CN114943174A - Fan output loss prediction method used under cold tide small sample condition - Google Patents
Fan output loss prediction method used under cold tide small sample condition Download PDFInfo
- Publication number
- CN114943174A CN114943174A CN202210406086.2A CN202210406086A CN114943174A CN 114943174 A CN114943174 A CN 114943174A CN 202210406086 A CN202210406086 A CN 202210406086A CN 114943174 A CN114943174 A CN 114943174A
- Authority
- CN
- China
- Prior art keywords
- data
- cold
- weather
- fan
- fan output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000013507 mapping Methods 0.000 claims description 13
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 238000012512 characterization method Methods 0.000 claims description 7
- 238000005259 measurement Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 101150049349 setA gene Proteins 0.000 claims description 3
- 238000004088 simulation Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 2
- 238000011160 research Methods 0.000 abstract description 7
- 238000004458 analytical method Methods 0.000 abstract description 3
- 230000004044 response Effects 0.000 abstract description 3
- 239000007787 solid Substances 0.000 abstract description 3
- 230000007812 deficiency Effects 0.000 abstract description 2
- 230000007246 mechanism Effects 0.000 abstract description 2
- 230000004907 flux Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 4
- 238000001556 precipitation Methods 0.000 description 4
- 230000005855 radiation Effects 0.000 description 4
- 230000001364 causal effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000003042 antagnostic effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000002045 lasting effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000010792 warming Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/06—Wind turbines or wind farms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Evolutionary Computation (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Algebra (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Air Conditioning Control Device (AREA)
Abstract
The invention provides a method for predicting the output loss of a fan under the condition of a small sample of cold tide. In order to make up the deficiency of the prediction research on the fan output loss under the extreme weather conditions such as the current cold tide, on one hand, the data volume of the extreme weather conditions such as the cold tide is expanded, and a wide and solid data base is provided for the prediction of the fan output loss; on the other hand, a prediction model of the fan output loss is established, and the fan output loss when a cold tide occurs is predicted. The method adopts a meta-learning-confrontation generation network to expand and enhance various required data in the cold tide period, researches the influence mechanism of each meteorological factor on the fan output in the cold tide period and the normal period respectively, establishes two models to predict the fan output value under the extreme weather condition and the non-extreme weather condition respectively, and subtracts the two models to obtain the loss value of the fan output. By the method, the full estimation and analysis of the output of the fan can be provided when cold tide events occur, and a proper and timely response strategy is provided.
Description
Technical Field
The invention belongs to the field of operation and control of power systems, and particularly relates to a method for predicting output loss of a fan under a small sample condition of cold tide.
Background
In the present stage, due to the energy crisis, global problems such as global warming are more prominent, and the energy structure is promoted to be transformed from the traditional energy such as coal to the novel energy such as wind power and photovoltaic. Wind power stands out from numerous new energy sources by virtue of wide industrial foundation, solid technical foundation and abundant resources, and becomes a pillar energy source of a power system. The magnitude of wind power output mainly depends on wind-weather resources of the local environment, abundant wind resources can promote the development of the wind power industry, and the wind power output is more when the wind resources are rich; when the wind resources are deficient, the wind power output is less. Therefore, the output of the fan cannot be manually controlled like traditional energy sources such as thermal power and the like, but has certain volatility, randomness and uncontrollable property, and threats and challenges are brought to the safe and stable operation of a power system.
The China occupies unique geographical position and topographic conditions, is influenced by Siberian high air pressure in winter, and can smoothly enter the inland from north to south to send rich wind energy resources; the east coastal region has abundant wind energy resources due to the geothermal difference effect between sea and land and the addition effect of monsoon climate. In the three north (northwest, northeast and north China), inner Mongolia, southwest and Xinjiang, etc., the wind energy resource is particularly rich. Therefore, wind power in China has great development potential and favorable development natural environment. However, in the above areas, especially in winter, the strong wind and the temperature reduction are often accompanied, and in dry northwest and north China, the strong wind often induces wind dust or sand storm. During cold air activities, the air temperature drops sharply, generally lasting for one to several days. The temperature reduction is more in northwest and north China, and frost is often caused at low temperature below zero, and the pneumatic performance and the mechanical performance of the wind turbine generator are affected due to icing of blades caused by rainfall. Therefore, although the wind energy resources are abundant, due to the influence of the terrain and the meteorological environment, extreme weather conditions such as cold tides and the like occasionally occur, once the extreme weather conditions occur, the great uncertainty of the output of the fan is caused, even the electrical equipment and the secondary measurement element of the unit can be damaged, the risk of large-scale power failure can be increased by forcing the generator to be cut when strong wind exists, the stability and the safety of the power system are further influenced, and the immeasurable loss is caused.
However, the existing researches on the output of the fan under extreme weather conditions such as cold tides are less, so that not only is enough research on the influence of the cold tides on the fan lacked, but also the estimated output of the fan is less when the cold tides are about to occur, so that the problem that the output of the fan is insufficient when the cold tides occur is solved, and a proper and timely response strategy is difficult to provide. Therefore, a method for predicting the fan output loss under the cold weather condition is needed at present.
Disclosure of Invention
In order to make up the deficiency of the prediction research on the output loss of the fan under the extreme weather conditions of cold tide and the like at present, the invention aims to: on one hand, the data volume of extreme weather conditions such as cold tides and the like is expanded, and a wide and solid data base is provided for the prediction of the output loss of the fan; on the other hand, a prediction model of the fan output loss is established, and the fan output loss when a cold tide occurs is predicted. The method adopts a meta-learning-confrontation generation network to expand and enhance various required data in the cold tide period, researches the influence mechanism of each meteorological factor on the fan output in the cold tide period and the normal period respectively, establishes two models to predict the fan output value under the extreme weather condition and the non-extreme weather condition respectively, and subtracts the two models to obtain the loss value of the fan output.
In order to achieve the purpose, the invention adopts the technical scheme that:
s1: constructing a cold tide weather characterization vector for distinguishing the occurrence of a cold tide event, and defining typical weather change parameters and a distinguishing threshold;
s2, constructing a data set containing the NWP weather forecast value and the fan output value of the day before;
s3: according to the influence degree of the NWP meteorological factors on the fan output value, selecting the meteorological factors with the maximum influence degree on the fan output as typical features respectively, and constructing a typical feature set;
s4: expanding the data set under the cold weather condition; carrying out random undersampling treatment on the samples under the normal weather condition;
s5, respectively establishing network mapping according to the extended sample set under the cold weather condition and the sample set under the normal weather condition obtained by random undersampling obtained in the S4; training to obtain a cold tide and normal weather neural network model;
s6, inputting the typical meteorological factor set of cold tide in NWP into the cold tide model in S5, inputting the typical meteorological factor set data of normal weather into the normal weather model in S5, and respectively obtaining predicted values of output of two fans; and subtracting the obtained fan output predicted value under the normal weather condition and the cold weather condition to obtain the fan output loss predicted value under the cold weather condition.
On the basis of the scheme, the NWP meteorological factors are 24, and specifically comprise air temperature, momentum flux, wind speed at 170m height, wind speed at 100m height, wind speed at 30m height, wind speed at 10m height, ground surface wind speed, wind direction at 170m height, wind direction at 100m height, wind direction at 30m height, wind direction at 10m height, ground surface wind direction, sea level air pressure, cloud amount, sensible heat flux, latent heat flux, short wave radiation, long wave radiation, ground surface air pressure, prt, large-scale precipitation, convective precipitation, 2m air temperature and humidity.
On the basis of the above scheme, in S1: based on NWP data of the day before, defining the cold tide weather characterization vector H of one day as shown in a formula (1):
in the formula, t n Is a temperature negative distance, delta t is a steep decrease degree of the gas temperature in one day, R day In order to accumulate the rainfall amount per day,is the daily average wind speed;
temperature negative distance:
the equation represents the difference between the average daily air temperature and the average air temperature over ten years of the day, where,the average air temperature in the day is shown,the average temperature in ten days representing the years of the day;
the steep drop degree of the air temperature in one day:
Δt=t max -t min (3)
in the formula, t max And t min Respectively taking the NWP air temperature maximum value and the NWP air temperature minimum value;
the discrimination threshold is shown in formula (4):
in the formula, x t For the temperature negative distance, x Δt A threshold value for determining the degree of steep drop of air temperature R To determine the threshold, χ, for the total rainfall in a day S Determining a threshold for wind speed;
if the recognition threshold is satisfied, such as formula (4), it is considered that the time period has a cold tide, otherwise, it is a normal weather condition.
On the basis of the above scheme, in S2: counting NWP meteorological data M ═ M of certain wind power plant area (i) ,i=1,2,...,24},M (i) Values representing different meteorological factors; counting actually measured output data { P } of a fan in the wind power plant; the data sets are shown in equations (5) and (6); wherein S n In the normal weatherData set under conditions, S c Is a data set under cold weather conditions;
in the formula, M c And P c Respectively NWP meteorological data and fan output data under the cold weather condition;
M n and P n Respectively NWP meteorological data and fan output data under normal weather conditions.
On the basis of the scheme, the time resolution is 15min, and the NWP meteorological data and the fan output data correspond to each other in a time sequence;
wherein the data set S c In which there are l pieces of data, data set S n With k pieces of data in it.
On the basis of the scheme, in the step S3, typical feature sets are respectively selected from the data sets of the NWP weather forecast value and the fan output value obtained in the step S2, and the typical feature sets are used for respectively representing the significant characteristics of two types of weather conditions;
firstly, constructing the following two regression models, wherein a formula (7) is a constrained regression, and a formula (8) is an unconstrained regression;
in the formula,andrespectively representing values of a fan output value constrained regression model and an unconstrained regression model at the time t, a and b respectively representing regression coefficients of a fan output self variable and a NWP meteorological variable, D being the maximum lag time number of the fan output, S being the maximum lag time number of the NWP factor,and delta t Is white noise;
calculating the residual square of the two regression models according to formula (9), and calculating F statistic according to formula (10);
respectively screening according to the F statistics to obtain meteorological factors with the confidence coefficient of more than 99 percent as the typical feature sets of the two weather conditions, wherein the typical feature set of normal weather is T n ={T (i) ,i=1,2,...,I n };
Equations (11) and (12) represent typical weather feature sets in normal weather conditions and in cold weather conditions.
On the basis of the above scheme, in S4: for the data set under the condition of normal weather, a random undersampling mode is adopted, and the data set is processed by S n Randomly selecting NWP meteorological data and fan actual measurement output data at part of time to obtain a new sample setAs shown in equation (13);
in the formula,representing the undersampled sample set, wherein K pieces of data are contained in the sample set, and K is known<k;
Data set S for cold weather conditions c Generating neural network pairs S using the countermeasures c Expanding the data set to obtain an expanded sample setThe anti-neural network is used for learning the data distribution characteristics under the cold weather condition, and more similar samples are generated in a simulation mode on the data distribution characteristics to form a new sample set together with the original sample set, as shown in a formula (14);
in the formula,representing the extended sample set, wherein the sample set has L pieces of data, namely L>l。
On the basis of the above scheme, in S5: respectively learning the mapping relation between the typical meteorological feature set and the actual measured force value of the fan under the cold weather condition and the normal weather condition by utilizing an extreme learning machine algorithm;
equation (15) represents a mapping between typical meteorological conditions and fan output during cold tides,
equation (16) represents a mapping between typical weather conditions and fan output under normal weather conditions, f 1 And f 2 Representing an extreme learning machine algorithm;
on the basis of the above scheme, in S6: from NWP meteorological data giving cold weatherInputting their typical feature sets under two weather conditions into the two models trained in the S5, as shown in equation (17):
in the formula,andtypical feature set data for NWP in cold weather and normal weather for the period to be predicted,the predicted value of the fan output under the normal weather condition,is a predicted value of fan output under cold weather conditions, P loss Is windAnd (4) predicting the machine output loss.
On the basis of the scheme, the data volume of the extended cold weather sample is equal to the data volume of the undersampled normal weather sample, namely, L is equal to K.
The prior art bases and data sources used by the present invention are:
1. the 24-dimensional NWP data of one day before the day is used and comprises 24 meteorological factors, specifically air temperature, momentum flux, wind speed at 170m height, wind speed at 100m height, wind speed at 30m height, wind speed at 10m height, ground surface wind speed, wind direction at 170m height, wind direction at 100m height, wind direction at 30m height, wind direction at 10m height, ground surface wind direction, sea level air pressure, cloud amount, sensible heat flux, latent heat flux, short wave radiation, long wave radiation, ground surface air pressure, prt, large-scale precipitation, convective precipitation, 2m air temperature and humidity.
2. And (4) actually measuring output data of the fan under the influence of cold tide.
3. The granger causal relationship verification method, a tool of mathematical statistics, was used.
4. Using a deep learning neural network algorithm: antagonistic neural networks, extreme learning machines.
The invention has the beneficial effects that:
due to the fact that the quantity of extreme weathers such as cold tides is small, and the data volume is insufficient, not only is enough research on the influence of the cold tides on the fan lacked, but also the estimated output of the fan is small when the cold tides are about to occur. The prediction method can provide a sufficient prediction analysis for the output of the fan when dealing with the cold tide event and provide a proper and timely response strategy.
Drawings
The invention has the following drawings:
FIG. 1 is a flow model for predicting output loss of a fan under a small sample condition of cold tide.
Detailed Description
The present invention is described in further detail below with reference to fig. 1.
A method for predicting output loss of a fan under a small sample condition of cold tide comprises the following steps:
s1, based on the definition and standard of the cold tide event, a weather characterization vector for distinguishing the cold tide event is constructed from the weather characteristics of the cold tide event, and typical weather change parameters and a distinguishing threshold value are defined to provide a distinguishing method for distinguishing the cold tide event.
S2, collecting operation information of the wind power plant and weather forecast information of the station, constructing a data set containing NWP weather forecast values and fan output values of the day before, and dividing the data set into two types according to the method in S1. One is a weather forecast and fan output data set under normal weather conditions, and the other is a weather forecast and fan output data set under extreme weather conditions such as cold tides and the like.
S3, analyzing meteorological factors influencing the output of the fan under two weather conditions, quantifying the influence degree of the 24 NWP meteorological factors on the output of the fan under the two weather conditions by using a mathematical statistics method, respectively selecting the meteorological factors influencing the output of the fan to be the maximum as typical characteristics, and constructing a typical characteristic set.
And S4, expanding the data set under the cold weather condition. And simulating the distribution conditions of similar meteorological data and fan output data under the cold weather condition by utilizing the generation countermeasure network so as to generate more cold weather data and corresponding cold fan output data. And carrying out random undersampling treatment on the samples under the normal weather conditions. This achieves a consistent number of samples of both types.
And S5, respectively establishing network mapping according to the extended sample set under the cold weather condition and the undersampled sample set under the normal weather condition obtained in the S4. Under the condition of cold tide, inputting a NWP typical meteorological factor set of the cold tide, and outputting a fan output value of the cold tide; under the condition of normal weather, the NWP typical meteorological factor set in normal weather is input, and the fan output value in the normal period is output. Two neural network models are obtained through training.
S6, predicting the occurrence of the cold tide event according to the method provided by S1, inputting the typical meteorological factor set of the cold tide in the NWP into the cold tide model in S5, inputting the typical meteorological factor set data of the normal weather into the normal weather model in S5, and respectively obtaining two predicted values of the fan output. And subtracting the obtained predicted value of the output of the fan under the normal weather condition and the cold weather condition to obtain the predicted value of the output loss of the fan under the cold weather condition.
On the basis of the above technical solution, in S1, criteria for analyzing the definition of cold tide events are studied, and the intensity and intensity of cold air activity are studied. And (3) excavating the characteristics of NWP meteorological data in the cold tide period, extracting and constructing a characterization vector capable of obviously characterizing the occurrence of the cold tide weather condition, and defining characterization parameters. Typical meteorological features of cold tides are large temperature dips, with strong winds, even snowstorms, low temperatures and heavy rainfall. And defining a cold tide weather characterization vector H of one day as shown in a formula (1-3) on the basis of NWP data of one day before the day.
Wherein,
temperature negative distance:
the expression represents the difference between the daily average air temperature and the average air temperature over the years of the day, wherein,the average air temperature in the day is shown,the average air temperature in each year of the year in which the day is located.
② the steep drop degree of the air temperature in one day:
Δt=t max -t min (3)
in the formula, t max And t min NWP air temperature maximum and air temperature minimum, respectively.
③R day Accumulating rainfall for each day;
Under the condition of cold weather, a certain temperature negative distance can be generated, the air temperature in one day is rapidly reduced, the phenomena of strong wind, strong rainfall and the like are often accompanied in partial areas of China, the discrimination threshold value of the above characteristic parameters is given at present for identifying the cold weather event, and the threshold value condition is shown as a formula (4)
In the formula, x t For the temperature negative distance, x Δt A threshold value for determining the degree of steep drop of air temperature R To determine the threshold, χ, for the total rainfall in a day S A threshold is determined for the wind speed. If the above judgment conditions are satisfied, it can be considered that the cold tide occurs in the period, otherwise, it is a normal weather condition.
Based on the above technical solution, in S2, NWP meteorological data M ═ M in a certain wind farm area is counted (i) ,i=1,2,...,24},M (i) Values representing different meteorological factors; and (4) counting the actually measured output data { P } of the fan in the wind power plant. The data sets are shown in equations (5) and (6). Wherein S n Is a data set under normal weather conditions, S c Is a data set in cold weather conditions.
In the formula, M c And P c Respectively NWP meteorological data and fan output data under the cold and tide weather conditions. M n And P n Respectively, normal weatherUnder the condition, the time resolution of the NWP meteorological data and the fan output data is 15min, 96 pieces of data are counted in one day, and the NWP meteorological data and the fan output data correspond to each other in a time sequence. Wherein the data set S c Among which is l pieces of data, a data set S n There are k pieces of data, and the value of l is generally much smaller than the value of k.
On the basis of the technical scheme, in S3, a typical feature set is selected from the meteorological factor sets in the two sample sets obtained in S2, so as to represent the significant features of the two types of weather conditions respectively. And verifying the causal relationship between the meteorological factors and the fan output under the two weather conditions by using a grand causal relationship verification method, and calculating to obtain the relevance between the 24 meteorological factors and the fan output. The calculation method is as follows, firstly, the following two regression models are constructed, wherein formula (7) is a constrained regression, and formula (8) is an unconstrained regression.
In the formula,andrespectively representing values of a fan output value constrained regression model and an unconstrained regression model at the time t, a and b respectively representing regression coefficients of a fan output self variable and a NWP meteorological variable, D being the maximum lag time number of the fan output, S being the maximum lag time number of the NWP factor,and delta t Is white noise.
The residual squares of the two regression models were calculated according to equation (9) and the F statistic was calculated according to equation (10).
Respectively screening according to the F statistic to obtain a meteorological factor with a confidence coefficient of more than 99% as a typical feature set of two weather conditions, wherein the typical feature set of normal weather is T n ={T (i) ,i=1,2,...,I n }; typical feature set of cold weather is T c ={T (i) ,i=1,2,...,I c },I n <24,I c < 24. The typical feature set under two weather conditions basically represents all information contained in 24 meteorological factors, and has the maximum reference contribution to the fan output value.
Equations (11) and (12) represent typical weather feature sets under normal weather conditions and cold weather conditions, and weather factors in the two typical feature sets are not exactly the same.
Based on the above technical solution, in S4, to solve the data distribution situation of the unbalanced data amount in the two types of data sets, the following measures are taken for the two types of data sets respectively. For the data set under the condition of normal weather, because the data volume is large, a random undersampling mode is adopted, and the data set is divided into S n NWP meteorological data and wind turbine real-time measurement value at part of time selected randomlyForce data, obtaining a new sample setAs shown in equation (13).
In the formula,representing the undersampled sample set, wherein K pieces of data are contained in the sample set, and K is known<k。
Data set S for cold weather conditions c And because the data volume is small, the requirements of later analysis and modeling are difficult to meet, and the difference between the data volume and the data volume under the normal weather condition is large. Thus, the neural network pair S is generated using the pair c Expanding the data set to obtain an expanded sample setThe anti-neural network is used for learning data distribution characteristics under cold weather conditions, and more similar samples are generated on the data distribution characteristics in a simulation mode to form a new sample set together with the original sample set, as shown in a formula (14).
In the formula,representing the extended sample set, wherein the sample set has L pieces of data, namely L>l。
The sample quantities of the two types of sample sets obtained in the way are consistent, namely K is L, so that the balance of sample data under the cold weather condition and the normal weather condition is ensured, the requirement of the data quantity of modeling in the subsequent steps is met, and the reasonability and the effectiveness of the sample sets are ensured.
In the above technical schemeIn step S5, an extreme learning machine algorithm is used to learn the mapping relationship between the typical weather feature set and the actual measured force value of the wind turbine under cold weather conditions and normal weather conditions. Equation (15) represents a mapping between typical meteorological conditions and fan output during cold tides. Equation (16) represents a mapping between typical meteorological conditions and fan output under normal weather conditions. f. of 1 And f 2 An extreme learning machine algorithm is represented.
Based on the above technical solution, in S6, NWP meteorological data for providing cold weather is usedInputting the typical feature set under the two weather conditions into the two models trained in S5 respectively, as shown in formula (17)
In the formula,andrespectively, typical feature set data of NWP in cold weather and normal weather for a period to be predicted,the predicted value of the output of the fan under the normal weather condition,is a predicted value of fan output under cold weather conditions, P loss And (5) predicting the output loss of the fan.
Can be regarded asThe output of the fan is reduced to the value required by the fan under normal weather conditions due to the influence of cold weatherThe loss prediction value of the present invention is therefore calculated based on the typical feature set of different weather conditions to derive this difference.
The above embodiments are only for illustrating the present invention and are not meant to be limiting, and those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, so that all equivalent technical solutions also belong to the scope of the present invention, and the scope of the present invention should be defined by the claims.
Those not described in detail in this specification are within the skill of the art.
Claims (10)
1. A method for predicting the output loss of a fan under the condition of a small sample in cold tide is characterized in that,
the method comprises the following steps:
s1: constructing a cold tide weather representation vector for distinguishing the occurrence of a cold tide event, and defining a typical weather change parameter and a distinguishing threshold;
s2, constructing a data set containing the NWP weather forecast value and the fan output value of the day before;
s3: according to the influence degree of the NWP meteorological factors on the fan output value, selecting the meteorological factors with the maximum influence degree on the fan output as typical features respectively, and constructing a typical feature set;
s4: expanding the data set under the cold weather condition; carrying out random undersampling treatment on samples under normal weather conditions;
s5, respectively establishing network mapping according to the extended sample set under the cold weather condition and the sample set under the normal weather condition obtained by random undersampling obtained in the S4; training to obtain a cold tide and normal weather neural network model;
s6, inputting the typical meteorological factor set of cold tide in NWP into the cold tide model in S5, inputting the typical meteorological factor set data of normal weather into the normal weather model in S5, and respectively obtaining predicted values of output of two fans; and subtracting the obtained fan output predicted value under the normal weather condition and the cold weather condition to obtain the fan output loss predicted value under the cold weather condition.
2. The method as claimed in claim 1, wherein the number of NWP meteorological factors is 24.
3. The method for predicting the fan output loss in the case of the small sample of the cold tide according to claim 2, wherein in the step S1: defining the weather characterization vector H of the cold tide in one day as shown in a formula (1) on the basis of NWP data of one day before the day:
in the formula, t n Is a temperature negative distance, delta t is a steep decrease degree of the gas temperature in one day, R day In order to accumulate the rainfall amount per day,is the daily average wind speed;
temperature negative distance:
the formula represents the average daily temperature and years of the dayThe difference between the average gas temperature and the average temperature in ten days, wherein,the average air temperature in the day is shown,the average temperature in ten days representing the years of the day;
the steep drop degree of the air temperature in one day:
Δt=t max -t min (3)
in the formula, t max And t min Respectively taking the NWP air temperature maximum value and the NWP air temperature minimum value;
the discrimination threshold is shown in formula (4):
in the formula, x t For the temperature negative distance judgment threshold value, χ Δt A threshold value for determining the degree of steep drop of air temperature R To determine the threshold, χ, for the total rainfall in a day S Determining a threshold for wind speed;
if the recognition threshold is met, such as equation (4), it is considered that a cold tide has occurred in the period, otherwise it is a normal weather condition.
4. The method for predicting the fan output loss in the case of the small sample of the cold tide according to claim 2, wherein in the step S2: calculating NWP meteorological data M ═ M of a certain wind power plant area (i) ,i=1,2,...,24},M (i) Values representing different meteorological factors; counting actually measured output data { P } of a fan in the wind power plant; the data sets are shown in equations (5) and (6); wherein S n Is a data set under normal weather conditions, S c Is a data set under cold weather conditions;
in the formula, M c And P c Respectively NWP meteorological data and fan output data under the cold weather condition;
M n and P n Respectively NWP meteorological data and fan output data under normal weather conditions.
5. The method for predicting the fan output loss under the condition of the small sample with the cold tide as claimed in claim 4, wherein the time resolution is 15min, and the NWP meteorological data and the fan output data correspond to each other in a time sequence;
wherein the data set S c Among which is l pieces of data, a data set S n There are k pieces of data.
6. The method as claimed in claim 2, wherein in S3, a typical feature set is selected from the NWP weather forecast value and the fan output value data set obtained in S2 for representing the significant features of two types of weather conditions;
firstly, constructing the following two regression models, wherein a formula (7) is a constrained regression, and a formula (8) is an unconstrained regression;
in the formula,andrespectively representing values of a fan output value constrained regression model and an unconstrained regression model at a time t, a and b respectively representing regression coefficients of a fan output self variable and a NWP meteorological variable, D being the maximum lag time number of the fan output, S being the maximum lag time number of a NWP factor,and delta t Is white noise;
calculating the residual squares of the two regression models according to formula (9), and calculating the F statistic according to formula (10);
respectively screening according to the F statistic to obtain a meteorological factor with a confidence coefficient of more than 99% as a typical feature set of two weather conditions, wherein the typical feature set of normal weather is T n ={T (i) ,i=1,2,...,I n };
Equations (11) and (12) represent typical weather feature sets in normal weather conditions and in cold weather conditions.
7. The method for predicting the fan output loss in the case of the small sample of the cold tide according to claim 2, wherein in the step S4: for the data set under the condition of normal weather, a random undersampling mode is adopted, and S is measured n Randomly selecting NWP meteorological data and fan actual measurement output data at part of time to obtain a new sample setAs shown in equation (13);
in the formula,the sample set after undersampling is represented, the sample set has K pieces of data, and K can be known<k;
Data set S for cold weather conditions c Generation of neural network pairs S using the pairs c Expanding the data set to obtain an expanded sample setThe anti-neural network is used for learning the data distribution characteristics under the cold weather condition, and more similar samples are generated in a simulation mode on the data distribution characteristics to form a new sample set together with the original sample set, as shown in a formula (14);
8. The method for predicting the fan output loss in the case of the small sample of the cold tide according to claim 2, wherein in the step S5: respectively learning the mapping relation between the typical meteorological feature set and the actual measured force value of the fan under the cold weather condition and the normal weather condition by utilizing an extreme learning machine algorithm;
equation (15) represents a mapping between typical meteorological conditions and fan output during cold tides,
equation (16) represents a mapping between typical meteorological conditions and fan output under normal weather conditions, f 1 And f 2 Representing an extreme learning machine algorithm;
9. the method for predicting fan output loss in the case of a small sample of cold tide as set forth in claim 1, wherein in S6: by NWP meteorological data giving out cold weatherInputting the typical feature set under two weather conditions into two models trained in the mapping model established in the S5, as shown in equation (17):
in the formula,andtypical feature set data for NWP in cold weather and normal weather for the period to be predicted,the predicted value of the fan output under the normal weather condition,is a predicted value of fan output under cold weather conditions, P loss And (5) predicting the output loss of the fan.
10. The method for predicting the fan output loss under the condition of the small sample in the cold tide as claimed in claim 7, wherein the data volume of the sample in the cold tide after expansion is equal to the data volume of the sample in the normal weather after undersampling, namely L is K.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210406086.2A CN114943174A (en) | 2022-04-18 | 2022-04-18 | Fan output loss prediction method used under cold tide small sample condition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210406086.2A CN114943174A (en) | 2022-04-18 | 2022-04-18 | Fan output loss prediction method used under cold tide small sample condition |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114943174A true CN114943174A (en) | 2022-08-26 |
Family
ID=82907229
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210406086.2A Pending CN114943174A (en) | 2022-04-18 | 2022-04-18 | Fan output loss prediction method used under cold tide small sample condition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114943174A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117631090A (en) * | 2024-01-25 | 2024-03-01 | 南京信息工程大学 | Cold tide identification method and device |
-
2022
- 2022-04-18 CN CN202210406086.2A patent/CN114943174A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117631090A (en) * | 2024-01-25 | 2024-03-01 | 南京信息工程大学 | Cold tide identification method and device |
CN117631090B (en) * | 2024-01-25 | 2024-05-14 | 南京信息工程大学 | Cold tide identification method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Dabbaghiyan et al. | Evaluation of wind energy potential in province of Bushehr, Iran | |
CN102663251B (en) | Physical prediction method for wind power station power based on computational fluid mechanics model | |
Tabrizi et al. | Performance and safety of rooftop wind turbines: Use of CFD to gain insight into inflow conditions | |
CN105492762B (en) | According to the method for its component life of wind turbine or the location determination of similar devices | |
Li et al. | Onshore and offshore wind energy potential assessment near Lake Erie shoreline: A spatial and temporal analysis | |
BoroumandJazi et al. | Technical characteristic analysis of wind energy conversion systems for sustainable development | |
CN104299044A (en) | Clustering-analysis-based wind power short-term prediction system and prediction method | |
CN106611243A (en) | Residual correction method for wind speed prediction based on GARCH (Generalized ARCH) model | |
CN110991701A (en) | Wind power plant fan wind speed prediction method and system based on data fusion | |
CN105825002A (en) | Method for modeling dynamic equivalence of wind power farm based on dynamic grey-relevancy analysis method | |
CN108062722B (en) | Mechanical power calculation method of mountain wind power plant model fan based on wind speed variation coefficient | |
CN114943174A (en) | Fan output loss prediction method used under cold tide small sample condition | |
Özkan et al. | Evaluation of wind loads and the potential of Turkey's south west region by using log-normal and gamma distributions | |
Cheggaga et al. | A neural network solution for extrapolation of wind speeds at heights ranging for improving the estimation of wind producible | |
CN115204709B (en) | Typhoon risk assessment method convenient for site selection of wind power plant | |
CN115204712B (en) | Offshore and coastal wind power plant site selection evaluation method | |
Corscadden et al. | The impact of variable wind shear coefficients on risk reduction of wind energy projects | |
Chaudhry et al. | Computational analysis to factor wind into the design of an architectural environment | |
Rodriguez et al. | CFD Design of Urban Wind Turbines: A Review and Critical Analysis | |
Khan et al. | Wind energy potential estimation for different regions of Bangladesh | |
Li et al. | Wind velocity prediction at wind turbine hub height based on CFD model | |
Dadioti | Numerical wind resource assessment in urban environments | |
Siddique et al. | Techno-economic study for 50 MW wind farm in Ormara, Balochistan | |
Sikkeland | Power prediction and wake losses in offshore wind farms with the dynamic wake meandering model | |
Abo-Elkhair et al. | New Performance Modeling Techniques for Photovoltaic Modules and Different Types of Wind Turbines |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |