JP2013222423A - Method and system for predicting power generation capacity, and method and system for managing health of wind power generation facilities - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to predict changes in velocity of the wind blowing on a wind power generation system and control output of a power system generator on the basis of the prediction.SOLUTION: A power generation capacity prediction system comprises; a data acquisition unit which acquires data on wind condition, weather information, and power generation capacity of a wind power generation system; a temporal database unit which stores past temporal data on wind condition, weather information, and power generation capacity of the wind power generation system acquired by the data acquisition unit; wind condition prediction unit which predicts future wind condition using the data on current wind condition, weather information, and power generation capacity of the wind power generation system acquired by the data acquisition unit and information stored in the temporal database unit; a power generation capacity prediction unit which predicts future power generation capacity of the wind power generation system using the wind condition information predicted by the wind condition prediction unit; and an output unit which outputs the information on the future wind condition predicted by the wind condition prediction unit and the information on the future power generation capacity of the wind power generation system predicted by the power generation capacity prediction unit.

Description

  The present invention predicts a power generation amount of a wind power generation facility and performs output control of a power system side generator based on a predicted value, and relates to a power generation amount prediction method and a system thereof. Furthermore, the present invention relates to a health management method for wind power generation equipment and a system for judging the health condition of the wind power generation equipment.

  A wind power generation facility (system) is a clean energy source that does not emit carbon dioxide during power generation. However, wind power generation facilities depend on the installed environment, and the amount of power generation increases or decreases depending on the wind conditions. When this wind power generation facility is linked to the grid, an unexpected increase or decrease in the amount of power generation causes a power instability factor in the grid linkage and becomes a social problem. For this reason, it is necessary to predict the power generation amount of the wind power generation facility and perform output control of the power system side generator based on the predicted value.

  In order to predict the wind conditions, the wind speed and direction of the wind generator were measured using a Doppler radar using radio waves, and the output value of the wind generator was predicted from these values. However, in general, since natural energy depends on the weather, the amount of power generation is uncertain and difficult to predict. The increase in the amount of power generation can be controlled (suppressed) by the wind power generator itself, but it is important to predict the decrease in the amount of power generation by some means and substitute it with another power generation facility. In particular, considering the prediction of the order of several tens of minutes later as a margin for performing output control of the power system side generator, the power generation amount prediction accuracy is not sufficient.

  Non-Patent Document 1 introduces a combined output smoothing system using a storage battery. As one method, a method of notifying a power company or the like in advance of a combined output of wind power generation output and storage battery output and performing constant output control is being studied. In order to implement constant output control, it is necessary to make an operation plan based on the forecast information of wind power generation after 2-3 days from a few hours ahead of time and notify the power company. When performing the notification operation, the predicted wind power generation output includes an error. Therefore, during actual operation, the storage battery is controlled while taking this error into consideration. Naturally, the larger the inverter output and kWh capacity of the storage battery, the easier it is to control the combined output as informed in advance, and the deviation rate is reduced. However, when a large-scale wind power generation facility is taken into consideration, the required storage battery becomes enormous. Therefore, when considering business feasibility, a storage battery with as little capacity as possible is desirable, and the output prediction of wind power generation is also accurate. A good approach is required.

  In Tohoku Electric Power Co., Ltd., 20 minutes starting at any time, the “maximum value-minimum value” of the combined output of wind power generation equipment (average value for 1 minute) may be 10% or less of the total rated output of the wind power generator. It has been established. For this purpose, it is said that it is necessary to introduce a storage battery such as a lead storage battery into a wind power generation facility.

  Patent Document 1 stores past data related to past power generation amount of a wind power generator, and wind power generation based on a statistical correlation between different times in the past data or a statistical correlation between different generator positions. A method is described in which a predicted value related to the power generation amount of a machine is calculated as time-series data including the occurrence probability. Similarly, Patent Document 2 similarly describes a method of predicting the wind direction using past data and controlling the direction of the windmill. Patent document 3 is also the same, and it is described that a wind condition is predicted and a wind turbine is controlled to aim at maximization of power generation and profit.

  Non-Patent Document 2 shows an example in which the next day power load prediction is performed by a Gaussian process. This is to predict the power load, not the amount of power generated by the wind power generation facility, and data such as wind conditions are not used.

  Non-Patent Document 3 shows a method for estimating the power generation output of a short time ahead and 10 minutes ahead in photovoltaic power generation that is linked to the grid based on the multiple regression equation based on the amount of solar radiation. However, a method of overlearning past data such as multiple regression is adopted, and there is a problem in estimation reliability.

  Furthermore, Non-Patent Document 4 shows that a multi-step prediction is constructed online and nonlinear prediction of wind speed on a scale from seconds to minutes is performed using past time-series data.

JP 2011-200040 A JP-A-2005-98181 JP 2008-64081 A

Ryoichi Tanikawa: Examination of wind power output prediction method for constant power output control method for wind power generation facilities with storage batteries, etc., 13th Japan Society of Mechanical Engineers Power and Energy Technology Symposium (June 2008) Shotaro Omi, Hiroyuki Mori: Short-term power load prediction expressing uncertainty by Gaussian process, IEEJ Transactions B, Vol. 126-B, pp. 202-208, 2006-02-01 Tatsuya Nagai: Output estimation method for short-term solar power plant based on solar radiation forecast, Outline of Master's thesis, Graduate School of Information Science, Hokkaido University (February 14, 2011) Yoshito Hirata et al .: Real-time prediction of local wind speed from second to minute scale, 2012 IEEJ National Convention Proceedings, 7th volume 7-064 p95, 2012-3-21

  The power generation amount prediction of the wind power generation facility on the order of several tens of minutes is realized as a margin for performing output control of the power system side generator. The present invention stores past wind condition time series data and past weather data time series data, and extracts past wind condition time series data similar to the current wind condition / weather time series data. Based on the wind condition prediction, the power generation amount is estimated.

  In the method based on the covariance matrix of past data disclosed in Patent Document 1, the past data is assumed to be Gaussian and random numbers are generated to obtain the distribution. However, the distribution is obtained rather than predicted. Not too much.

  The method disclosed in Patent Document 2 merely outputs an average behavior of past data as a predicted value.

  In addition, the method disclosed in Patent Document 3 does not consider performing output control of the power system side generator.

  The method disclosed in Non-Patent Document 1 inputs RSM-GPV of the Japan Meteorological Agency, performs a wide range of prediction simulation including wind power generation equipment by a weather prediction simulation model, and is involved in the wind power generation equipment of interest. It is not possible to obtain highly accurate data.

  The method disclosed in Non-Patent Document 2 does not predict the power generation amount of the wind power generation facility as described above.

  In the method disclosed in Non-Patent Document 3, a method is used in which the power generation output is estimated for a short time and 10 minutes ahead by a method of overlearning past data such as multiple regression based on the amount of solar radiation. And there is a problem with the estimated reliability.

  Further, the method for predicting the local wind speed in real time disclosed in Non-Patent Document 4 is to predict the future wind speed in the order of minutes based on the wind information obtained from the wind power generator. Therefore, information around the wind power generator is not reflected, and it is difficult to increase the reliability of the prediction.

  Furthermore, when considering system linkage, the effect of the wind conditions is significant, but there are other projects that should be taken into account, such as deterioration and failure of the wind power generation equipment itself, and deterioration of the storage battery. However, none of the above-mentioned patent documents and non-patent documents can determine the health status of wind power generation facilities and storage batteries while predicting the amount of power generation that depends on the influence of wind conditions. The generator output cannot be controlled.

  Therefore, in the present invention, past wind condition time series data, weather time series data, etc. are stored, and past wind condition time series data similar to the current wind condition / weather time series data is extracted, The present invention provides a power generation amount prediction method and system for predicting wind conditions several tens of minutes ahead based on the above and estimating the power generation amount of wind power generation facilities. Furthermore, the present invention provides a health management method for wind power generation equipment and a system for judging the health status of wind power generation equipment and storage batteries while performing wind condition prediction and power generation amount estimation.

  In order to solve the above-described problems, in the present invention, in a method for predicting the amount of power generated by a wind power generation facility, past points of a point where the wind power generation facility is installed and points around the point where the wind power generation facility is installed The wind condition ahead of the present is predicted using the wind time series data and the current wind condition data, and the power generation beyond the present is predicted based on the predicted wind condition.

  Further, in order to solve the above-described problems, in the present invention, in a method for determining the health state of a wind power generation facility, a point where the wind power generation facility is installed and points around the point where the wind power generation facility is installed Measure wind condition data and power generation amount of wind power generation equipment, extract the feature quantity of the observation data including the measured wind condition data and power generation quantity, and extract the feature quantity and the data stored in the database Is used to detect an abnormality or sign of an abnormality in the wind power generation facility, and based on the result of the detection or sign, the health state of the wind power generation facility is determined.

  In order to solve the above-described problem, in the present invention, the power generation amount prediction system includes a wind condition and weather information at a point where the wind power generation facility is installed and points around the point where the wind power generation facility is installed, and A data acquisition unit that acquires power generation data, the wind power generation facilities acquired by this data acquisition unit, and the past time-series wind conditions and weather at the points around the point where the wind power generation facilities are installed Time series database section that stores information and power generation data, wind conditions and weather information at the location where the current wind power generation equipment acquired by the data acquisition section is installed, and around the location where this wind power generation equipment is installed And wind power generation using the wind energy information predicted by the wind condition prediction unit and the wind condition prediction unit for predicting the wind condition beyond the present using the power generation amount data and the information stored in the time series database unit. The equipment is now The power generation amount prediction unit that predicts the previous power generation amount, the information on the wind state ahead of the present predicted by the wind state prediction unit, and the information on the power generation amount beyond the present of the wind power generation facility predicted by the power generation amount prediction unit And an output unit for outputting.

  Furthermore, in order to solve the above-described problems, in the present invention, a wind power generation facility health management system includes a wind condition at a point where the wind power generation facility is installed and a point around the point where the wind power generation facility is installed. Wind condition data measurement unit that measures wind power and obtains wind condition data, power generation amount measurement part that measures the amount of power generated by the wind power generation facility, and wind condition data and power generation that are measured and acquired by the wind condition data measurement unit Using the observation data feature quantity extraction unit that extracts the feature quantity of the observation data including the power generation amount measured by the quantity measurement unit, the feature quantity extracted by this observation data feature quantity extraction unit, and the data stored in the database An abnormality detection unit for detecting an abnormality or a sign of abnormality of the wind power generation facility, and a health state determination unit for determining the health state of the wind power generation facility based on the result of the abnormality detection or the sign of abnormality performed by the abnormality detection unit. Configuration It was.

  ADVANTAGE OF THE INVENTION According to this invention, the fluctuation | variation of the wind speed concerning a wind power generation equipment can be estimated, and output control of an electric power system side generator can be performed based on it. Moreover, according to this invention, the time of the wind speed concerning wind power generation equipment falling can be estimated and issued.

  Further, the method of notifying a power company or the like in advance of the combined output of the wind power generation output and the storage battery output and performing the constant output control has an economical effect of requiring a storage battery with as little capacity as possible.

  At the same time, it is possible to determine the health state of the wind power generation facility and the storage battery, and to control the output of the power system side generator in preparation for occurrence of an abnormality.

FIG. 1 is a graph showing an example of time series data of wind power among wind conditions affecting wind power generation equipment and time series data of power generation amount of the wind power generation equipment obtained as a result. FIG. 2 is a plan view showing an example of a wind power generation facility and a wind condition measuring device arranged so as to surround it. FIG. 3 is a block diagram showing a configuration of a system for predicting a future wind condition from time series data related to a past wind condition according to an embodiment of the present invention. FIG. 4 is a pie chart showing the behavior of wind speed and wind direction as time-series wind condition data in a polar coordinate format. FIG. 5 is a graph showing the behavior of wind speed and wind direction as time series wind condition data in a time series format. FIG. 6 is a graph for explaining estimation by a regression method such as a Gaussian process or a prediction method. FIG. 7 is a graph showing an example of past data for wind condition prediction by the Gaussian process. FIG. 8A is a graph showing an example of a result of wind condition prediction performed by a Gaussian process. FIG. 8B is a graph showing an example in which information on the dispersion of predicted values is added to the result of wind condition prediction by the Gaussian process and displayed. FIG. 8C is a graph showing an example in which information on the dispersion of the predicted value is added to the information on the result of predicting the behavior of the wind speed and the wind direction by the Gaussian process. FIG. 9 is a block diagram showing a configuration of a system for predicting a future wind condition from time-series data related to a past wind condition according to the first modification of the embodiment of the present invention. FIG. 10 shows a prediction using the identification unit 17 by the recognition engine in the wind condition prediction unit 151 in the system for predicting the future wind condition from the time series data related to the past wind condition according to the first modification of the embodiment of the present invention. It is a block diagram which shows the operation | movement block in performing. FIG. 11 is a graph showing an example of an identification method by the recognition engine of the identification unit 17 according to the first modification. FIG. 12 is a graph showing an example of past data for wind condition prediction according to the first modification. FIG. 13 is a graph showing an example of a result of wind condition prediction according to the first modification. FIG. 14 is a block diagram showing a configuration of a system for predicting a future wind condition from time series data related to a past wind condition according to the second modification of the present invention. FIG. 15A is a plan view showing another example of a wind power generation facility and a wind condition measuring device arranged to surround the wind power generation facility. FIG. 15B shows a wind power generation facility and a wind condition measuring device arranged so as to surround the wind power generation facility and an air flow measuring device arranged so as to surround it. FIG. FIG. 16 is a graph showing the wind speed obtained by converting the wind condition data obtained by the wind condition measuring instrument into the wind power generation position in the second modification. FIG. 17 is a graph showing the wind speed obtained by converting the difference (gradient) in the atmospheric pressure obtained from the result of atmospheric pressure measurement in Modification 2 into the wind power generation position. FIG. 18 is a diagram for explaining the concept of a particle filter according to the second modification. FIG. 19 is a plan view showing an example of a wind condition measuring device arranged so as to surround the wind power generation facility and another wind power generation facility. FIG. 20 is a block diagram for explaining an abnormality detection method for a wind power generation facility. FIG. 21 is a block diagram showing the configuration of a system for predicting wind conditions using the movement of clouds.

  The present invention relates to a method for predicting wind speed fluctuations in a wind power generation facility and issuing a report based on the predicted wind speed to control the output of a power system side generator. Predict future wind conditions from time series data. In addition, it predicts and reports when the wind speed applied to the wind power generation facility decreases.

  For this reason, a wind condition measuring device is arranged so as to surround the wind power generation facility, the data is handled as time series data, and weather data such as atmospheric pressure is also used.

  Specifically, (1) Prediction using the Gaussian process, which is a nonlinear regression technique that can predict future wind conditions from past time-series data and evaluate their reliability, (2) Past Prediction applied to time-series data using k-NN recognition method for time-series data, (3) Modeling dynamics for time-series data, obtaining a state space model, and predicting with particle filter A highly accurate power generation amount prediction method is provided.

  Further, in the present invention, wind condition measuring devices are arranged so as to surround the wind power generation equipment at a necessary ratio, and the data is used as time series data. We also decided to use weather data such as atmospheric pressure.

Moreover, in this invention, the health state of a wind power generation installation or a storage battery is judged, estimating the electric power generation amount.
Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 shows time-series data of wind speed 1 and power generation amount 2 in a wind power generation facility targeted by the present invention. The wind speed is measured by a wind condition measuring instrument and is observed together with the wind direction (not shown). The horizontal axis is time, and data is acquired in units of 1 minute, for example. In FIG. 1, a decrease in the amount of power generation is seen in a region 3 surrounded by a square. When this decrease occurs, it is necessary to quickly improve the output of the power system side generator so as to meet the power demand forecast. Therefore, it is desirable to output a power generation amount decrease forecast when a decrease in the power generation amount is predicted. In the present embodiment, a method for outputting a forecast of the power generation amount decrease will be described.

  FIG. 2 shows an arrangement example of the wind power generation equipment 4 and a plurality of wind condition measuring devices 5a to 5h provided around the wind power generation equipment 4. The wind gauge measures the wind speed and direction. The height from the ground is also an important parameter. In this example, although the arrangement | positioning which surrounds the focused wind power generation equipment was shown, the wind condition measuring device should just be arrange | positioned in consideration of the main wind direction. FIG. 2 shows an example in which a plurality of wind condition measuring devices 5a to 5h are arranged at equal distances from the wind power generation equipment 4, but the wind power generation equipment actually depends on the topography and the surrounding environment. In many cases, the distance from 4 and the interval between the wind condition measuring devices 5a to 5h are not the same. Of course, the wind condition measuring device may be anything as long as it can measure the wind condition, and may be a wind power generation facility itself. In any case, time series data of wind speed and direction is acquired from a plurality of wind condition measuring means.

  FIG. 3 is a block diagram illustrating a wind condition and power generation amount prediction system and a prediction method for predicting a future wind condition and a power generation amount from time-series data related to the past wind conditions according to the present invention. The wind condition and power generation amount prediction system includes a time series wind condition / weather / power generation amount data acquisition unit 10, a time series database 11, a wind condition prediction unit 15, a power generation amount prediction unit 20, and a display unit 26. Yes.

  In the time-series wind condition / weather / power generation data acquisition unit 10, current time-series wind condition / weather / power generation data is obtained from the wind condition measuring devices 5a to 5h and the wind power generation equipment 4 itself. This is stored in the time series database 11 every time it is acquired, and can be used as past information. Data accumulated in the time series database 11 includes wind condition data 12, weather data 13, and power generation amount data 14. The weather data 13 stores weather such as rain and fine weather, temperature, humidity, and atmospheric pressure arrangement.

  The wind condition prediction unit 15 refers to the data stored in the time series database 11 and makes a wind condition prediction by the similar time series data selection unit 16 and the identification unit 18 by the regression engine.

  The similar time-series data selection unit 16 is similar to the past time-series wind condition data 12 and the current time-series wind condition / weather / power generation data input from the time-series wind condition / weather / power generation data acquisition unit 10. Meteorological data 13 and power generation amount data 14 are selected from the database 11.

  The identification unit 18 performs future prediction on the selected past time-series wind condition data 12, weather data 13, and power generation amount data 14 by a regression equation to obtain wind condition data 25. The regression method will be described later with reference to FIGS.

  The power generation amount prediction unit 20 performs power generation amount prediction from the obtained wind condition data 25. The display unit 26 displays the wind condition data stored in the time series database 11, the power generation amount reduction report predicted by the power generation amount prediction unit 20, and the power generation amount waveform on the screen and outputs them. .

  An example of display will be described with reference to FIGS. FIG. 4 gives a way of expressing time-series wind condition data obtained from, for example, the wind condition measuring instrument 5a. In the pie chart 6 of FIG. 4, the wind speed is represented by a length from the central origin (for example, unit m / s), and the wind direction (azimuth) is represented by an angle. It shows wind condition data for a certain period.

  On the other hand, in the graph 8 of the Cartesian coordinate system in FIG. 5, the wind direction (azimuth 7) is represented by angle information as in the case of the pie chart 6 in FIG. 4, but the wind speed is represented by the length from the horizontal axis ( It is represented by a vector of arrows) and is displayed as time series data. In this example, since the frequency of the south wind and the north wind is small, the direction is arranged so that the east wind and the west wind can be well understood. Conversely, if the frequency of the south and north winds is large and the frequency of the east and west winds is small, the orientation should be arranged so that the south and north winds can be clearly understood. You can create a graph. As a power generation amount prediction system, the wind condition is represented by these notations.

  The display method is one expression form of wind conditions. Therefore, in the prediction described later, this expression form may be treated as an observation sensor signal (vector). For example, the wind speed and direction are represented by a radius vector centered on the origin of FIG. In FIG. 5, time-series arrow vectors are observation data.

  Next, a wind condition prediction method will be described with reference to FIGS. FIG. 6 is a diagram for explaining estimation by a nonlinear regression method such as a Gaussian process or a prediction method. Learning data 21 and a regression function 22 are drawn. Here, the learning data 21 is past time-series wind condition / weather / power generation data that is similar to the current time-series wind condition / weather / power generation data. Here, the similarity is selected within a possible range.

  There are many commentary articles on the Gaussian process, but here is a report from the academic conference “Takusaku Ozaki, Toshikazu Wada, Shunji Maeda, Hisae Shibuya, Pattern Recognition and Media Understanding (PRMU) , Image Engineering (IE), 133-138 (2011.5). The characteristic of the Gaussian process is that data similar to learning data can be selected and output, and its reliability can be output as variance.

In the Gaussian process, given an output t = (t ₁ , t ₂ ,... T _n ) ^T corresponding to an input vector x ₁ , x ₂ ,... X _n , an output t _{n + 1} for a new input vector x _{n + 1.} Predict.

Here, σ represents a dispersion parameter, and β represents noise. From (Equation 1), an output t _est is obtained as a predicted value.

  "Kai Goebe: Prognostics in Battery Health Management, IEEE Instrumentation And Measurement Magazine (2008), Volume: 11, Issue: 4, Pages: 33-40" covers the Gaussian process and the following A technique for estimating Remaining-useful-life (RUL) of a storage battery using a particle filter is introduced.

  One of the technical differences is that as far as possible in the past, or if the data is time-scaled (sampling) in order to ensure accuracy, the capacity increases. This is a point where such a huge amount of similar time series data is selected one by one in advance, which reduces the load of the calculation shown in (Equation 1), particularly the inverse matrix calculation, and makes a short-term (short time) prediction. Is possible.

FIG. 7 shows an example of wind condition prediction by the Gaussian process. In the figure, the upper waveform 701 is the electric energy, and the lower waveform 702 is the wind speed. Portions 711 and 712 surrounded by a square are time series data up to the present time. The period is set as one month. Of course, the longer one can cover various phenomena and expect high accuracy. The waveform of the time series, collectively as vectors, denoted as X, X ₁ is waveform data, X ₂ wind speed represents the waveform data of the amount of power.

  FIG. 8A shows the prediction stage. With respect to the waveform 701 of the electric energy and the waveform 702 of the wind speed, ◯ marks 721 and 722 indicate the wind condition prediction results by the Gaussian process from the waveforms surrounded by the squares 711 and 712, respectively. Later, data on the right (later in time) than the waveform enclosed by the squares 711 and 712 will actually be obtained. Here, the time increment is t.

In this graph, it is expected that at time t ₂ , the wind speed decreases and the power generation amount is greatly reduced, and it becomes impossible to secure the prescribed power generation amount. The prediction result is displayed on the screen of the display unit 26 and a warning is issued. At the same time, information on the prediction of power generation reduction is transmitted to the upper power control system.

Further, FIG. 8B shows the wind condition prediction result by the Gaussian process from the waveforms surrounded by the squares 711 and 712 with respect to the electric energy waveform 701 and the wind velocity waveform 702, respectively. _The graph which added and displayed the dispersion _| distribution information of _est is also shown. In the graph of FIG. 8B, as dispersion, ± 3σ values 811 and 812 with respect to the predicted electric energy value 810 at each time after the current time 801, and ± 3σ values 821 and 822 with respect to the predicted wind speed value 820 are shown. The displayed state is shown.

Further, in FIG. 8C, the dispersion of the output _test shown in (Expression 1) is added as a wind condition prediction result by the Gaussian process to the graph showing the result of predicting the wind direction and the wind speed described in FIG. Shows the displayed graph. In the graph of FIG. 8C, the wind direction and the wind speed at each time before the current time 850 estimated from the data in the range enclosed by the square 852 among the wind direction and wind speed data 851 before (past) the current time 850. The information 860 is displayed on a graph, and the color is displayed according to the size of the dispersion as the dispersion information of the output _test shown in (Equation 1). In other words, a reference value for dispersion is set in advance, and when the estimated wind direction and wind speed dispersion is smaller than the preset dispersion value, it is displayed with a red arrow, and when it is greater than the preset dispersion value, it is displayed with a blue arrow. . Instead of displaying color-coded according to the degree of dispersion, it may be displayed in shades.

[Modification 1]
Next, modification example 1 of wind condition prediction will be described with reference to FIGS.
In FIG. 9, the structure of the wind condition and power generation amount prediction system in the modification 1 is shown. In this system, the identification unit 18 by the regression engine in the wind condition prediction unit 15 of the wind condition and power generation amount prediction system described in FIG. 3 in the first embodiment is used as the identification unit 17 by the recognition engine as the wind condition prediction unit 151. It is a replacement.

  FIG. 10 shows an operation block when the wind condition prediction unit 151 performs prediction using the recognition unit 17 by the recognition engine. The wind condition data, weather data, and power generation data are input from the time series wind condition / weather / power generation data acquisition unit 10 to the recognition engine 17, and the past time series wind data 12, weather data 13, and power generation data 14 are input. Is read from the database 11 and recognized. The past time-series wind condition data 12, weather data 13, and power generation amount data 14 stored in the database 11 are used as teaching learning data for determining parameters of the recognition engine of the identification unit 17 by the recognition engine.

  In the recognition unit 17 by the recognition engine, the current wind condition data, weather data / power generation data input from the time series wind condition / weather / power generation data acquisition unit 10, and the past time series wind conditions read from the database 11. Using the data 12, the weather data 13, and the power generation amount data 14, a wind condition prediction value 251 is obtained by a recognition engine in which parameters are set by learning, and the result is output to the power generation amount prediction unit 201 side.

  FIG. 11 shows an example of the identification method by the recognition engine of the identification unit 17 by the recognition engine. Here, the k-NN method is mentioned, but the time trajectory is targeted in the sense that it is applied to time-series data. This is vectorized for a predetermined period with respect to the past time series data. As an unknown pattern, time-series data is vectorized for the period determined up to the present time, and the distance from the vector of the past time-series data is obtained, and the vector of past time-series data close to the unknown pattern based on the magnitude of this distance Select multiple items. For example, from five selected vectors of past time series data, five pieces of data that are several steps ahead are read out, and the center of gravity and the center of gravity with the inverse of the distance as the weight are predicted values of wind conditions and power generation amount. And The numerical value of 5 is merely an example.

FIG. 12 shows an example of wind condition prediction by the recognition engine 17. In the same figure, similarly to FIG. 7, the upper waveform 1201 is the electric energy, and the lower waveform 1202 is the wind speed. The portion surrounded by the squares 1211 and 1212 is time-series data up to the present time. The period is set to one month or two weeks. The target time series data may include not only wind conditions but also weather data. In other words, if the weather is similar and the wind conditions are similar, the data that is several steps ahead in time may be trusted as the predicted value. In the case of FIG. 12, as described in the first embodiment with reference to FIG. 7, time-series waveforms are collected as vectors and expressed as X, X ₁ is wind speed waveform data, and X ₂ is an electric energy waveform. Represents the data.

  FIG. 13 shows the prediction stage. With respect to the waveform 1201 of the electric energy and the waveform 1202 of the wind speed, the wind condition prediction results by the recognition engine are indicated by ◯ marks 1221 and 1222 from the waveforms surrounded by the squares 1211 and 1212, respectively. Later, data on the right (later in time) than the waveform enclosed by the squares 1211 and 1212 will be actually obtained. Here, the time increment is set to 1.

[Modification 2]
Next, the method of the modification 2 of a wind condition prediction is demonstrated using FIGS.
FIG. 14 shows an example of prediction by the particle filter 19.

  In FIG. 14, the structure of the wind condition and electric power generation amount prediction system in the modification 2 is shown. In this system, the identification unit 18 by the regression engine in the wind condition prediction unit 15 of the wind condition and power generation amount prediction system described in FIG. 3 in the first embodiment is used as the identification unit 19 by the particle filter as the wind condition prediction unit 152. It is a replacement.

  FIG. 10 shows an operation block in the case where the wind condition prediction unit 152 performs prediction using the particle filter identification unit 19. The wind condition data, weather data, and power generation data are input from the time series wind condition / weather / power generation data acquisition unit 10 to the recognition engine 17, and the past time series wind data 12, weather data 13, and power generation data 14 are input. Is read from the database 11 and recognized. The past time-series wind condition data 12, weather data 13, and power generation amount data 14 stored in the database 11 are used as teaching learning data for determining parameters of the recognition engine of the identification unit 17 by the recognition engine.

  In the recognition unit 17 by the recognition engine, the current wind condition data, weather data / power generation data input from the time series wind condition / weather / power generation data acquisition unit 10, and the past time series wind conditions read from the database 11. Using the data 12, the weather data 13, and the power generation amount data 14, a wind condition prediction value 251 is obtained by a recognition engine in which parameters are set by learning, and the result is output to the power generation amount prediction unit 201 side.

  (Equation 2) shows a model for prediction of time series data. (Equation 3) to (Equation 6) are each component models.

  It is important to consider a model that reflects the dynamics of wind power generation. Hereinafter, each component will be described.

  FIG. 15A shows an example in which the wind conditions are measured around the wind power generation facility 4 using the wind condition measuring devices 51a to 51g. Wind condition measuring instruments 51a to 51g are arranged at necessary locations according to the topography and the like. In the same figure, the wind condition which influences the position of the wind power generation equipment 4 is shown. It can be expressed as the sum of vector components at the position C of the wind power generation equipment 4 according to the wind speed and the wind direction.

  Similarly, FIG. 15B shows an example in which the wind condition is measured around the wind power generation facility 4 using the wind condition measuring instruments 51a to 51g. In this case, the “air column” 30 surrounded by the wind condition measuring devices 51a to 51g is considered, and the input / output of air to the “air column” 30 is considered, and the wind force received by the entire air column 30 is represented. The central arrow (wind vector) 150 in the figure is considered to be the wind speed of the “air column” 30 as a whole. Of course, it may be considered as a pressure gradient. The “air column” 30 does not have to be a cylinder, and has a shape that depends on the actual arrangement of the air flow measuring instruments 51a to 51g in a plan view. The height of the “air column” 30 is a height within a measurable range of the wind condition measuring devices 51a to 51g.

  In FIG. 16, the conversion wind speed of this wind condition vector 150 in the position C of the wind power generation equipment 4 is shown with a graph. The function fw makes it non-linear. This is a component corresponding to (Equation 3). FIG. 17 shows the difference (gradient) in atmospheric pressure at the position C of the wind power generation equipment 4 from the atmospheric pressure arrangement. From the pressure difference (gradient), the function fp is used to make non-linearity. This is a component corresponding to (Equation 4). (Equation 5) and (Equation 6) are models showing fluctuations different from the trend component. Although the difference is expressed as the second floor difference, any number of floor differences may be used. V is a noise term.

  From these, it is shown that state models such as a system model represented by (Equation 7) and an observation model represented by (Equation 8) are generated.

For details, see Tomoyuki Higuchi: Particle Filter, Journal of IEICE Vol.88, No.12,2005. FIG. 18 illustrates the operation of the particle filter. As a result, a prediction result as shown in FIG. 13 is obtained.

  FIG. 19 shows an example in which another wind power generation facility 4b is provided instead of the wind condition measuring device 51a. In this case, the same approach can be used.

  In the above-described embodiment, the first modification, and the second modification, the Gaussian process, the recognition engine, the particle (particle) filter, and the like have been described. However, these may be used in combination. For example, the output can be processed from various viewpoints such as instantaneousness, reliability, and the like, such as the earliest predicted time, the latest time, and the average time.

  According to the present embodiment, it is possible to predict the fluctuation of the wind speed applied to the wind power generation facility 4 and perform output control of the power system side generator based on the predicted fluctuation. Further, according to the present embodiment, it is possible to predict and report the time point at which the wind speed applied to the wind power generation equipment 4 decreases.

  Further, the method of notifying a power company or the like in advance of the combined output of the wind power generation output and the storage battery output and performing the constant output control has an economical effect of requiring a storage battery with as little capacity as possible. Furthermore, a charge / discharge plan for the storage battery can be established with high accuracy. In addition, when it is predicted that the combined output of the wind power generation output and the storage battery output will not reach the specified output due to the decrease in the wind speed, it is possible to predict the point in time and send information to the upper power generation control system. it can.

  In addition, the wind power generation facility can be maintained by using the wind condition data.

  Specifically, the transition of the power generation amount with respect to the wind condition data is accumulated, and the health condition of the wind power generation facility such as abnormality detection and abnormality sign can be determined on the observation data including the wind condition data and the power generation amount. Of course, if the control parameters of the wind power generation facility are also added to the observation data, a more reliable health check can be performed. As a realization method, for example, a subspace method described in JP 2010-191556 A can be used.

  FIG. 20 shows the method. Here, the wind condition data and the power generation amount are referred to as sensor data 40. Sensor data consisting of wind condition data and power generation amount is input (S201), features of each sensor data 40 are extracted (S202), and attention is paid to the similarity between the sensor data 40, and compact learning data consisting of normal cases. (S203), the generated learning data is modeled by a subspace method (LSC: Local Subspace Classifier) (45), and based on the distance relation between the observed sensor data and the subspace, the divergence degree of the observed sensor data is calculated. Obtaining (S204), the health state of the sensor data observed as an abnormal measure is determined (S205). Thereby, the health state of a wind power generation facility can be judged, estimating the electric power generation amount. More specifically, it is possible to monitor the health status of wind power generation facilities such as breakdowns due to natural disasters such as lightning and typhoons, deterioration of gears and bearings of gearboxes, and fatigue of blades. Naturally, a monitoring sensor signal separately attached to the wind power generation facility can also be used. These monitoring sensor signals are for failure detection such as temperature, pressure, rotation speed, and voltage.

  The wind power generation facility health condition determination method can also be applied to the storage battery health condition determination. In this way, it is possible to detect in advance power generation reductions due to wind speed reduction, wind power generation equipment deterioration or abnormalities, storage battery deterioration or power generation reductions, etc. The output control of the side generator can be performed, and the power supply service to the customer can be performed smoothly.

  In addition, grid connection requires the same quality as the power supplied by the power company, so if the voltage generated by wind power generation becomes overvoltage or undervoltage, or if the frequency rises or falls, the power company The quality of the entire system is adversely affected. A relay that detects these problems is installed, and when an abnormality in voltage or frequency is detected, it is immediately disconnected from the power company system. Depending on the content and type of anomaly sign, it can be used as a control input to the relay.

  Although the said Example demonstrated the wind power generation equipment, solar power generation is also natural energy and the method mentioned above is applicable to solar radiation amount etc. as an object.

  Finally, wind condition prediction using cloud movement will be explained. The type of cloud is identified, the height of the cloud is calculated from it, and then the movement of the cloud is measured. FIG. 21 shows the configuration. Reference numeral 51 denotes an observation camera system, and reference numeral 52 denotes a cloud type / motion measurement system. The moving speed and direction are calculated for each cloud height. These cloud movement information storage areas are provided in the time series database 11 of FIG. 14 and input to the wind condition prediction unit 15 to be used for wind speed prediction. Here, the low-level clouds such as stacked clouds, cumulus clouds, cumulonimbus clouds, and the like are clouds having an altitude of about 2000 m from the ground, and the altitude and movement are measured for these clouds.

  As the camera used for the observation camera system 51, an omni camera having a wide field of view is suitable. If there is anything else that can be used, it can be similarly realized by inputting it to the wind condition prediction unit 15 of FIG. If it is along a railway line, input the wind measurement data. If the airfield is nearby, enter wind measurement data obtained near the runway. Similarly, observation data from the Japan Meteorological Agency is input to the wind condition prediction unit 15. It is only necessary to add these components to the model formula (Formula 2) for prediction of time series data.

  The above-described embodiment can be used as input data to a simulation for predicting a wind condition at an arbitrary point based on inputs such as terrain conditions, wind condition observation data, surface roughness, and wind turbine installation conditions. . Simulation software includes those from RIAM-COMPACT, LAWEPS developed by NEDO, and WAsP developed by Royal RISO Laboratories. Coarse spatial resolution is a weak point, which can be interpolated. Being able to use each other is an advantage.

  In the simulation, it is said that the complexity of the terrain affects the wind condition prediction accuracy. For example, 1. RIX (Ruggedness Index), 2. Roughness, 3. Valley density, 4. High dispersion, 5. Gradient, 6. Laplacian of altitude, 7. Surface area ratio, 8. Average ratio There are high, 9. terrain steepness index, 10. measured value in wave number domain (http://homepage3.nifty.com/chacocham/Wind_Note/note/note_019.htm). The method for predicting the future wind conditions from the time series data related to the past wind conditions described in the present example is modeled at each point, so the influence is included in the above factors. It is characterized by not being affected. In other words, it is effective in any terrain.

  2 ... Wind power generation equipment 4 ... Wind power generation equipment 5 ... Wind condition measuring instrument 10 ... Time series wind condition / meteorological / power generation data acquisition part 11 ... Time series database 15 ... Wind condition prediction part 20 ... Power generation quantity prediction part 26 ... Display unit 51 ... Observation camera system 52 ... Cloud type, motion measurement system.

Claims (11)

A method for predicting the amount of power generated by a wind power generation facility,
Predict the wind conditions beyond the present using the past wind current time series data and the current wind condition data at the point where the wind power generation facility is installed and the points around the point where the wind power generation facility is installed, A power generation amount prediction method characterized by predicting a power generation amount ahead of the present based on the predicted wind conditions.   The power generation amount prediction method according to claim 1, wherein the wind power generation facility includes a storage battery, and predicts a power generation amount including the storage battery ahead of the present based on the predicted wind condition.   Prediction of wind conditions beyond the present using current wind data and time series data of past wind conditions at a point where the wind power generation facility is installed and points around the site where the wind power generation facility is installed The method for predicting the amount of power generation according to claim 1, wherein this is performed using a Gaussian process.   Results of forecasting wind conditions beyond the present using the past wind current time series data and current wind condition data at the point where the wind power generation facility is installed and the points around the point where the wind power generation facility is installed The power generation amount prediction method according to claim 1, further comprising: displaying a result of predicting a power generation amount ahead of the present time determined based on the predicted wind condition as time series data.   Results of forecasting wind conditions beyond the present using the past wind current time series data and current wind condition data at the point where the wind power generation facility is installed and the points around the point where the wind power generation facility is installed The power generation amount prediction method according to claim 4, wherein time series data is displayed together with information on variance of predicted values. A method for determining the health status of wind power generation equipment,
Measure wind condition data at the location where the wind power generation facility is installed and the amount of power generated by the wind power generation facility, extract feature values of the observation data including the measured wind condition data and power generation amount, and extract the extracted features An abnormality detection or an abnormality sign of the wind power generation facility is detected using the amount and data stored in the database, and a health state of the wind power generation facility is determined based on a result of the abnormality detection or anomaly prediction. A method for health management of wind power generation facilities. A data acquisition unit for acquiring wind conditions, weather information, and power generation amount data at a point where the wind power generation facility is installed and a point around the point where the wind power generation facility is installed;
Time series storing past time series wind conditions, meteorological information, and power generation amount data of the point where the wind power generation facility acquired by the data acquisition unit and the point around the point where the wind power generation facility is installed A database section;
Stored in the time series database unit, the wind condition, weather information, and power generation amount data at the point where the current wind power generation facility acquired by the data acquisition unit and the point around the point where the wind power generation facility is installed are stored. A wind condition prediction unit that predicts a wind condition ahead of the present using
A power generation amount prediction unit that predicts a power generation amount ahead of the current wind power generation facility using information on the wind state predicted by the wind state prediction unit;
An output unit that outputs information on the wind conditions ahead of the current predicted by the wind condition prediction unit and information on the amount of power generation ahead of the current of the wind power generation equipment predicted by the power generation amount prediction unit; A featured power generation prediction system.   The wind power generation facility includes a storage battery, and the power generation amount prediction unit predicts a power generation amount including the storage battery ahead of the present based on information on the wind state prior to the present predicted by the wind state prediction unit. The power generation amount prediction system according to claim 7.   The wind condition predicting unit predicts a wind condition ahead of the present time by using the past time series wind condition information and the current wind condition information stored in the time series database unit. The power generation amount prediction system according to claim 7, wherein:   The output unit uses the past time series wind information and the current wind condition information stored in the time series database unit to predict a wind condition ahead of the present and the predicted wind condition. The power generation amount prediction system according to claim 7, wherein a result of predicting a power generation amount ahead of the present time determined based on is displayed on the screen as time series data. A method for determining the health status of wind power generation equipment,
A wind condition data measurement unit for measuring wind conditions at a point where the wind power generation facility is installed and a point around the point where the wind power generation facility is installed, and obtaining wind condition data;
A power generation amount measuring unit for measuring the power generation amount generated by the wind power generation facility;
An observation data feature amount extraction unit for extracting feature amounts of observation data including the wind condition data measured and acquired by the wind condition data measurement unit and the power generation amount measured by the power generation amount measurement unit;
An anomaly detection unit that performs an anomaly detection or an anomaly sign of the wind power generation facility using the feature amount extracted by the observation data feature amount extraction unit and the data stored in the database;
A health management system for a wind power generation facility, comprising: a health state determination unit that determines a health state of the wind power generation facility based on a result of abnormality detection or an abnormality sign performed by the abnormality detection unit.

Images