TWI662423B - Display system and method for wind power prediction - Google Patents

Abstract

A wind power generation prediction method includes: capturing a plurality of input data from a plurality of servers, the input data including at least one real-time wind speed data and at least one historical wind speed data; and using the input data to mold at least one weight of a type of neural network model And at least one partial weight value; according to the weight and partial weight value, the input data is calculated by a neural network-like model to obtain at least one wind speed prediction data; according to a correction value, the wind speed prediction data is calculated to obtain At least one wind speed prediction correction data is calculated; the wind speed prediction correction data is calculated according to a predicted effective range value to obtain at least one predicted effective range data; the wind speed prediction correction data is calculated to obtain at least one wind power prediction correction Data; and calculating the predicted effective range data to obtain at least one predicted effective range data of wind power generation.

Description

Wind power generation prediction display system and method

The invention relates to a wind power forecast display system and method, and more particularly to a wind power forecast display system and method that can perform prediction value correction and prediction effective range labeling.

Wind power is an intermittent energy source. If the power generated by wind power is incorporated into the grid system, the stability of the original power system will be reduced. At the same time, the unknown supply of wind power will increase the difficulty of power dispatch and related operating costs. The main cause of wind turbine output instability is that when the wind speed changes, the wind turbine output also changes. Due to the very complex and highly non-linear relationship between the effects of terrain, temperature, air pressure, latitude and other factors on wind speed, it will cause considerable difficulties when using traditional physical models or statistical methods to predict wind speed. When using a linear or non-linear time series model to predict wind speed in the early days, only the historical data was used to make wind speed predictions at multiple leading time points. The prediction accuracy was poor, and it was not easy to combine weather data to perform multiple leading time points. .

Compared to this, neural network-like artificial intelligence prediction can learn the relationship between input and output, without the need to provide mathematical functions for conversion, and can complete complex non-linear mapping, so it is more suitable for wind speed prediction. However, when using neural networks to make wind speed predictions, Reliable correction method can correct the prediction result. At the same time, neural-like networks also lack a method that can calculate the effective range of the prediction to indicate the effective range of the prediction result.

Therefore, in order to improve the accuracy of forecasting wind power generation, a wind power forecast display method capable of correcting the prediction result is required. At the same time, there is also a need for a wind power forecast display method that can indicate the effective range of the forecast results.

In order to solve the above problems, an object of the present invention is to provide a wind power generation prediction display system and method.

Another object of the present invention is to provide a wind power generation prediction display system and method capable of correcting prediction results.

Still another object of the present invention is to provide a wind power generation prediction display system and method that can indicate the effective range of prediction results.

According to the above object, the present invention provides a wind power forecast display system, including: a prediction module including: an input data storage module, which retrieves and stores a plurality of input data from a plurality of servers, and the input data includes at least one Real-time wind speed data and at least one wind speed historical data; a type of neural network learning module that models at least one weight and at least one partial weight based on the input data; a prediction data calculation module based on the weight and partial weight, Input data for calculation to obtain at least one wind speed prediction data; a prediction data correction module for calculating wind speed prediction data according to a correction value to obtain at least one wind speed prediction correction data; and a wind power calculation module , Calculating input data to obtain real-time wind power data and historical data of wind power, calculating wind speed prediction data to obtain wind power prediction data, The wind speed prediction correction data is calculated to obtain a wind power prediction correction data; the wind power prediction display system further includes: a database that receives and stores input data and / or wind speed prediction data and / or wind speed prediction correction data And / or real-time data of wind power and / or historical data of wind power and / or forecast data of wind power and / or forecast correction data of wind power; and a user operation interface, receiving at least one user operation condition, and according to the user operation Condition to display input data and / or wind speed forecast data and / or wind speed forecast correction data and / or real-time wind power data and / or wind power historical data and / or wind power forecast data and / or wind power forecast correction data.

In an embodiment that achieves the above object, the input data includes at least one real-time data of wind speed weather forecast and at least one historical data of wind speed weather forecast.

In the embodiment for achieving the above purpose, the correction value is obtained by calculating the input data and the wind speed prediction data by using a root mean square error calculation formula or an average absolute error calculation formula.

In an embodiment that achieves the above purpose, the prediction module further includes: a prediction effective range calculation module that calculates wind speed prediction correction data according to a predicted effective range value to obtain at least one predicted effective range data.

In an embodiment that achieves the above purpose, the wind power generation computing module calculates the predicted effective range data to obtain the wind power predicted effective range data.

In the embodiment that achieves the above purpose, the predicted effective range value is obtained by calculating input data and / or wind speed prediction data and / or wind speed prediction correction data by using a standard deviation calculation formula.

In the embodiment that achieves the above purpose, the predicted effective range data includes a predicted effective range upper threshold value data and a predicted effective range lower threshold value data, and the wind power generation forecast The measured effective range data includes a critical value data of a wind power forecast effective range and a critical value data of a wind power forecast valid range.

In the embodiment that achieves the above purpose, when the critical value data under the predicted effective range is a value less than zero, the critical value data under the predicted effective range is modified to a value of zero, and when the critical value data under the predicted effective range of the wind power generation is less than When the value is zero, the critical value data in the effective range of the wind power forecast is modified to a value of zero.

In the embodiment that achieves the above purpose, the database receives and stores the predicted effective range data and the wind power predicted effective range data, and the user operation interface displays the input data and / or wind speed prediction data and / or according to the user operating conditions. Wind speed forecast correction data and / or forecast effective range data and / or wind power real-time data and / or wind power historical data and / or wind power forecast data and / or wind power forecast correction data and / or wind power forecast valid range data.

In an embodiment that achieves the above purpose, the display input data and / or wind speed prediction data and / or wind speed prediction correction data and / or prediction effective range data and / or real-time wind power data and / or historical wind power data and / The wind power forecast data and / or wind power forecast correction data and / or wind power forecast effective range data are displayed in a graphical manner.

In an embodiment that achieves the above purpose, wherein the user operation condition includes an indication of the displayed wind field, the user operation interface displays input data and / or wind speed prediction data associated with the indicated wind field according to the user operation condition And / or wind speed forecast correction data and / or real-time wind power data and / or wind power historical data and / or wind power forecast data and / or wind power forecast correction data.

In the embodiment that achieves the above purpose, the indication of the displayed wind field is an indication of a plurality of wind fields, and the user operation interface displays the input data associated with the indicated plurality of wind fields and / Or wind speed forecast data and / or wind speed forecast correction data and / or the sum of wind power real-time data and / or the sum of wind power historical data and / or the sum of wind power forecast data and / or wind power The sum of predicted correction data.

According to the purpose of the present invention, a wind power forecast display system is further provided, including: a prediction module including: an input data storage module that retrieves and stores a plurality of input data from a plurality of servers, and the input data includes at least A wind speed real-time data and at least one wind speed historical data; a type of neural network learning module, based on the input data, molds at least one weight and at least one partial weight; a prediction data calculation module, based on the weight and partial weight, Calculate input data to obtain at least one wind speed prediction data; a prediction effective range calculation module to calculate wind speed prediction data according to a predicted effective range value to obtain at least one predicted effective range data; and a wind force The power generation computing module calculates input data to obtain real-time wind power data and historical wind power data, calculates wind speed prediction data to obtain wind power prediction data, and calculates valid range data. To obtain a wind power forecast effective range data; the wind power forecast display system and Including: a database that receives and stores input data and / or wind speed forecast data and / or forecast effective range data and / or real-time wind power data and / or wind power historical data and / or wind power forecast data and / or wind power Predicting effective range data; and a user operation interface, receiving at least one user operating condition, and displaying input data and / or wind speed prediction data and / or predicted effective range data and / or wind power generation in real time according to the user operating condition Data and / or historical data on wind power and / or forecast data on wind power and / Or wind power forecast valid range data.

In an embodiment that achieves the above object, the input data includes at least one real-time data of wind speed weather forecast and at least one historical data of wind speed weather forecast.

In the embodiment that achieves the above purpose, the predicted effective range value is obtained by calculating input data and wind speed prediction data by using a standard deviation calculation formula.

In an embodiment that achieves the above purpose, the predicted effective range data includes a predicted effective range upper threshold data and a predicted effective range lower threshold data, and the wind power predicted effective range data includes a wind power predicted effective range upper threshold data. And a critical value data of the effective range of wind power forecast.

In the embodiment that achieves the above purpose, when the critical value data under the predicted effective range is a value less than zero, the critical value data under the predicted effective range is modified to a value of zero, and when the critical value data under the predicted effective range of the wind power generation is less than zero The value of the critical value data in the valid range of the wind power forecast is modified to a value of zero.

In an embodiment that achieves the above purpose, the display input data and / or wind speed prediction data and / or predicted effective range data and / or real-time wind power data and / or historical wind power data and / or wind power prediction data and / Or the effective range data of wind power forecast are displayed graphically.

In an embodiment that achieves the above purpose, wherein the user operation condition includes an indication of the displayed wind field, the user operation interface displays input data and / or wind speed prediction data associated with the indicated wind field according to the user operation condition. And / or forecast effective range data and / or real-time data of wind power and / or historical data of wind power and / or forecast data of wind power and / or wind power Electric forecast effective range data.

According to the purpose of the present invention, the indication of the displayed wind field is an indication of displaying a plurality of wind fields, and the user operation interface displays the input data associated with the indicated plurality of wind fields and / or according to the operating conditions of the user. Wind speed forecast data and / or forecast effective range data and / or wind power real-time data added value and / or wind power historical data added value and / or wind power forecast data added value and / or wind power forecast effective The sum of the range data.

According to the purpose of the present invention, a wind power generation prediction method is further provided. The method includes: extracting a plurality of input data from a plurality of servers, and the input data includes at least one real-time wind speed data and at least one historical wind speed data; At least one weight and at least one partial weight of the neural network model; according to the weight and partial weight, the input data is calculated by the neural network-like model to obtain at least one wind speed prediction data; according to a correction value, the Calculate the wind speed prediction data to obtain at least one wind speed prediction correction data; operate on the wind speed prediction correction data according to a predicted effective range value to obtain at least one prediction effective range data; perform calculation on the wind speed prediction correction data, To obtain at least one wind power forecast correction data; and to perform calculation on the predicted effective range data to obtain at least one wind power forecast effective range data.

In an embodiment that achieves the above object, the input data includes at least one real-time data of wind speed weather forecast and at least one historical data of wind speed weather forecast.

In the embodiment for achieving the above purpose, the correction value is obtained by calculating the input data and the wind speed prediction data by using a root mean square error calculation formula or an average absolute error calculation formula.

In an embodiment that achieves the above purpose, the predicted effective range value is calculated by using a standard deviation calculation formula on input data and / or wind speed prediction data and / or wind speed prediction correction data. And get.

In an embodiment that achieves the above purpose, the predicted effective range data includes a predicted effective range upper threshold data and a predicted effective range lower threshold data, and the wind power predicted effective range data includes a wind power predicted effective range upper threshold data. And a critical value data of the effective range of wind power forecast.

In the embodiment that achieves the above purpose, when the critical value data under the predicted effective range is a value less than zero, the critical value data under the predicted effective range is modified to a value of zero, and when the critical value data under the predicted effective range of the wind power generation is less than zero The value of the critical value in the valid range of the wind power forecast is modified to a value of zero.

The foregoing aspects and other aspects of the present invention will become more apparent from the detailed description of the following non-limiting specific embodiments and with reference to the accompanying drawings.

100‧‧‧ class neural network system

210‧‧‧ training materials

220‧‧‧Data Server

230‧‧‧wind power forecast display system

250‧‧‧ Prediction Module

252‧‧‧Input data storage module

254‧‧‧ class neural network learning module

256‧‧‧ Forecast Data Operation Module

257‧‧‧ Forecast data correction module

258‧‧‧ Prediction valid range calculation module

259‧‧‧wind power computing module

260‧‧‧user interface

262‧‧‧operation module

264‧‧‧ Forecast data display module

270‧‧‧Database

280‧‧‧users

300‧‧‧flow chart

310-360‧‧‧step

400‧‧‧wind speed prediction system

401‧‧‧current wind speed

402‧‧‧The first hour wind speed

403‧‧‧First 2 hours wind speed

Wind speed weather forecast after 404‧‧‧1 hour

Wind speed weather forecast after 405‧‧‧2 hours

Weather forecast for 406‧‧‧3 hours later

410‧‧‧ class neural network

422‧‧‧1 hour forecast wind speed

424‧‧‧2 hours forecast wind speed

426‧‧‧3 hours forecast wind speed

428‧‧‧48 hours forecast wind speed

429‧‧‧corrected value

430‧‧‧ Calibration Module

Predicted corrected wind speed after 442‧‧‧1 hour

Predicted corrected wind speed after 444‧‧‧2 hours

446‧‧‧3 hours after forecasting corrected wind speed

Predicted corrected wind speed after 448‧‧‧48 hours

449‧‧‧Valid range value

450‧‧‧ Prediction Effective Range Calculation Module

Effective range of forecast wind speed after 462‧‧‧1 hour

Effective range of predicted wind speed after 464‧‧‧2 hours

Effective range of predicted wind speed after 466‧‧‧3 hours

Effective range of predicted wind speed after 468‧‧‧48 hours

500‧‧‧wind speed prediction system

520‧‧‧Wind-speed power generation conversion curve-like neural network model

540‧‧‧ class neural network

542‧‧‧wind speed

544‧‧‧Wind power generation

610‧‧‧Vertical axis

620‧‧‧horizontal axis

631‧‧‧ historical forecast correction data

632‧‧‧Historical observation data of wind power generation

633‧‧‧ Upper threshold data of historical forecast effective range of wind power generation

634‧‧‧Critical data of the lower limit of historical forecast effective range of wind power generation

635‧‧‧ Wind power forecast correction data

636‧‧‧ Upper threshold data of the effective range of wind power forecast

637‧‧‧Critical data of the lower limit of wind power forecast effective range

638‧‧‧ Forecast data of wind power generation

The first figure is a schematic diagram of a general neural network system.

The second figure is a system architecture diagram of a specific embodiment of a wind power forecast display system according to the present invention.

The third figure is a flowchart of a specific embodiment of the wind power prediction method of the present invention.

FIG. 4A is a system model diagram of a specific embodiment of wind speed prediction according to the present invention.

FIG. 4B is a system model diagram of another specific embodiment of wind speed prediction according to the present invention.

The fifth figure is a system model diagram of a specific embodiment for constructing a wind speed power generation conversion curve.

The sixth figure is a preview of a specific embodiment of the wind power forecast display system of the present invention. Schematic representation of test results.

The wind power generation prediction display system and method of the present invention utilize the characteristics of a neural network that can learn the relationship between input data and output data without providing a mathematical function for conversion. Neural network-like training to model the weights of each input node and each hidden layer neuron, the weights of each hidden layer neuron and output node, and the partial weights of each node, and then cooperate with real-time retrieved wind speed data As input data of each input node, wind speed prediction is performed, and based on the obtained predicted wind speed, a predicted wind power generation amount is calculated through a wind speed power generation conversion curve. The following will be further explained with the illustration.

The first figure is a schematic diagram of a general neural network system. The wind power generation prediction display system of the present invention is trained by using a neural network system 100 such as shown in the figure to model each weight and each partial weight. In the figure, the neural network-like system 100 is in the first place. The layer has an input node S1, an input node S2, a hidden layer neuron S3, a hidden layer neuron S4, a hidden layer neuron S5 in the second layer, and an output node S6 in the third layer.

It should be understood that the neural network system 100 is only an example here. The present invention is not limited to using a single layer of hidden layer neurons, but may use one to multiple layers of hidden layer neurons as required. The number of input nodes, the number of hidden layer neurons, and the number of output nodes of the neural network system 100 are not intended to limit the present invention. The number of input nodes, the number of hidden layer neurons, and the number of output nodes of the present invention It can be adjusted to any number as required.

Please continue to refer to the first figure. Each input node of the neural network system 100 has a corresponding weight to each hidden layer neuron. For example, the input node S1 and the hidden layer neuron S5 have a weight W15, and the input node S2 and the hidden layer The layer neuron S5 has a weight W25. At the same time, each hidden layer neuron of the neural network system 100 and the output node also have corresponding weights. For example, the hidden layer neuron S3 and the output node S6 have a weight W36, and the hidden layer neuron S5 and the output node S6 have a weight. Weight W56. In addition, each input node, hidden layer neuron, and output node have their own partial weights. For example, the input node S2 has a partial weight θ2, the hidden layer neuron S5 has a partial weight θ5, and the output node S6 has a partial weight θ6. . The calculation method of the transmission value transmitted from one node to the next node is as follows. Assuming that a total of n nodes transmit their respective transmission values to a node Y, the formula for the transmission value y from node Y to the next node is: Among them, Wi is the weight corresponding to the i-th node and node Y among the n nodes of node Y, Xi is the transmission value of the i-th node to node Y, and θ is the bias of node Y. Weight.

Taking node S5 as an example, the weight corresponding to input node S1 and hidden layer neuron S5 is W15, the weight corresponding to input node S2 and hidden layer neuron S5 is W25, and the partial weight of hidden layer neuron S5 is θ5 . Let the transmission value of input node S1 to hidden layer neuron S5 be X1, and the transmission value of input node S2 to hidden layer neuron S5 be X2, then the transmission value of hidden layer neuron S5 to output node S6 is: (W15.X1 + W25.X2) -θ5

In the process of training the neural network system 100, a large amount of known input data and output data are first used as training input data and output data to train the neural network system 100, thereby modeling each weight And each partial weight. Traditional gradient The descent algorithm is used to correct the individual weights. After modeling the weights and partial weights, the neural network system 100 can be used to predict the output data with real-time input data.

The second figure is a system architecture diagram of a specific embodiment of the wind power forecast display system according to the present invention. As shown, the wind power forecast display system 230 includes a forecast module 250; a user operation interface 260; and a database 270 . The prediction module 250 includes an input data storage module 252 for capturing and storing input data to a plurality of servers, and the input data includes at least real-time wind speed data and historical wind speed data; a type of neural network learning module 254, according to The input data stored in the input data storage module 252 is used to mold at least one weight and at least one partial weight; a predictive data operation module 256 is used to model the weights and Partial weights are calculated on the input data to obtain prediction data, where the prediction data includes at least wind speed prediction data; a prediction data correction module 257 calculates wind speed prediction data based on a correction value to obtain a wind speed prediction Correction data; and a prediction effective range calculation module 258, which calculates wind speed prediction data and / or wind speed prediction correction data according to a predicted effective range value to obtain predicted effective range data. The weights, partial weights, wind speed prediction data, wind speed prediction correction data, and effective range data are all stored in the database 270. The user operation interface 260 includes an operation module 262 for receiving user operation conditions, and the user operation conditions include prediction data and / or historical data and / or valid range data selected by the user 280 for browsing; and a prediction data display Module 264, according to user operating conditions, go to the database 270 to retrieve the forecast data and / or historical data and / or effective range data of the selected browse, and display the forecast data and / or historical data of the selected browse and / or Valid range data for user 280 to view.

Among them, the prediction module 250 is a 210-pair neural network with a large amount of training data. The learning module 254 performs training, thereby modeling each weight and each partial weight. Then, the input data storage module 252 periodically retrieves and stores various types of required real-time data to one or more data servers 220, and then the predicted data calculation module 256 stores the data stored in the input data storage module 252. As input data, the weights and partial weights modeled by the neural network learning module 254 are used to further calculate the prediction data, and the prediction data is stored through the database 270. The user can determine the range of the predicted data and / or historical data and / or predicted effective range data to be displayed through the operation module 262. Then, the user operation interface 260 will go to the database 270 according to the predicted data range selected by the user. Retrieve the prediction data and / or historical data and / or prediction effective range data, and present the prediction data and / or historical data to the user through the prediction data display module 264.

The third figure is a flowchart of a specific embodiment of the wind power prediction method of the present invention. The flowchart 300 of the wind speed prediction method in the figure includes the following steps: First, step 310 is performed to retrieve input data from a plurality of servers. In this embodiment, the input data retrieved at step 310 includes real-time wind speed data, historical wind speed data, real-time wind speed weather forecast data, and historical wind speed weather forecast data.

After step 310 is completed, step 320 is performed to model the weights and partial weights of the neural network-like model based on the input data. In this embodiment, the input nodes of the neural network are the current wind speed, the first hour wind speed, the first two hours wind speed, the wind speed weather forecast after one hour, the wind speed weather forecast after two hours, and the wind speed weather forecast after three hours. forecast. The number of hidden neurons is 5 to 20. The output node of the neural network is the predicted wind speed after 1 to 48 hours. Next, step 330 is performed to perform prediction operation according to the neural network model based on the modeled weights and partial weights. Gives 1 to 48 Wind speed forecast data in hours.

It should be understood that the number of input nodes, the number of hidden layer neurons, and the number of output nodes described herein are not intended to limit the present invention. The number of input nodes, the number of hidden layer neurons, and the number of output nodes of the present invention It can be adjusted to any number as required. At the same time, the present invention is not only used to predict the wind speed after 1 to 48 hours, but it can be used to predict wind speed for at least several hours to several days and weeks according to demand. In addition, the time interval for forecasting is not limited to only one hour, but the time interval for forecasting can be set to make a forecast every few minutes or hours or days according to demand.

After step 330 is completed, step 340 is performed, a correction value is calculated, and the prediction result is further corrected by the correction value. In a specific embodiment, the correction value is calculated by using a root mean square error (RMSE) calculation formula. The formula of the root mean square error expression is: Among them, f (i) is the observed wind speed, y (i) is the predicted wind speed, and N is the number of samples taken. In this specific embodiment, the root-mean-square error calculation formula is used to calculate the predicted wind speed values from 1 hour from February 2016 to July 2016 and the observed wind speed values at that time, thereby obtaining The root mean square error value of the predicted wind speed after 1 hour is used as the correction value of the wind speed prediction data after 1 hour. Then the wind speed prediction value after 1 hour is added to the correction value of wind speed prediction data after 1 hour, which is the wind speed prediction correction value after 1 hour. In the above manner, the respective correction values of the wind speed prediction data after 1 hour to the wind speed prediction data after 48 hours can be calculated, and the wind speed prediction correction data after 1 to 48 hours can be further calculated.

In another specific embodiment, the correction value is calculated by a mean absolute error (MAE: Mean Absolute Error) calculation formula. The formula of the average absolute error expression is: Among them, f (i) is the observed wind speed, y (i) is the predicted wind speed, and N is the number of samples taken. In this specific embodiment, the average absolute error calculation formula is used to calculate the wind speed prediction value and the wind speed observation value measured at that time for all one hour from February 2016 to July 2016, thereby obtaining The average absolute error value of the predicted wind speed after 1 hour is used as the correction value of the predicted wind speed data after 1 hour. Then the wind speed prediction value after 1 hour is added to the correction value of wind speed prediction data after 1 hour, which is the wind speed prediction correction value after 1 hour. In the above manner, the respective correction values of the wind speed prediction data after 1 hour to the wind speed prediction data after 48 hours can be calculated, and the wind speed prediction correction data after 1 to 48 hours can be further calculated.

After step 340 is completed, step 350 is performed to calculate the effective range value, and the predicted effective range is calculated based on the effective range value. The predicted effective range includes the upper critical value of the predicted effective range and the lower critical value of the predicted effective range. In a specific embodiment, the effective range value is calculated by a standard deviation (σ: Standard Deviation) calculation formula. The formula for the standard deviation expression is: Among them, e (i) is the prediction error value, which is the value obtained by subtracting the wind speed observation value from the wind speed observation value, N is the number of samples taken, μ is the average value of the prediction error value, and the formula of μ is: Among them, e (i) is the prediction error value, that is, the value obtained by subtracting the wind speed observation value from the wind speed observation value, and N is the number of samples taken. In this specific embodiment, the standard deviation calculation formula is used for all wind speed prediction data from February 2016 to July 2016 that have the same wind speed prediction value after one hour and the wind speed observation values measured at that time. Calculation to obtain the standard deviation of the predicted wind speed after 1 hour, and use this as the valid range value of the wind speed prediction data after 1 hour. For example, for all the wind speed forecast data with a wind speed forecast value of A from February 2016 to July 2016 and the measured wind speed observation values at that time, you can get the wind speed forecast value A after 1 hour. The standard deviation of the wind speed forecast data is used as the valid range value of the wind speed forecast data after one hour when the wind speed forecast value is A after one hour. By calculating the wind speed forecast data of wind speed forecast value B and the wind speed observation value measured at that time from February 2016 to July 2016, you can get the wind speed forecast value B after 1 hour. The standard deviation of the wind speed prediction data is used as the valid range value of the wind speed prediction data after one hour when the wind speed prediction value after one hour is B. Then, after calculating the effective range value of the wind speed prediction data after one hour, the predicted wind speed value after one hour is added to the effective range value of the wind speed prediction data after one hour, which is the predicted wind speed after one hour. The critical value of the predicted effective range of. The predicted value of the wind speed after one hour is subtracted from the effective range value of the wind speed prediction data after one hour by two times, which is the lower critical value of the predicted effective range of the predicted wind speed after one hour. It should be noted that when the lower critical value of the predicted effective range is less than zero, the lower critical value of the predicted effective range less than zero is modified to zero. With the above method, the effective range values of the wind speed prediction data after 1 hour to the wind speed prediction data after 48 hours can be calculated, and the upper critical values of the respective prediction effective ranges of the wind speed prediction data after 1 to 48 hours can be further calculated. Data and critical value data for the forecast effective range.

In another specific embodiment, the prediction is performed on the wind speed prediction correction data. Effective range calculation. In this specific embodiment, the standard deviation calculation formula is used for all wind speed prediction correction data from February 2016 to July 2016 that have the same wind speed prediction correction value after one hour and the wind speed observations measured at that time. The value is calculated to obtain the standard deviation of the predicted and corrected wind speed after one hour, and this is used as the effective range value of the predicted and corrected wind speed after one hour. For example, for all the wind speed forecast correction data of C and the wind speed observation correction value measured at that time from February 2016 to July 2016, one can obtain the wind speed forecast correction value after one hour. It is the standard deviation of the wind speed prediction correction data of C, and this is taken as the effective range value of the wind speed prediction correction data of 1 hour after the wind speed prediction correction value of C. And for all wind speed forecast correction data of D from 1st February 2016 to July 2016, the wind speed prediction correction data D and the wind speed observation value measured at that time are calculated to obtain the wind speed prediction correction value after 1 hour. The standard deviation of the wind speed prediction correction data for D is used as the effective range value of the wind speed prediction correction data for D after 1 hour. Then, after calculating the effective range value of the wind speed prediction correction data after one hour, the wind speed prediction correction value after one hour is added to twice the effective range value of the wind speed prediction correction data after one hour, which is one hour later. The critical value of the forecast effective range of the forecast corrected wind speed. The predicted value of the wind speed correction after 1 hour is subtracted from the effective range value of the wind speed prediction correction data after 1 hour, which is the lower critical value of the predicted effective range of the prediction correction wind speed after 1 hour. It should be noted that when the lower critical value of the predicted effective range is less than zero, the lower critical value of the predicted effective range less than zero is modified to zero. In this way, the effective range values of the wind speed prediction correction data after 1 hour to the wind speed prediction correction data after 48 hours can be calculated, and the respective prediction effective ranges of the wind speed prediction correction data after 1 to 48 hours can be calculated. Upper critical value data and lower critical value data of the forecast effective range.

After step 350 is completed, step 360 is performed, and the wind power generation prediction result is calculated by using the wind speed power generation conversion curve, wherein the wind speed power generation conversion curve representing the relationship between the wind speed and the power generation amount is provided by the wind turbine manufacturer. At the same time, when the wind speed power generation conversion curve of the fan is inaccurate due to long-term use of the fan, the neural speed-like system can be used to mold the wind speed power generation conversion curve. And through the wind speed power generation conversion curve of the fan, the wind speed prediction data from 1 to 48 hours can be used to calculate the wind power prediction data from 1 to 48 hours, and the wind speed prediction correction data from 1 to 48 hours can be used to calculate 1 to 48 hours. For the subsequent wind power forecast correction data, the upper critical value data of the wind power forecast effective range after 1 to 48 hours is calculated based on the forecast effective range upper threshold data of 1 to 48 hours, and the forecast valid range after 1 to 48 hours is calculated. The lower critical value data calculates the lower critical value data of the valid range of wind power generation prediction after 1 to 48 hours. It should be noted that when the critical value of the effective range of the wind power forecast is less than zero, the critical value of the effective range of the wind power forecast less than zero is modified to zero.

FIG. 4A is a system model diagram of a specific embodiment of wind speed prediction according to the present invention. The wind speed prediction system 400 in the figure is a wind speed prediction after 1 to 48 hours by a trained neural network 410. First, the current wind speed 401, the wind speed 402 in the first hour, the wind speed 403 in the first two hours, the wind speed weather forecast 404 in one hour, the wind speed weather forecast 405 in two hours, and the wind speed weather forecast 406 in three hours are used as input data. A neural network 410 is used to calculate the predicted wind speed 422 after 1 hour, the predicted wind speed 424 after 2 hours, the predicted wind speed 426 after 3 hours, and the predicted wind speed 428 after 48 hours. Then, the correction module 430 calculates a correction value 429 through a root mean square error calculation formula or an average absolute error calculation formula, and corrects the predicted wind speed from 1 to 48 hours by using the correction value 429 to obtain a prediction after 1 hour. Corrected wind speed 442, predicted after 2 hours Corrected wind speed 444, predicted after 3 hours The corrected wind speed is 446, and the predicted wind speed is 448 after 48 hours. Then, the prediction effective range calculation module 450 calculates the effective range value 449 through the standard deviation calculation formula, and uses the effective range value 449 to calculate the predicted and corrected wind speed after 1 to 48 hours to obtain the effective range of the predicted wind speed after 1 hour. 462, the effective range of the predicted wind speed after 2 hours is 464, the effective range of the predicted wind speed after 3 hours is 466, and the effective range of the predicted wind speed after 48 hours is 468.

FIG. 4B is a system model diagram of another specific embodiment of wind speed prediction according to the present invention. The wind speed prediction system 500 in the figure is a wind speed prediction after 1 to 48 hours by a trained neural network 410 or the like. First, the current wind speed 401, the wind speed 402 in the first hour, the wind speed 403 in the first two hours, the wind speed weather forecast 404 in one hour, the wind speed weather forecast 405 in two hours, and the wind speed weather forecast 406 in three hours are used as input data. A neural network 410 is used to calculate the predicted wind speed 422 after 1 hour, the predicted wind speed 424 after 2 hours, the predicted wind speed 426 after 3 hours, and the predicted wind speed 428 after 48 hours. Then, the effective range calculation module 450 calculates the effective range value 449 through the standard deviation calculation formula, and uses the effective range value 449 to calculate the predicted wind speed after 1 to 48 hours to obtain the effective wind range 462 after one hour. The effective range of wind speed is 464 after 2 hours, the effective range of wind speed is 466 after 3 hours, and the effective range of wind speed is 468 after 48 hours.

The fifth figure is a system model diagram of a specific embodiment for constructing a wind speed power generation conversion curve. In this embodiment, the wind speed power generation curve-like neural network model 520 uses wind speed 542 as the input node, the number of hidden layer neurons of the neural network 540 is 5, and the output node is the corresponding wind speed of the input wind speed. Power generation 544. With the corresponding power generation amount under different wind speeds, the wind speed power generation conversion curve can be modeled by the neural network model 520 of the wind speed power generation conversion curve.

The sixth figure is a preview of a specific embodiment of the wind power forecast display system of the present invention. Schematic representation of test results. In this embodiment, the user requested to display the prediction result at 23:30 on August 14, 2016, and the range of prediction data selected by the user is A, B, and C. The wind field has been on August 14, 2016 14 : From 00, wind power forecast correction data after 1 to 48 hours and wind power forecast correction data from 1 to 48 hours. After the selection is completed, the prediction result is displayed in the form of a line chart. In the figure, the vertical axis 610 indicates the amount of wind power, and the horizontal axis 620 indicates time. The displayed data includes wind power historical forecast correction data 631, wind power historical observation data 632, wind power historical forecast effective range upper threshold data 633, wind power historical forecast effective range lower threshold data 634, wind power forecast correction data 635, The upper critical value data 636 of the effective range of the wind power forecast, the lower critical value data 637 of the effective range of the wind power forecast, and the observation data 638 of the wind power forecast.

Among them, the wind power historical forecast correction data 631 are the wind power forecast correction data for A wind farm at 1 hour and 14:00 on August 14 at 13:00, and the wind power forecast for B wind farm at 1 hour. The sum of the correction data and the correction data of the wind power forecast for 1 hour after the C wind field. The historical observation data of wind power generation 632 are the observation data of A wind farm wind power, B wind farm wind power observation data, and C wind farm wind power observation data at 15:00 on August 12 and 14:00 on August 14. Add value. The upper threshold data 633 of the effective range of the historical forecast of wind power generation is the upper threshold value data of the effective range of the historical forecast of wind power at 14:00 on August 12 to 13:00 on August 14, and the wind B The combined value of the critical value data of the valid range of the historical forecast of wind power generation after one hour of the field and the valid range of the historical range of the wind power forecast of the C wind field after one hour. The lower threshold data 634 of the effective range of the historical forecast of wind power generation is the lower critical value data of the effective range of the historical forecast of wind power generation at 14:00 on August 12th to 13:00 on August 14th. field Addition of the critical value data of the effective range of the historical forecast of wind power generation after 1 hour and the critical value data of the effective range of the historical forecast of wind power generation after 1 hour of the C wind farm.

In addition, the wind power forecast correction data 635 is the phase of A wind farm wind power forecast correction data, B wind farm wind power forecast correction data, and C wind farm wind power forecast correction data starting at 14:00 on August 14. Value added, up to the sum of A wind farm wind power forecast correction data, B wind farm wind power forecast correction data, and C wind farm wind power forecast correction data 48 hours later. The upper threshold value data of the effective range of wind power forecasting starts at 14:00 on August 14th, and after 1 hour, the upper threshold value data of the effective range of wind power forecast of A wind farm, the upper threshold value data of effective range of wind power forecasting and The sum of the critical value data in the effective range of the C wind farm wind power forecast, 48 hours later, the critical value data of the A wind farm wind power forecast effective range, the B critical data of the wind farm wind power forecast effective range, and the C wind Addition of critical value data in the effective range of wind farm forecast. Data 637 for the lower critical value of the effective range of the wind power forecast is from 14:00 on August 14, 1 hour later. The data for the lower critical value of the effective range of the wind power forecast for A wind farm, the data for the lower critical value of the effective range of the wind power forecast for wind farm and The sum of the critical value data in the effective range of the wind farm wind power forecast, the critical value data in the effective range of the wind farm A wind farm forecast, the critical value data in the valid range of the wind farm wind power forecast, and the C wind 48 hours later. Addition of critical value data in the valid range of wind power generation forecast. The wind power forecast observation data 638 are the data of the wind power observation data of A wind field, the wind power observation data of B wind field, and the wind power observation data of C wind field at each time point from 14:00 on August 14 to 23:00 on August 14. Add value.

It should be understood that although the wind power historical forecast correction data 631 here is the wind power forecast correction data for the first hour of each time point, the present invention is not limited to using each time Wind power forecast correction data for the first hour of the point, and if necessary, the wind power forecast correction data for the first few hours or days, weeks, and months of the time point can be used. In addition, users are not only able to choose to display the forecast data of multiple wind farms, but they can choose to display the forecast data of a single wind farm according to their needs. The wind power forecast display system of the present invention also displays forecast data associated with the wind farm selected by the user. For example, when the wind field selected and displayed by the user is only the D wind field, the historical wind power observation values at each time point displayed will be the historical wind power observation data of the D wind field at the corresponding time point. The displayed wind power forecast correction value at each time point will be the D wind farm wind power forecast correction data at the corresponding time point.

So far, the wind power generation prediction display system and method of the present invention have been described by the above description and drawings. It should be understood, however, that the specific embodiments of the present invention are for illustration purposes only, and various changes can be made without departing from the scope and spirit of the patent application of the present invention, which should be included in the patent scope of the present invention. Therefore, the specific embodiments described in this specification are not intended to limit the present invention. The true scope and spirit of the present invention are disclosed in the following patent applications.

Claims (23)

A wind power generation prediction display system includes: a prediction module including: an input data storage module that retrieves and stores a plurality of input data from a plurality of servers, and the input data includes at least one real-time wind speed data and at least one wind speed Historical data; a type of neural network learning module that models at least one weight and at least one partial weight based on the input data; a predictive data calculation module that performs input data based on the weight and the partial weight Calculation to obtain at least one wind speed prediction data; a prediction data correction module to calculate the wind speed prediction data according to a correction value to obtain at least one wind speed prediction correction data; a prediction effective range calculation module according to A predicted effective range value, calculating the wind speed prediction correction data to obtain at least one predicted effective range data; and a wind power calculation module, calculating the input data to obtain a wind power real-time data and a The historical data of wind power generation, the wind speed prediction data is calculated to obtain a wind power prediction data, and the The speed prediction correction data is calculated to obtain a wind power prediction correction data; a database receives and stores the input data and / or the wind speed prediction data and / or the wind speed prediction correction data and / or the wind power real-time data And / or the wind power historical data and / or the wind power forecast data and / or the wind power forecast correction data; and a user operation interface, receiving at least one user operation condition, and according to the user operation condition, to Display the input data and / or the wind speed prediction data and / or the wind speed prediction correction data and / or the wind power real-time data and / or the wind power historical data and / or the wind power prediction data and / or the wind power prediction Correction data; wherein the operation of the wind speed prediction data to obtain the wind power prediction data, and the operation of the wind speed prediction correction data to obtain the wind power prediction correction data, using a wind speed Power conversion curve to perform; wherein the wind speed power conversion curve is based on a wind speed power conversion curve like a neural network It is obtained by modeling; wherein the model of the neural network model of the wind speed power generation conversion curve model is independent of the model of the neural network learning module; wherein the predicted effective range data includes a predicted effective range upper threshold data And a critical value of the predicted effective range. The wind power forecast display system according to item 1 of the scope of patent application, wherein the input data includes at least one real-time wind speed weather forecast data and at least one historical wind speed weather forecast data. The wind power forecast display system according to item 1 of the scope of the patent application, wherein the correction value is obtained by calculating the input data and the wind speed prediction data by using a root mean square error calculation formula or an average absolute error calculation formula. The wind power forecast display system according to item 1 of the scope of the patent application, wherein the wind power computing module calculates the forecast effective range data to obtain the wind power forecast valid range data. The wind power forecast display system according to item 1 of the scope of the patent application, wherein the valid value range of the forecast is performed on the input data and / or the wind speed forecast data and / or the wind speed forecast correction data by using a standard deviation calculation formula. Calculated. The wind power forecast display system according to item 4 of the scope of patent application, wherein the wind power forecast effective range data includes a wind power forecast effective range upper threshold data and a wind power forecast effective range lower threshold data. The wind power forecast display system according to item 6 of the scope of patent application, wherein when the critical value data under the predicted effective range is a value less than zero, the critical value data under the predicted effective range is modified to a value of zero; when the wind power When the critical value data in the predicted effective range is less than zero, the critical value data in the predicted effective range of the wind power generation is modified to a value of zero. The wind power forecast display system according to item 4 of the scope of patent application, wherein the database receives and stores the forecast effective range data and the wind power forecast effective range data, and the user operation interface is based on the user operating conditions, To display the input data and / or the wind speed prediction data and / or the wind speed prediction correction data and / or the forecast effective range data and / or the wind power real-time data and / or the wind power historical data and / or the wind power The forecast data and / or the wind power forecast correction data and / or the wind power forecast valid range data. The wind power forecast display system according to item 8 of the scope of the patent application, wherein the display displays the input data and / or the wind speed forecast data and / or the wind speed forecast correction data and / or the forecast effective range data and / or the The real-time wind power data and / or the wind power historical data and / or the wind power forecast data and / or the wind power forecast correction data and / or the wind power forecast effective range data are displayed graphically. According to the wind power forecast display system described in item 1 of the scope of patent application, wherein the user operation condition includes an indication of the displayed wind field, the user operation interface displays the wind energy indicated by the user according to the user operation condition. The field-associated input data and / or the wind speed prediction data and / or the wind speed prediction correction data and / or the wind power real-time data and / or the wind power historical data and / or the wind power prediction data and / or the Wind power forecast correction data. According to the wind power forecast display system described in item 10 of the scope of patent application, wherein the displayed wind field indication is an indication of a plurality of wind fields, and the user operation interface displays and instructs according to the user operation conditions. The input data and / or the wind speed prediction data and / or the wind speed prediction correction data and / or the wind power real-time data and / or the wind power historical data associated with the plurality of wind farms Value and / or the added value of the wind power forecast data and / or the added value of the wind power forecast correction data. A wind power generation prediction display system includes: a prediction module including: an input data storage module that retrieves and stores a plurality of input data from a plurality of servers, and the input data includes at least one real-time wind speed data and at least one wind speed Historical data; a type of neural network learning module that models at least one weight and at least one partial weight based on the input data; a predictive data calculation module that performs input data based on the weight and the partial weight Calculation to obtain at least one wind speed prediction data; a prediction effective range calculation module to calculate the wind speed prediction data according to a predicted effective range value to obtain at least one predicted effective range data; and a wind power calculation model Group, calculating the input data to obtain a real-time wind power data and a historical wind power data, calculating the wind speed prediction data to obtain a wind power prediction data, and calculating the prediction effective range data, To obtain a valid range of wind power forecast data; a database to receive and store the input data and / or The wind speed forecast data and / or the forecast effective range data and / or the wind power real-time data and / or the wind power historical data and / or the wind power forecast data and / or the wind power forecast valid range data; and a use User operation interface, receiving at least one user operation condition, and displaying the input data and / or the wind speed prediction data and / or the prediction effective range data and / or the wind power real-time data and / or according to the user operation condition Or the wind power historical data and / or the wind power prediction data and / or the wind power prediction effective range data; wherein the operation of the wind speed prediction data to obtain a wind power prediction data is to use a wind speed The power generation conversion curve is performed; wherein the wind speed power generation conversion curve is modeled according to a neural network model of the wind speed power generation conversion curve model; wherein the wind speed power generation conversion curve neural network model is modeled independently of the neural network. Model of the road learning module; wherein the prediction effective range data includes a prediction upper range critical value data and a prediction Data valid threshold range; the effective range of the wind power forecast wind power forecast data includes a threshold value of the valid range data and a lower threshold of wind power forecast data valid range. The wind power forecast display system according to item 12 of the scope of the patent application, wherein the input data includes at least one real-time wind speed weather forecast data and at least one historical wind speed weather forecast data. The wind power forecast display system according to item 12 of the scope of the patent application, wherein the forecast effective range value is obtained by calculating the input data and the wind speed forecast data through a standard deviation calculation formula. The wind power forecast display system according to item 12 of the scope of patent application, wherein when the critical value data under the forecast effective range is a value less than zero, the critical value data under the forecast valid range is modified to a value of zero; when the wind power When the critical value data in the predicted effective range is less than zero, the critical value data in the predicted effective range of the wind power generation is modified to a value of zero. The wind power forecast display system according to item 12 of the scope of the patent application, wherein the display displays the input data and / or the wind speed forecast data and / or the forecast effective range data and / or the wind power real-time data and / or the Historical wind power data and / or the wind power forecast data and / or the wind power forecast effective range data are displayed in a graphical manner. According to the wind power forecast display system described in item 12 of the scope of patent application, wherein the user operation condition includes an indication of the displayed wind field, the user operation interface displays and indicates the wind power according to the user operation condition. The field-related input data and / or the wind speed prediction data and / or the prediction effective range data and / or the wind power real-time data and / or the wind power historical data and / or the wind power prediction data and / or the Data on the effective range of wind power forecasts. According to the wind power forecast display system described in item 17 of the scope of patent application, wherein the displayed wind field indication is an indication of a plurality of wind fields, and the user operation interface displays and instructs according to the user operation conditions. The input data and / or the wind speed prediction data and / or the predicted effective range data and / or the wind power real-time data and / or the wind power historical data associated with the plurality of wind farms Value and / or the added value of the wind power forecast data and / or the added value of the wind power forecast valid range data. A wind power generation prediction method includes: capturing a plurality of input data from a plurality of servers, the input data including at least one real-time wind speed data and at least one wind speed history data; and using the input data to mold at least one type of neural network model. A weight and at least one partial weight; according to the weight and the partial weight, the input data is calculated by the neural network model to obtain at least one wind speed prediction data; according to a correction value, the wind speed Calculating the prediction data to obtain at least one wind speed prediction correction data; calculating the wind speed prediction correction data according to a predicted effective range value to obtain at least one prediction effective range data; calculating the wind speed prediction correction data, To obtain at least one wind power forecast correction data; and to perform calculations on the forecast effective range data to obtain at least one wind power forecast correction range data; wherein the wind speed prediction correction data is calculated to obtain at least A wind power forecast correction data is made by using a wind speed power conversion curve; The rapid power generation conversion curve is modeled according to a neural network model of a wind speed power conversion curve; the molding system of the neural network model of the wind speed power conversion curve is independent of the neural network model; The predicted effective range data includes a predicted effective range upper critical value data and a predicted effective range lower critical value data; the wind power predicted effective range data includes a wind power predicted effective range upper threshold value and a wind power predicted effective range below Threshold data. The method for predicting wind power generation as described in item 19 of the scope of patent application, wherein the input data includes at least one real-time wind speed weather forecast data and at least one historical wind speed weather forecast data. The method for predicting wind power generation according to item 19 of the scope of the patent application, wherein the correction value is obtained by calculating the input data and the wind speed prediction data by using a root mean square error calculation formula or an average absolute error calculation formula. The method for predicting wind power generation according to item 19 of the scope of patent application, wherein the effective value range of the prediction is calculated by using a standard deviation calculation formula on the input data and / or the wind speed prediction data and / or the wind speed prediction correction data. And get. The method for predicting wind power generation as described in item 19 of the scope of patent application, wherein when the critical value data under the effective range of the prediction is a value less than zero, the critical value data under the effective range of the prediction is modified to a value of zero; When the critical value data in the effective range is less than zero, the critical value data in the effective range of the wind power forecast is modified to a value of zero.