JP4688097B2 - Wind generator operating state determination method - Google Patents

Description

  The present invention relates to a method for determining an operating state of a wind power generator in a grid interconnection system in which wind power generators are grid-connected.

  Alternating power generators widely used as wind power generators are roughly classified into two types: pole-switching induction generators and variable speed generators. A pole-switching induction generator switches the number of revolutions by switching the number of poles, and usually has two kinds of revolutions. The pole switching type induction generator has power quality problems such as a sudden voltage drop due to inrush current when switching the number of poles, and the voltage drop varies depending on the number of poles after switching. It is necessary to grasp this.

  On the other hand, the variable speed generator is a super-synchronous Serbius induction generator that changes the rotation speed according to the wind speed, or is connected to the system through a frequency converter and AC excitation is applied to operate the synchronous generator at a variable speed. is there. In variable speed generators, it is necessary to perform simulations in advance and examine whether there is a problem in power quality or the like in order to perform complex control that changes the rotational speed in accordance with the wind speed.

  In other words, regardless of the type of pole-switching induction generator or variable-speed generator, modeling data that is the basis for that type requires generator measurement data, and one of the basic data is the number of revolutions. There is. The number of revolutions cannot be measured directly by the power company with which the wind power generator is linked, and the data obtained is only the voltage and current in the two-phase wattmeter method with the watt hour meter at the demarcation point with the grid interconnection system. Absent.

  As means for obtaining the rotational speed of the wind power generator, the pulsation of the output power P (t) of the wind power generator due to the rotation of the blade of the wind turbine can be used. That is, in the vicinity of the tower of the windmill, the wind speed tends to be weakened due to the effect of the tower body, and as a result, the output of the wind power generator is reduced, so that the output power P (t) of the wind power generator pulsates. A tower shadow effect occurs.

For example, two pulsation frequency (theoretical vibration frequency) f _t1, f _t2 of wind power generator of the number of poles switching type is a blade rotation speed n _i (rpm), when the number of blades N _B, given by: It is done.

[Equation 1]
f _ti = n _i N _B / 60 (1)
(I = 1, 2)
Thus, if the pulsation frequency f _ti of the output power P (t) of the wind power generator is known, the rotation speed of the wind power generator can be known. In order to simplify the following description, simply referred to as f _t the theoretical vibration frequencies.

Next, the estimation method of the theoretical vibration frequency f _t. Now, it is assumed that the output power P (t) of the wind power generator is a sine wave as shown in equation (2). P _m is the peak value of the output power P (t), and θ ₀ is the phase.

[Equation 2]
_{P (t) = P m cos} (2πf t t + θ 0) ... (2)
A square wave window is applied to both ends of the time domain t + T / 2 to t−T / 2 with respect to the wind power generator output P (t) of the above expression around the time t to perform Fourier transform. In this case, it is assumed that the integration time T has a relationship shown in the equation (3).

[Equation 3]
2πft _t = 2πN (3)
(N is the number of vibration cycles)
Then, the absolute value | χ (t, ω) | of the Fourier transform value is expressed by equation (4).

Here, ω = ωt + δω, ωt = 2πft, θ (t) = ωt + θ ₀ , δω: angular frequency error.

Next, the cos (δωT) term of the equation (4) is macrourin expanded and approximated to the 4th order term, and δω is sufficiently small, the {δω / (2ω _t + δω)} ² term is ignored and 2ω _t + δω≈2ω. If approximated with _t , the absolute value | χ (t, ω) | of the Fourier transform value is approximated as the following equation (5).

Furthermore, if the fourth term on the right side of equation (5) is sufficiently small and ignored, the approximate equation shown in equation (6) below is obtained.

Considering up to the second term in the root sign from this equation (6), when δω = 0, the absolute value of the Fourier transform value | χ (t, ω) | To do. When considering up to the third term, the angular frequency error δω that maximizes the absolute value | χ (t, ω) | of the Fourier transform value is given by the following (7) from the extreme value condition regarding the angular frequency error δω. It is shown in relation to the formula.

  As can be seen from the equation (7), the angular frequency error δω is a time function and has the following characteristics.

(A) Since the angular frequency error δω is inversely proportional to the square N ² of the vibration period number N, it decreases rapidly if the integration time T, which is the window length, is increased during Fourier transformation.

(B) it is proportional to _{cos (2ω t t + 2θ 0} ), when Fourier transformation, the trend of the angular frequency error δω vibrates at twice harmonics of f _t.

From Equation (7), the maximum value of the vibration error ε (ε = δω / ω _t ) due to the number of vibration periods N is when cos (2ω _t t + 2θ ₀ ) = 1, and is given as shown in Table 1 below.

  As shown in Table 1, when the vibration period number N is N = 1, the vibration error ε is as large as 15.2%, indicating that the practicality is poor. For this reason, a vibration period number N = 3 in consideration of suppression of vibration error ε has been adopted from a practical aspect.

  However, the vibration of the output power P (t) of the wind power generator due to the tower shadow effect of the windmill does not continue for three periods, and the vibration frequency may not be estimated. For this reason, a method that can be estimated even when the vibration period number N is 2 or less is desired. When the number of vibration cycles N is N = 3, the vibration frequency is averaged over three cycles.

  For this reason, even in a section where the vibration does not persist in the data (hereinafter, referred to as a non-good quality section), there may be a result that the vibration appears at first glance. Therefore, an estimation method that can automatically determine a non-quality section of data is also desired.

  Accordingly, the inventors of the present applicant have invented a method for determining the operating state of a wind power generator that can accurately estimate the vibration frequency of the output power of the wind power generator that is grid-connected with a short number of vibration cycles. A patent application was filed under No.-168021.

That is, through a band filter using the moving average method around the theoretical vibration frequency f _t of the output power signal of the wind power generator, the absolute of the Fourier transform value by performing integration time domain (1 / f _t) the Fourier transform of s The frequency having the maximum value is obtained as the estimated vibration frequency signal f ₁ , and the estimated vibration frequency f _k (k = 1/2 ⁱ ) including the error vibration of 2 ^j harmonics is sequentially passed through the high-pass filter by the moving average method. The estimated vibration frequency signal f _{k + 1} (k + 1 = 1/2 ^{i + 1} ) is obtained by performing a Fourier transform in the integration time region (1/2 ⁱ f _t ) seconds, and further, the estimated vibration frequency signal f _{k + 1} seeking integration time domain (2/2 i + ¹ f _t) estimated vibration subjected to Fourier transform of the second frequency signal _{f m (m = 2/2} i + 1) for the final frequency through numerical digital low-pass filter to determine the estimated value f _T.

  As a result, the vibration of the output power of the wind turbine generator is more accurately extracted, the section where the vibration of the output power of the wind power generator is sustained is more accurately extracted, and the frequency estimate of the vibration of the output power of the wind turbine generator is extracted. Can be estimated more accurately.

  In this Japanese Patent Application No. 2004-168021, when the output power vibration of the wind power generator continues to some extent, the frequency estimation of the vibration can be accurately estimated, but when the duration of the output power vibration is short, It was determined that the non-good quality section in which the vibration did not persist was determined, and that there were few good quality sections.

  The object of the present invention is to accurately estimate the vibration frequency of the wind power generator even when the duration of the output power vibration of the wind power generator connected to the grid is short. It is an object to provide a method for determining an operating state of a wind power generator capable of determining an operating state such as the relationship of

  The wind generator operating state determination method according to the invention of claim 1 is a time series of output power signals obtained by sampling an output power signal of a pole-switching wind generator including a measured vibration component at a predetermined period. Data is input, and the time series data is passed through a band-pass filter based on a moving average method centered on a theoretical vibration frequency that is theoretically estimated as vibration of the output power signal. The time series data of the removed output power signal is accumulated, the time series data of the accumulated output power signal is applied to the autoregressive model of the damped oscillation prepared in advance, and output from the coefficient of the autoregressive model to which the time series data is applied The vibration frequency of the power signal is calculated, the validity of the vibration frequency obtained based on the predetermined condition is determined, and the vibration frequency satisfying the predetermined condition is determined as the vibration frequency estimated value. Extracted Te, and identifies the rotational speed of the wind turbine based on the vibration frequency estimate.

  According to a second aspect of the present invention, there is provided a method for discriminating an operating state of a wind power generator by sampling an output power signal of a variable speed type wind power generator including a measured vibration component at a predetermined period and time series data of the output power signal. And the center frequency of each region obtained by equally dividing the existence region in which the vibration of the output power signal is present into n equal parts is defined as the theoretical frequency that is theoretically estimated as the vibration of the output power signal. The time-series data is passed through a band-pass filter based on the moving average method centering on the frequency, and the time-series data of the output power signal from which the low-frequency and high-frequency components are removed by the band-pass filter is accumulated for each theoretical frequency. Apply the time series data of the accumulated output power signal for each theoretical frequency to the auto-regression model of the damped vibration prepared in advance, and the coefficient of the auto-regression model applying the time series data for each theoretical frequency. Calculate the vibration frequency of the output power signal, determine the appropriateness of each vibration frequency obtained based on a predetermined condition, extract each vibration frequency satisfying a predetermined condition as each vibration frequency estimate, The rotational speed of the wind power generator is specified based on the estimated vibration frequency.

  According to a third aspect of the present invention, there is provided a method for determining an operating state of a wind power generator, wherein the output power signal of the wind power generator including the measured vibration component is passed through the band filter twice. It is characterized by that.

  A wind power generator operating state determination method according to the invention of claim 4 is the invention according to any one of claims 1 to 3, wherein the output power signal of the wind power generator that has passed through the bandpass filter is equal to or less than a predetermined threshold value. The output power signal in the section is excluded.

  According to a fifth aspect of the present invention, there is provided a method for determining an operating state of a wind power generator according to any one of the first to fourth aspects, wherein the sine of a coefficient including a sine term indicating vibration among the coefficients of the autoregressive model. The validity of the vibration frequency is determined under the condition that the absolute value of the term is 1 or less.

  According to a sixth aspect of the present invention, there is provided a method for determining an operating state of a wind power generator according to any one of the first to fifth aspects, wherein the coefficient includes a time constant indicating a damped oscillation among the coefficients of the autoregressive model. The validity of the vibration frequency is determined under the condition that the value of the time constant is positive.

  According to a seventh aspect of the invention, there is provided the wind power generator operating state determination method according to the sixth aspect of the invention, wherein the time constant range of the coefficient including the time constant indicating the damped oscillation is the coefficient of the autoregressive model. The validity of the vibration frequency is determined under a condition larger than the value estimated based on the time series data.

  The wind power generator operating state determination method according to the invention of claim 8 is the invention according to claim 6, wherein the coefficient range including the time constant indicating the damped oscillation among the coefficients of the autoregressive model is the time series data. The validity of the vibration frequency is determined under a condition that is equal to or greater than the value estimated based on the above and less than a value obtained by adding a margin value to the value when the time constant is infinite.

  According to a ninth aspect of the present invention, there is provided the method for determining an operating state of a wind power generator according to any one of the first to eighth aspects, wherein among the coefficients of the autoregressive model, variation of a coefficient including a time constant indicating a damped oscillation. The validity of the vibration frequency is determined under the condition that the minute is smaller than a predetermined threshold value.

  According to a tenth aspect of the present invention, there is provided a method for determining an operating state of a wind power generator according to any one of the first to ninth aspects, wherein the mean value of the vibration frequency in the estimation region is ± It is characterized by targeting vibration frequencies within 15%.

  The wind turbine generator operating state determination method according to the invention of claim 11 is the invention according to any one of claims 1 to 10, wherein the vibration frequency is kept constant in the estimation region as the determination of the validity of the vibration frequency. The vibration frequency whose time length is below a predetermined time length is excluded.

  According to the present invention, the autoregressive model of the damped oscillation of the output power signal is created based on the time series data of the output power signal, and the oscillation frequency of the output power signal is calculated from the coefficient of the autoregressive model. The vibration frequency of the output power signal can be calculated even with the time series data. Further, since the damped vibration waveform model is adopted a priori as the autoregressive model, the resolution is not defined by the sampling interval for obtaining time series data. In addition, the calculation amount for calculating the vibration frequency is small. Furthermore, since a damped vibration waveform model is assumed, the validity of the vibration frequency as an estimation result can be easily determined from the correspondence with the time constant of the wind power generator obtained from actual data.

  Embodiments of the present invention will be described below. FIG. 1 is a flowchart of a wind generator operating state determination method according to an embodiment of the present invention. This embodiment is intended for a pole-switching wind power generator. The output power of the pole number switching type wind power generator is measured by a watt hour meter of a grid interconnection system in which the pole number switching type wind power generator is linked. The measured output power signal of the pole-switching wind generator includes a pulsation due to the tower shadow effect.

First, the output power signal P (t) of the pole-switching wind power generator including the measured vibration is read at a predetermined sampling period (S1), and the theoretical vibration theoretically estimated as the vibration of the output power signal. An output power signal P _B (t) from which a low-frequency component and a high-frequency component are removed is obtained through a band-pass filter based on the moving average method centering on the frequency f _t (S2). Then, time-series data P _{B, n} (t) (n = 1, 2,... N) with a predetermined sampling number is accumulated (S3).

Next, an autoregressive model of the damped oscillation of the output power signal is created based on the time series data P _{B, n} (t) of the output power signal (S4). Since the power P _B (t) after application of the bandpass filter is generally damped vibration in the vibration region, a quadratic model is adopted as an autoregressive model ARM (Auto Regressive Model). That is, as will be described later, the relationship between a plurality of time-series data P _{B, n} (t) using two coefficients, a coefficient including a sine term indicating vibration and a coefficient including a time constant indicating damped vibration. Create an autoregressive model by deriving the formula.

Then, the vibration frequency f of the output power signal P _B (t) is calculated from the two coefficients of the relational expression of the autoregressive model (S5), and the validity of the calculated vibration frequency f is based on a predetermined condition. Determine (S6). A method for determining the validity of the vibration frequency f will be described later. This determination to extract satisfies oscillation frequency predetermined as a vibration frequency estimate f _T (S7), to identify the rotational speed of the wind power generator on the basis of the oscillation frequency estimate f _T (S8). By substituting the estimated vibration frequency f _T into f _ti in the equation (1), the rotational speed n _i of the wind power generator is specified.

  Hereinafter, the processing contents of steps S1 to S7 will be described in detail. First, the band filter will be described. When estimating the rotational speed of the wind power generator, the vibration due to the tower shadow effect included in the output signal P (t) of the wind power generator is extracted, but the low frequency vibration is superimposed on the vibration due to the tower shadow effect. In some cases, when low-frequency vibration is superimposed, a fundamental wave vibration error occurs. Therefore, a bandpass filter is applied to suppress this and remove high frequency vibration. In other words, since there are only two types of theoretical vibration frequencies given by the equation (1) in the pole number switching type wind power generator, a band-pass filter using these as the frequencies having the maximum pass rate is applied.

As the band filter in this case, a combination of a moving average method low-pass filter and a high-pass filter whose phase does not have frequency characteristics is used. Band filter frequency of the filter center and the theoretical vibration frequency f _t, and versus frequency (f) characteristic of the passing ratio K _B is adopted as given by the following equation.

  In addition, the bandpass filter is continuously applied twice to the output power signal P (t) of the wind power generator as necessary. Thereby, the low frequency vibration and high frequency vibration included in the output power signal P (t) of the wind power generator are removed.

Figure 2 is a vs. frequency characteristic diagram of the passing rate K _B of the bandpass filter used in the embodiment of the present invention. As shown in FIG. 2, bandpass filter frequency of the center and the theoretical vibration frequency f _t, a filter passing ratio K _B is 0.4.

FIG. 3 is a frequency characteristic diagram of the output power signal P _B (t) of the wind power generator when the bandpass filter is applied to the measured output power signal P (t) of the wind power generator twice in succession. is there. FIG. 3 shows the measured output power signal P (t) of the wind power generator and the output power signal P _B (t) of the wind power generator after applying the bandpass filter. Although P (t) and P _B (t) having only about two cycles of vibration are shown in a comparative diagram, it can be seen that they are sufficiently smoothed by the bandpass filter.

Next, of the output power signal P _B of the wind turbine that has passed through the band-pass filter (t), a predetermined threshold value P _B, the output power signal P _B of _C following section (t) is optionally excluded . This is because the reliability of the estimation result is low when the vibration amplitude due to the tower shadow effect is small, and it is desirable to exclude the estimation result in a region where the pulsation of the output power signal P _B (t) is small. . For this reason, the threshold value P _{B, C} for the amplitude value of the output power signal P _B (t) as an exclusion index is set as follows.

First, assuming that the generator is constant at the rated voltage V _N and the power factor is 1, the power pulsation is determined by the current pulsation. Thus, jump width of the effective power corresponding to 1-bit A / D converter Δp has an input range V _in the M-bit A / D converter, the rated current I _N to the voltage converted input signal level V _When using _I , the following equation (9) is given.

Here, β = 2 ^M−1 (V _I / V _in ).

Furthermore, given that passing ratio _{K B} of the bandpass filter through the band pass filter 2 times of 0.4, jumping width Delta] p _B of the output power signal P _B (t) becomes 0.4Derutapi. Accordingly, as shown in FIG. 4, the threshold value P _{B, C} of the output power signal P _B (t) has a critical value at the minimum resolution (Δp _B / 2) of the A / D converter. Since the actual trend is not a sine wave, the threshold value P _{B, C} is set as shown in Equation (10) using the margin α (α ₀ ≧ α ≧ 1).

Next, since the time series data P _{B, n} (t) of the output power signal P _B (t) is obtained for each sampling period Δt, the time series data P _{B, n} (t) for each sampling period Δt is sequentially provided. (N = 1, 2,... N) are accumulated.

Since the vibration region of the power P _B (t) after application of the bandpass filter is damped vibration, the autoregressive model adopts a quadratic model as shown in the equation (11).

Here, P is the peak value of the output power signal P _{B (t),} omega is the output power signal P angular frequency of _B (t), theta is the output power signal P _B of the (t) phase, T is the time of the attenuation constant It is. Then, rewriting equation (11) yields equation (12).

Substituting the time nΔt (n = 1, 2,... N) at each sampling time into t in the equation (12), each time series data P _{B, n} (t) (n = 1, 2,... N) ) Now, when time (n−2) Δt is substituted for t in equation (12), time-series data P _{B, n−2} is represented by equation (13).

Then, when an operator Z (Z = ^ελΔt ) for advancing time is introduced, the relationship between the time series data P _{B, n-1} and the time series data P _{B, n-2} is expressed by equation (14).

By ^calculating P _h ελ ^{(n−2) Δt} and (P _h ελ ^{(n−2) Δt} ) ^* from the equations (13) and (14), the equation (15) is obtained.

On the other hand, the relationship between the time-series data P _{B, n} and the time-series data P _{B, n-2} is expressed by equation (16) when expressed using an operator Z (Z = ^ελΔt ) that advances time.

Substituting P _h ε ^{λ (n−2) Δt} and (P _h ε ^{λ (n−2) Δt} ) ^* obtained in equation (15) into equation (16) yields equation (17).

The sum of the advance time operator Z and its complex conjugate Z ^* (Z + Z ^*) to a _1, the product (ZZ ^*) with operators Z advancing the time and its complex conjugate Z ^* putting the a _2, ( 18) Equation is obtained.

As can be seen from the equation (18), the time series data P _{B, n} is obtained by multiplying the time series data P _{B, n-1} before one sampling by the coefficient a ₁ and the time series data P _B before two samplings. _{, n−2} multiplied by the coefficient a ₂ , and the coefficient a ₁ and the coefficient a ₂ are expressed by equation (19).

As can be seen from the equation (19), the coefficient a ₁ is a coefficient including a sine term indicating vibration, and the coefficient a ₂ is a coefficient including a time constant T indicating damped vibration. The coefficients a ₁ and a ₂ can be obtained from four pieces of time series data. That is, the equation (18) consisting of three time series data P _{B, n} , P _{B, n-1} and P _{B, n-2} and the three time series data P _{B, n-1} before one sampling. , P _{B, n-2} , P _{B, n-3} , the coefficients a ₁ and a ₂ can be obtained by simultaneous equations with the equation (18). In other words, a simultaneous equation is established from the four time series data P _{B, n} , P _{B, n−1} , P _{B, n−2} , P _{B, n−3} using the equation (18), and the coefficients a ₁ , a ₂ can be obtained.

In the embodiment of the present invention, the coefficients a ₁ and a ₂ are obtained by the least square method using N continuous data strings in consideration of the measurement accuracy. Then, from the obtained coefficients a _1, a _2, obtaining the oscillation frequency using the equation (19). First, the decay time constant T is obtained by the equation (20) from the second equation of the equation (19).

Also, the vibration frequency f is obtained by substituting the second equation into the first equation (19) and ω = 2πf, so that the vibration frequency f is obtained by equation (21).

FIG. 5 is a trend diagram of the coefficient a ₂ including the time constant T indicating the output power signal P _B , the vibration frequency f, and the damped vibration of the wind power generator. As shown in FIG. 5, the output power signal P _B of the wind power generator pulsates between 146 s and 149 s. On the other hand, the calculated vibration frequency f rises in the vicinity of 145.85 s, and substantially constant vibration continues in the range of 0.9 Hz to 1.0 Hz between 145.9 s and 147.85 s. And it falls in the vicinity of 147.85 s, and a substantially constant vibration continues in the range of 0.8 Hz to 0.9 Hz again between 147.9 s and 147.4 s. Furthermore, it falls near 147.4 s, and substantially constant vibration continues in the range of 0.65 Hz to 0.75 Hz between 147.6 s and 148.15 s. After 148.3 s, the vibration frequency f varies greatly, and the period during which substantially constant vibration continues is short.

Judging from the characteristics of the vibration frequency f, it can be estimated that the vibration frequency f between 145.9 s and 148.15 s is the vibration frequency of pulsation due to the tower shadow effect of the output power signal P _B of the wind power generator.

  In the embodiment of the present invention, in order to obtain the vibration frequency f with high accuracy, the appropriate vibration frequency f is determined by providing a criterion for determining the validity of the vibration frequency f that is the measurement result. In other words, since the vibration caused by the tower shadow effect is handled as a damped vibration by the autoregressive model, it is easy to correspond to the actual waveform, and the validity criteria of the obtained vibration frequency f can be determined clearly. Can be expected.

First, as a first criterion, since the absolute value of the cos term in the equation (20) is 1 or less, those not satisfying this are excluded. That is, the vibration frequency f that does not satisfy the condition of the expression (22) is excluded.

  Next, the vibration of the tower shadow effect will be damped because the control system of the wind power generator works and brakes. Therefore, a criterion that the time constant T is positive is introduced from the condition of the damped vibration.

Furthermore, as a second reference using the estimated time constant from the actual data, the estimated attenuation time constant T of the actual data is normally set in the range of T ≧ 1 seconds because the blade rotation is mechanical. The sampling period Δt is set to about 0.02 s. As a result, the value of the coefficient a ₂ when the time constant is infinite is 1 from the second expression of the expression (19).

However, since there is a beat-like waveform, the coefficient a ₂ may be much larger than 1. In addition, there is a case where T <1 because the control system responds immediately when the output changes suddenly. Accordingly, the reference based on the range of the coefficient a ₂ is set as follows.

[Equation 23]
a _{2, max} ≧ a ₂ ≧ a _{2, min} (23)
For example, a2 _{, max} is a value of 1 when the time constant T is infinite, for example, a value of about 1.02 to 1.01 including 0.02 to 0.01, and a2 _{, min} Is about 0.96 to 0.99. Thereby, for example, the vibration frequency f between 145 s to 146 s and 149 s to 150 s can be eliminated.

Next, when the vibration attenuation is smooth, the coefficient a ₂ is substantially constant. Therefore, as a third reference, the vibration frequency f in a region where the absolute value of the change in the coefficient a ₂ is smaller than a predetermined threshold value. Is adopted. In this case, a coefficient [a ₂ ] smoothed by the moving average method is obtained as preprocessing, and the smoothed coefficient [a ₂ ] is used to calculate the amount of change in the smoothed coefficient [a ₂ ]. The absolute value | Δ [a _{2, k} ] | is determined by being smaller than a predetermined threshold value [a _{2, c} ]. Expression (24) is the determination expression.

[Equation 24]
Δ [a _{2, k} ] = | [a _{2, k} ] − [a _{2, k−1} ] | <[a _{2, c} ] (24)
Here, k is a time step, and [a _{2, c} ] is a threshold value of [a _{2, k} ].

FIG. 6 is a trend diagram of the coefficient [a ₂ ] smoothed by the moving average method and the absolute value | Δ [a _{2, k} ] | of the change in the smoothed coefficient [a ₂ ]. By appropriately setting the threshold value [a _{2, k} ] of the absolute value | Δ [a _{2, k} ] | of the change of the coefficient [a ₂ ], for example, between 145 s to 146 s and 149 s to 150 s The vibration frequency f in between can be eliminated.

Furthermore, since these criteria are strict and there are few areas of estimated values obtained, areas that seem to be appropriate for inspection may be excluded. Therefore, as a preventive measure, the following fourth reference for determining whether or not the value of the vibration frequency f in the region adjacent to the estimation region is appropriate is introduced. That is, with respect to the average value f _av of the vibration frequency f in the estimated region, the estimated region is expanded to a region where the value of the vibration frequency f in the adjacent region satisfies the condition of the expression (25). As a result, the validity of the vibration frequency is determined for vibration frequencies within ± 15% of the average value of vibration frequencies in the estimation region.

[Equation 25]
_{1.15f av ≧ f ≧ 0.85f av ...} (25)
In addition, as a determination of the validity of the vibration frequency, the vibration frequency f in which the constant time length of the vibration frequency f in the estimation region is equal to or less than a predetermined time length is excluded.

FIG. 7 is a distribution diagram of the obtained vibration frequency estimation value T _T. The distribution of the vibration frequency estimated value f _T obtained by applying the validity criterion to the vibration frequency f which is the measurement result is shown. The vibration frequency estimation value f _T is distributed in the range of 0.8 Hz to 1.0 Hz between 146.2 s and 147.0 s. It is distributed in almost the same range as that determined by inspection from the characteristics of the vibration frequency f. In the case of FIG. 7, it can be seen that the vibration frequency estimation value f _T is approximately 0.9 Hz. The rotational frequency n _i of the wind power generator is calculated and specified by substituting the estimated vibration frequency f _T into f _ti in the equation (1).

In the above description, the case of the pole number switching type wind power generator has been described. Next, the case of applying to a variable speed type wind power generator will be described. The variable-speed generator, the tower - shadow effect theory oscillation frequency by f _t is different from the generator of the number of poles switching type, generally the frequency f ₀ corresponding to the synchronous speed of about ± 30% of the area at the center fluctuate. Therefore, the existence region where the oscillation of the output power signal is estimated to be present, that is, the region of about ± 30% around the frequency f ₀ corresponding to the synchronization speed is divided into n equal parts, and the center frequency of each region divided into n parts. _{Is the} theoretical frequency f _ti (i = 1, 2, 3,... N) theoretically estimated as the vibration of the output power signal P (t). Practically, for example, it is divided into three, and in each divided area, the processing shown in FIG. 1 centering on each theoretical frequency f _t1 , f _t2 , f _t3 is performed.

FIG. 8 is a characteristic diagram of the characteristics of the output power P (t) of the pole number switching type wind power generator and the characteristics of the output power P (t) of the variable speed type wind power generator. As shown in FIG. 8 (a), around the output power P characteristic of the (t) is the theoretical frequency f _t of the wind power generator of the number of poles switching type tower - becomes a characteristic shown vibration caused by the shadow effect, variable As shown in FIG. 8B, the output power P (t) of the variable speed wind power generator has a very small output pulsation amount due to the tower shadow effect as compared with the pole number switching type. This is because the control system of the wind power generator performs control for suppressing fluctuations in the output power P (t).

Therefore, when the process shown in FIG. 1 is applied to the output power P (t) of the variable speed wind power generator as it is, for example, a band filter centered on the frequency f ₀ corresponding to the synchronous speed is used. When applied twice continuously, there is a concern that the vibration in this vicinity is overlooked because there is about 35% attenuation at the end of the frequency range. In order to cope with the attenuation at this end, the region of about ± 30% is divided into n equal parts around the frequency f ₀ corresponding to the synchronization speed, and the center frequency of each divided region is divided into the output power signal P. The theoretical frequency f _ti (i = 1, 2, 3,... N) theoretically estimated as the vibration of (t) is assumed.

For example, the end attenuation is suppressed to about 10% by dividing the frequency existence range into three equal parts. Furthermore, by simulating the center frequency of each region to the theoretical frequency and performing frequency estimation in the processing of FIG. 1 for each region, frequency estimation is possible even with a variable speed type wind power generator. is there. In this case, the pseudo theoretical frequency theoretical frequency f _ti (i = 1, 2, 3) is defined as follows.

Here, f _tmax is the maximum value of the frequency by the tower shadow effect, f _tmin is the minimum value of the frequency by the tower shadow effect.

Next, also in the frequency estimation of the variable speed type wind power generator, the output power signal of the section below the predetermined threshold value P _{B, C} among the output power signal P _B (t) of the wind power generator that has passed through the bandpass filter. P _B (t) is excluded as necessary.

FIG. 9 is a trend diagram of the frequency estimation value f _T when the predetermined threshold value P _{B, C} is not applied. As can be seen from FIG. 9, when the predetermined threshold value P _{B, C} is not applied, it can be seen that there are estimated frequencies that are substantially different at the same time after 150 seconds, which is physically unreasonable. The reason is that when the vibration amplitude is small, the reliability of the estimation result is low, and it is desirable to exclude the estimation result in a region where the pulsation of the output power signal P _B (t) is small.

Therefore, as shown in FIG. 10, a predetermined threshold value P _{B, C} is applied to the output power signal P _B (t) that has passed through the bandpass filter, and vibrations in the time domain where the vibration amplitude is small are excluded. As a result, as shown in FIG. 11, physically unreasonable those are excluded, the trend of the frequency estimate f _T of the wind power generator can be obtained.

  As described above, according to the embodiment of the present invention, the autoregressive model of the damped oscillation of the output power signal is created based on the time series data of the output power signal, and the output power signal is calculated from the coefficient of the autoregressive model. Since the vibration frequency is calculated, the vibration frequency of the output power signal can be calculated even with time series data of a short vibration period.

  That is, since the Fourier transform is a constant vibration model, for example, when the control system of the wind power generator starts to suppress and attenuates vibrations caused by changes in wind speed, especially when the duration of vibration is short However, in the embodiment of the present invention, since the damped vibration waveform model is adopted a priori as the autoregressive model, the vibration of the output power signal can be obtained even with time series data of a short vibration period. The frequency can be calculated.

On the other hand, since the coefficients a ₁ and a ₂ of the autoregressive model are obtained by the method of least squares, the resolution is not defined by the sampling interval for obtaining the time series data, and is a calculation for calculating the vibration frequency as compared with the case of the Fourier transform. The amount is small. Furthermore, since a damped vibration waveform model is assumed, the validity of the vibration frequency as an estimation result can be easily determined from the correspondence with the time constant of the wind power generator obtained from actual data.

The flowchart of the operating state discrimination | determination method of the wind power generator concerning embodiment of this invention. Versus frequency characteristic diagram of the passing rate K _B of the bandpass filter used in the embodiment of the present invention. In the embodiment of the present invention, the output power signal P _B (t) of the wind power generator when the band-pass filter is applied to the measured output power signal P (t) of the wind power generator twice in succession. Frequency trend diagram. Trend diagram showing a relationship between a predetermined threshold value P _{B, C} the output power passed through the bandpass filter signal P _B (t) in the embodiment of the present invention. In the embodiment of the present invention, trend diagram of the coefficient a _2, including a time constant T indicates output power signal P _B of the wind power generator, the oscillation frequency f and damping vibrations. In the embodiment of the present invention, the trend diagram of the coefficient [a ₂ ] smoothed by the moving average method and the absolute value | Δ [a _{2, k} ] | of the change of the smoothed coefficient [a ₂ ] . In the embodiment of the present invention, the distribution diagram of the obtained vibration frequency estimation value T _T. The characteristic view of the characteristic of the output electric power P (t) of a pole number switching type wind generator and the characteristic of the output electric power P (t) of a variable speed type wind power generator. Trends diagram of a frequency estimate f _T when the output power signal P _B of the wind turbine that has passed through the band-pass filter (t) does not apply a predetermined threshold value P _B, the _C. Trend diagram showing output power signal P _B that has passed through the band-pass filter (t) with a predetermined threshold value P _B, the _C. Predetermined threshold value to the output power signal P _B of the wind turbine that has passed through the band-pass filter (t) P _B, trend diagram of a frequency estimate f _T in the case of applying the _C.

Explanation of symbols

S1 ... Output power reading process, S2 ... Band filtering process, S3 ... Time series data accumulation process, S4 ... Autoregressive model creation process, S5 ... Vibration frequency calculation process, S6 ... Vibration frequency validity determination process, S7 ... Vibration frequency estimation Value extraction process, S8 ... rotational speed characteristic process

Claims (11)

  The output power signal of the pole-switching wind power generator that includes the measured vibration component is sampled at a predetermined period, and time series data of the output power signal is input, which is theoretically estimated as the vibration of the output power signal. The time series data is passed through a band filter based on the moving average method centering on the theoretical vibration frequency, and the time series data of the output power signal from which the low frequency component and high frequency component are removed by the band filter is integrated, and the integrated output Apply the power signal time-series data to the damped vibration auto-regression model prepared in advance, calculate the vibration frequency of the output power signal from the coefficient of the auto-regression model to which the time-series data is applied, and obtain it based on the predetermined condition. The validity of the obtained vibration frequency is determined, the vibration frequency satisfying the predetermined condition is extracted as the vibration frequency estimated value, and the wind turbine generator is rotated based on the vibration frequency estimated value. Operating state discrimination method of a wind power generator, wherein the identifying the.   Presence that the output power signal of the variable speed type wind generator including the measured vibration component is sampled at a predetermined period and the time series data of the output power signal is input, and the vibration of the output power signal is estimated to exist The center frequency of each region obtained by dividing the region into n equal parts is the theoretical frequency that is theoretically estimated as the vibration of the output power signal, and the time series data is applied to the bandpass filter based on the moving average method centered on each theoretical frequency. The time-series data of the output power signal from which the low-frequency component and the high-frequency component are removed by the band filter for each theoretical frequency is integrated, and the time-series data of the output power signal for each integrated theoretical frequency is collected. Applied to the autoregressive model of damping vibration prepared in advance, the vibration frequency of the output power signal is calculated from the coefficient of the autoregressive model applying the time series data for each theoretical frequency, and based on the predetermined condition. The validity of each obtained vibration frequency is determined, each vibration frequency satisfying a predetermined condition is extracted as each vibration frequency estimated value, and the rotation speed of the wind power generator is specified based on each vibration frequency estimated value. A method for determining the operating state of a wind power generator.   The wind power generator operating state determination method according to claim 1 or 2, wherein an output power signal of the wind power generator including the measured vibration component is passed through the band filter twice.   4. The wind power generator according to claim 1, wherein output power signals of a section below a predetermined threshold are excluded from the output power signals of the wind power generator that have passed through the bandpass filter. 5. Operation state discrimination method.   5. The validity of the vibration frequency is determined under the condition that the absolute value of the sine term of the coefficient including the sine term indicating vibration is 1 or less among the coefficients of the autoregressive model. The wind power generator operating state determination method according to any one of the above.   6. The validity of the vibration frequency is determined under the condition that the value of the time constant of the coefficient including the time constant indicating the damped vibration among the coefficients of the autoregressive model is positive. The wind power generator operating state determination method according to any one of the above.   Among the coefficients of the autoregressive model, the validity of the vibration frequency is determined under the condition that the range of the coefficient including the time constant indicating the damped vibration is larger than the value estimated based on the time series data. The operating state discrimination method for a wind power generator according to claim 6.   Among the coefficients of the autoregressive model, the range of the coefficient including the time constant indicating the damped oscillation is not less than the value estimated based on the time series data, and the margin value is a value when the time constant is infinite. The method of determining an operating state of a wind power generator according to claim 6, wherein the validity of the vibration frequency is determined under a condition that is less than or equal to a value that takes into account.   9. The validity of a vibration frequency is determined under a condition that, among the coefficients of the autoregressive model, a variation of a coefficient including a time constant indicating a damped vibration is smaller than a predetermined threshold value. The operating state discrimination method of the wind power generator as described in any one of.   10. The wind power generation according to claim 1, wherein the determination of the appropriateness of the vibration frequency is for a vibration frequency within ± 15% of an average value of the vibration frequency in the estimation region. Method for determining the operating state of the machine. 11. The vibration frequency having a predetermined length of vibration frequency that is not longer than a predetermined time length is excluded as the determination of the validity of the vibration frequency. The operation state discrimination method of the wind power generator of one description.

Images