(l\right)dl---\left(19\right)$

Fixing θ, γ, σ₀, σ asks likelihood equation by above formula and to make it be 0, solves μ₀=9.683, by its band likelihood equation B=1, γ=1, σ can be obtained further₀=2.667, σ=0.139, so far, the coefficient of equation is the most known.

The present invention compared with prior art has the advantages that:

The algorithm of the present invention is concisely reasonable, and calculating process is stably accurate, and the present invention proposes the efficiency decay in power station and declines Subtract the forecast model of phase, compensate for existing model not enough, the most more to the consideration of photovoltaic plant efficiency decay under open-air conditions Mend the vacancy in this field, be more beneficial for the marketization of photovoltaic plant.Meanwhile, also compensate for existing photovoltaic system running status to comment Estimate the problem that model accuracy has much room for improvement, be conducive to realizing further the health control technology of photovoltaic system.

Accompanying drawing explanation

The present invention is described in detail below in conjunction with the accompanying drawings with detailed description of the invention；

Fig. 1 is the Forecasting Methodology flow chart of the present invention；

Fig. 2 is actual measurement sample power signature tune line chart；

Fig. 3 is the denoising procedure chart of power curve；

Fig. 4 is measured power signal denoising result curve figure；

Fig. 5 is EFF sequence EMD exploded view；

Fig. 6 is that EFF sequence EMD extracts result figure；

Fig. 7 is the decay of photovoltaic plant efficiency by inputoutput test and model prediction Comparative result figure.

Detailed description of the invention

For the technological means making the present invention realize, creation characteristic, reach purpose and be easy to understand with effect, below in conjunction with Detailed description of the invention, is expanded on further the present invention.

As it is shown in figure 1, the natural law during i represents a week in Fig. 1, a kind of photovoltaic plant efficiency decay prediction side of the present invention Method, comprises the following steps:

A. at interval of one group of power station running environment ginseng of actual measurement in a minute in good weather day 8:00AM to 4:00PM segment Number and output are as actual measurement sample.Set up power station simulation platform, actual measurement sample environment parameter is input to power station simulation and puts down In platform, obtain simulation sample.Regard two kinds of sample datas as signal source, respectively measured power signal and simulated power signal are entered Row denoising, by the power station delivery efficiency EFF value in the moment corresponding to simulation sample power calculation of the actual measurement sample power after denoising.

B. week for unit of time EFF value sequence EMD method carried out pretreatment, and obtain it and decline percentage ratio D_EFFAs chain rate correction factor, and the EFF value using it for next week calculates.

C. gained EFF value in step b is saved as historical series.The time is reasonably selected according to actual prediction time limit length Interval, chooses EFF sample point from historical series at timed intervals, is calculated by efficiency decay forecast model, obtains final Predictive value.

Preferably, in step a, good weather pattern includes: fine day, irradiation weather clear to cloudy, low.

Further, in step a, the denoising method of measured power signal and simulated power signal uses wavelet method, its phase Related parameter is chosen as follows:

Owing to station output is relatively big by such environmental effects, actual measurement sample fluctuation per minute is relatively big, therefore takes chi Degree function is:

Wherein, ψ (τ) is scaling function, and τ is the time, and unit is s (second).

Then the translation function of ψ (τ) is { ψ (τ-k) }_k∈Z, then

Make u_k=(-1)^1-kv_1-k, then

(3) formula is converted to frequency-domain expression, i.e.It two is entered flexible translation can obtain:

Wherein, v_j,k(τ) being the translation space constituted based on ψ (τ), k is integer, j²=-1 is unit imaginary number, formula (4) being wavelet basis function, in formula, coefficient is well-determined.

Further, in step a, power station simulation platform is the emulation system built based on Matlab software Simulink module System, its topological structure is consistent with actual Power station structure and component parameter with model parameter.

Yet further, in step a, the computational methods of EFF are:

$EFF=\frac{P}{P_s}---\left(24\right)$

Wherein, P is measured power, P_sFor simulated power

Further, chain rate coefficient D in step b_EFFRegeneration period be limited to per week, its computational methods are:

$D_{EFF}=\frac{{\mathrm{EFF}}_{max,w}-{\mathrm{EFF}}_{\min ,w}}{{\mathrm{EFF}}_{max,w}}---\left(25\right)$

Wherein, EFF_max,wIt is the maximum of EFF sequence, EFF in a week_min,wIt it is the minima of EFF sequence in a week.

Further, step b circulates correction factor D_EFFRegeneration period be limited to per week, its modification method is:

${\mathrm{EFF}}_n=\frac{P}{D_{EFF}{P}_s}---\left(26\right)$

Wherein, EFF_nEFF value for next week.

Yet further, in step b, EFF sequence uses EMD method to process.

Further, in step c, efficiency attenuation model computational methods are:

Defined function X (t) represents the power station performance number in t.X (t) meets:

$X\left(t\right)=X\left(0\right)+a{\∫}_0^t\mathrm{\λ}(t;\mathrm{\θ})dt+\mathrm{\σ}B\left(\mathrm{\τ}\right(t;\mathrm{\γ}\left)\right)---\left(27\right)$

In formula: t is time variable, unit is the moon；γ is Brownian movement coefficient；θ is degeneration factor；X (0) represents that power station is pre- Survey the efficiency of initial time；λ(t；θ) represent the linear main attenuation of efficiency of plant；τ(t；γ) it is the most non-about time t Subtraction function, B (τ (t；γ)) it is non-linear Brownian movement, describes amount of degradation uncertainty on a timeline；A is coefficient of deviation, Characterize power station component quality diversity；σ is diffusion coefficient, represents the degree of fluctuation of uncertain decay.

Long according to efficiency of plant degradation period, in lifetime, year stable specific of attenuation, takes X (0)=0,τ(t；γ)=t^γ, then 8 formulas are reduced to:

X (t)=at^θ+σB(t^γ) (28)

Setting the failure threshold in power station as D, even X (t) is more than D, then judge product failure.Efficiency die-away time accordingly It is represented by:

T=inf{t, X (t) >=D} (29)

Wherein, inf represents the infimum in power station attenuation sequence X (t), and T is the power station out-of-service time.

Then, on the premise of given a, the efficiency of plant attenuation probability density function of (9) formula is:

$f_T\left(t\right)\≅\frac{1}{A_T^{\′}}{g}_T\left(t\right)---\left(30\right)$

Wherein

$g_T\left(t\right)=\frac{\mathrm{\γ}}{\sqrt{2{\mathrm{\πt}}^2}}\sqrt{\frac{1}{{\mathrm{\σ}}^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}_0^2{t}^{\mathrm{\γ}}}}(D-\frac{{\mathrm{\γt}}^{\mathrm{\θ}}-{\mathrm{bt}}^{\mathrm{\θ}}}{\mathrm{\γ}}\frac{{\mathrm{\σ}}_0^2{\mathrm{Dt}}^{\mathrm{\θ}}+{\mathrm{\μ}}_0{\mathrm{\σ}}^2{t}^{\mathrm{\θ}}}{{\mathrm{\σ}}_0^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}^2{t}^{\mathrm{\γ}}})\×\exp (-\frac{{(D-{\mathrm{\μ}}_0{t}^{\mathrm{\θ}})}^2}{2({\mathrm{\σ}}_0^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}^2{t}^{\mathrm{\γ}})})---\left(31\right)$

$A_T^{\′}={\∫}_0^{\mathrm{\∞}}{g}_T\left(t\right)dt---\left(32\right)$

Wherein, g_TT () is T moment power station failure condition probability density, A'_TExpress for the full probability that power station residual life is T Formula.σ₀For the standard deviation of coefficient of deviation a, μ₀The average of coefficient of deviation a；

Make D_h=D-X (t_h), t_hFor a certain moment point, k is integer, D_hFor t_hMoment residual life, then photovoltaic plant time Carve t_hResidue efficiency phase (L_h) it is regarded as X (t_h) arrive D_hTime span, i.e.

$f_{L_h}\left(l\right)\≅\frac{1}{A_{L_h}^{\′}}{g}_{L_h}\left(l\right)---\left(33\right)$

Wherein

$\begin{array}{c}{g}_{L_h}\left(l\right)=\frac{\mathrm{\γ}{(t_h+l)}^{\mathrm{\γ}-1}}{\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})\sqrt{2\mathrm{\π}B}}\exp (-\frac{{(D_h-\mathrm{\Δ}\mathrm{\Λ}(t_h+l;\mathrm{\θ}\left)\right)}^2}{2B})\×\\ (D_h-C\frac{\mathrm{\Δ}\mathrm{\Λ}(t_h+l;\mathrm{\θ})D_h+{\mathrm{\σ}}^2\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})}{B})\end{array}---\left(34\right)$

B=Δ Λ (t_h+l；θ)²+σ²Δτ(t_h+l；γ) (35)

$C=\mathrm{\Δ}\mathrm{\Λ}{(t_h+l;\mathrm{\θ})}^2-\frac{\mathrm{\θ}{(t_h+l)}^{\mathrm{\θ}-\mathrm{\γ}}\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})}{\mathrm{\γ}}---\left(36\right)$

$A_{L_h}^{\′}={\∫}_0^{\mathrm{\∞}}{g}_{L_h}\left(l\right)dl---\left(37\right)$

Fixing θ, γ, σ₀, σ asks likelihood equation by above formula and to make it be 0, solves μ₀=9.683, by its band likelihood equation B=1, γ=1, σ can be obtained further₀=2.667, σ=0.139, so far, the coefficient of equation is the most known.

Embodiment 1:

1. the data acquisition used by the present embodiment surveys platform from Hohai University open air 10kWp.First have chosen 2014 November 3, raw sample data, extracted 1 group of data, totally 480 groups of works at interval of 1 minute within 8:00AM to the 4:00PM time period For actual measurement sample, its power signal is carried out denoising.Original power signal curve as in figure 2 it is shown, processing procedure as it is shown on figure 3, In Fig. 3, s is original power signal, and a4 is main signal, and d1, d2, d3, d4 are 4 kinds and are entrained in the Gauss of different frequency in main signal White noise, 5 signal overlaps constitute original power signal, s=a together₄+d₁+d₂+d₃+d₄, power signal curve after denoising As shown in Figure 4.It can be seen that the sample curve after denoising eliminates the interference of outside noise signal, overall smoothness increases, Burr disappears.Sample environment parameter is input in power station simulation platform, obtains 480 groups of simulation sample under same ambient parameter, adopt By same method, its power signal is carried out denoising.When calculating corresponding with the measured power signal after denoising to simulated power signal The power station delivery efficiency EFF value carved.

2. choose data in 8:00AM to the 4:00PM time period on November 9,3 days to 2014 November in 2014, by step a In method data are carried out pretreatment, then carry out decomposing and smoothing processing by EFF value sequence EMD method.EMD processing procedure As it is shown in figure 5, signal is original EFF sequence in Fig. 5, i.e. without the EFF sequence of EMD algorithm process, imf1 to imf8 is The component that EMD algorithm decomposites, after gradually rejecting these components, remaining res is a monotonic function, is at EMD algorithm The new EFF sequence obtained after reason.EMD extracts result as shown in Figure 6.It can be seen that EMD carries out spy to the characteristic parameter of EFF Levy extraction, by the EFF sequence stationaryization of non-stationary and dullnessization, more accurately grasp the characteristic signal of primary signal.According to EMD extracts result, calculates D_EFF, D can be obtained_EFF=(0.99945-0.99932) ÷ 0.99945 × 100%=0.013%, i.e. originally Week, efficiency of plant decayed 0.013%, as chain rate correction factor, and used it for the EFF value calculating in next week.

3. choose and be total to bimestrial actual measurement sample in December, 2014 in November, 2014, enter by the method for step a and step b Row processes, and obtains final EFF sequence.With sky for interval, choose one group of data in the same time in every day, as forecast sample, defeated Enter in efficiency decay forecast model.Obtain the actual measurement platform delivery efficiency decay predictive value in January, 2015 to November, and by it Contrasting with measured value, result is as shown in Figure 7.It can be seen that the Forecasting Methodology described in this patent is effective.

The ultimate principle of the present invention and principal character and advantages of the present invention have more than been shown and described.The industry describes The principle of the present invention is simply described, without departing from the spirit and scope of the present invention, the present invention also have various change and Improving, these changes and improvements both fall within scope of the claimed invention.Claimed scope is by appended power Profit claim and equivalent thereof define.

Claims (9)

1. a photovoltaic plant efficiency decay Forecasting Methodology, it is characterised in that comprise the following steps:

A. in good weather day 8:00AM to 4:00PM segment, at interval of one group of power station running environment parameter of actual measurement in 1 minute With output as actual measurement sample；Set up power station simulation platform, actual measurement sample environment parameter is input to power station simulation platform In, obtain simulation sample；Regard actual measurement sample and simulation sample data as signal source, respectively to measured power signal and emulation merit Rate signal carries out denoising, by the power station delivery efficiency in the moment corresponding to simulation sample power calculation of the actual measurement sample power after denoising EFF value；

B. EFF value sequence EMD method to be processed for unit of time in week, and EFF value decline percentage ratio D is obtained_EFFMake For chain rate correction factor, and by D_EFFEFF value for next week calculates；

C. gained EFF value in step b is saved as historical series, according to actual prediction time limit length, between the rational selection time Every, from historical series, choose EFF sample point at timed intervals, calculated by efficiency decay forecast model, obtain final pre- Measured value.

A kind of photovoltaic plant efficiency decay Forecasting Methodology the most according to claim 1, it is characterised in that: in described step a Good weather pattern includes: fine day, irradiation weather clear to cloudy, low.

A kind of photovoltaic plant efficiency decay Forecasting Methodology the most according to claim 1, it is characterised in that: in described step a Measured power signal uses wavelet method with the denoising method of simulated power signal, and its scaling function is chosen as follows:

Wherein, ψ (τ) is scaling function, and τ is the time, and unit is s (second).

A kind of photovoltaic plant efficiency decay Forecasting Methodology the most according to claim 3, it is characterised in that: described wavelet method Wavelet basis function result of calculation as follows:

Wherein, v_j,k(τ) being the translation space constituted based on ψ (τ), k is integer, j²=-1 is unit imaginary number, and formula (4) is little Ripple basic function, in formula, coefficient is well-determined.

A kind of photovoltaic plant efficiency decay Forecasting Methodology the most according to claim 1, it is characterised in that: in described step a Power station simulation platform is the analogue system built based on Matlab software Simulink module, the topology of described power station simulation platform Structure is consistent with actual Power station structure and component parameter with model parameter.

A kind of photovoltaic plant efficiency decay Forecasting Methodology the most according to claim 1, it is characterised in that: in described step a The computational methods of EFF value are:

$EFF=\frac{P}{P_s}---\left(5\right)$

Wherein, P is measured power, P_sFor simulated power.

A kind of photovoltaic plant efficiency decay Forecasting Methodology the most according to claim 6, it is characterised in that: in described step b Chain rate correction factor D_EFFRegeneration period be limited to per week, D_EFFComputational methods be:

$D_{EFF}=\frac{{\mathrm{EFF}}_{max,w}-{\mathrm{EFF}}_{min,w}}{{\mathrm{EFF}}_{\max ,w}}---\left(6\right)$

Wherein, EFF_max,wIt is the maximum of EFF sequence, EFF in a week_min,wIt it is the minima of EFF sequence in a week.

A kind of photovoltaic plant efficiency decay Forecasting Methodology the most according to claim 7, it is characterised in that: next week described The modification method of EFF value be:

${\mathrm{EFF}}_n=\frac{P}{D_{EFF}{P}_s}---\left(7\right)$

Wherein, EFF_nEFF value for next week.

A kind of photovoltaic plant efficiency decay Forecasting Methodology the most according to claim 1, it is characterised in that: in described step c Efficiency attenuation model computational methods are:

Defined function X (t) represents the power station performance number in t, and X (t) meets:

$X\left(t\right)=X\left(0\right)+a{\∫}_0^t\mathrm{\λ}(t;\mathrm{\θ})dt+\mathrm{\σ}B\left(\mathrm{\τ}\right(t;\mathrm{\γ}\left)\right)---\left(8\right)$

In formula: t is time variable, unit is the moon；γ is Brownian movement coefficient；θ is degeneration factor；X (0) represents at the beginning of the prediction of power station The efficiency in moment beginning；λ(t；θ) represent the linear main attenuation of efficiency of plant；τ(t；γ) for subtract letter about the most non-of time t Number, B (τ (t；γ)) it is non-linear Brownian movement, describes amount of degradation uncertainty on a timeline；A is coefficient of deviation, characterizes Power station component quality diversity；σ is diffusion coefficient, represents the degree of fluctuation of uncertain decay；

Long according to efficiency of plant degradation period, in lifetime, year stable specific of attenuation, takes τ (t；γ)=t^γ, then formula (8) is reduced to:

X (t)=at^θ+σB(t^γ) (9)

Setting the failure threshold in power station as D, even X (t) is more than D, then judge product failure, and corresponding efficiency die-away time can table It is shown as:

T=inf{t, X (t) >=D} (10)

Wherein, inf represents the infimum in power station attenuation sequence X (t), and T is the power station out-of-service time；

Then, on the premise of given a, the efficiency of plant attenuation probability density function of formula (9) is:

$f_T\left(t\right)\≅\frac{1}{A_T^{\′}}{g}_T\left(t\right)---\left(11\right)$

Wherein

$g_T\left(t\right)=\frac{\mathrm{\γ}}{\sqrt{2{\mathrm{\πt}}^2}}\sqrt{\frac{1}{{\mathrm{\σ}}^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}_0^2{t}^{\mathrm{\γ}}}}(D-\frac{{\mathrm{\γt}}^{\mathrm{\θ}}-{\mathrm{bt}}^{\mathrm{\θ}}}{\mathrm{\γ}}\frac{{\mathrm{\σ}}_0^2{\mathrm{Dt}}^{\mathrm{\θ}}+{\mathrm{\μ}}_0{\mathrm{\σ}}^2{t}^{\mathrm{\θ}}}{{\mathrm{\σ}}_0^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}^2{t}^{\mathrm{\γ}}})\×\exp (-\frac{{(D-{\mathrm{\μ}}_0{t}^{\mathrm{\θ}})}^2}{2({\mathrm{\σ}}_0^2{t}^{2\mathrm{\θ}}+{\mathrm{\σ}}^2{t}^{\mathrm{\γ}})})---\left(12\right)$

$A_T^{\′}={\∫}_0^{\mathrm{\∞}}{g}_T\left(t\right)dt---\left(13\right)$

Wherein, g_TT () is T moment power station failure condition probability density, A'_TFor the full probability expression formula that power station residual life is T, σ₀For the standard deviation of coefficient of deviation a, μ₀The average of coefficient of deviation a；

Make D_h=D-X (t_h), t_hFor a certain moment point, k is integer, D_hFor t_hMoment residual life, then photovoltaic plant is at moment t_h Residue efficiency phase (L_h) it is considered X (t_h) arrive D_hTime span, i.e.

$f_{L_h}\left(l\right)\≅\frac{1}{A_{L_h}^{\′}}{g}_{L_h}\left(l\right)---\left(14\right)$

Wherein

$\begin{array}{c}{g}_{L_h}=\frac{\mathrm{\γ}{(t_h+l)}^{\mathrm{\γ}-1}}{\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})\sqrt{2\mathrm{\π}B}}\exp (-\frac{{(D_h-\mathrm{\Δ}\mathrm{\Λ}(t_h+l;\mathrm{\θ}\left)\right)}^2}{2B})\×\\ (D_h-C\frac{\mathrm{\Δ}\mathrm{\Λ}\left(t_h+l;\mathrm{\θ}\right)D_h+{\mathrm{\σ}}^2\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})}{B})\end{array}---\left(15\right)$

B=Δ Λ (t_h+l；θ)²+σ²Δτ(t_h+l；γ) (16)

$C=\mathrm{\Δ}\mathrm{\Λ}{(t_h+l;\mathrm{\θ})}^2-\frac{\mathrm{\θ}{(t_h+l)}^{\mathrm{\θ}-\mathrm{\γ}}\mathrm{\Δ}\mathrm{\τ}(t_h+l;\mathrm{\γ})}{\mathrm{\γ}}---\left(17\right)$

$A_{L_h}^{\′}={\∫}_0^{\mathrm{\∞}}{g}_{L_h}\left(l\right)dl---\left(19\right)$

Fixing θ, γ, σ₀, σ asks likelihood equation by above formula and to make it be 0, solves μ₀=9.683, its band likelihood equation is entered one Step can obtain b=1, γ=1, σ₀=2.667, σ=0.139, so far, the coefficient of equation is the most known.