CN116090216A - Power generation power prediction method and system based on typical correlation analysis - Google Patents

Power generation power prediction method and system based on typical correlation analysis Download PDF

Info

Publication number
CN116090216A
CN116090216A CN202310027000.XA CN202310027000A CN116090216A CN 116090216 A CN116090216 A CN 116090216A CN 202310027000 A CN202310027000 A CN 202310027000A CN 116090216 A CN116090216 A CN 116090216A
Authority
CN
China
Prior art keywords
historical
weather
predicted
power
typical
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
Application number
CN202310027000.XA
Other languages
Chinese (zh)
Inventor
李宝聚
赖晓文
付小标
温亚坤
侯嘉琪
孙勇
郭雷
王志伟
王尧
庄冠群
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Tsintergy Technology Co ltd
State Grid Jilin Electric Power Corp
Sichuan Energy Internet Research Institute EIRI Tsinghua University
Original Assignee
Beijing Tsintergy Technology Co ltd
State Grid Jilin Electric Power Corp
Sichuan Energy Internet Research Institute EIRI Tsinghua University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Tsintergy Technology Co ltd, State Grid Jilin Electric Power Corp, Sichuan Energy Internet Research Institute EIRI Tsinghua University filed Critical Beijing Tsintergy Technology Co ltd
Priority to CN202310027000.XA priority Critical patent/CN116090216A/en
Publication of CN116090216A publication Critical patent/CN116090216A/en
Pending legal-status Critical Current

Links

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
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01W—METEOROLOGY
    • G01W1/00—Meteorology
    • G01W1/10—Devices for predicting weather conditions
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/10—Numerical modelling
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/10—Power transmission or distribution systems management focussing at grid-level, e.g. load flow analysis, node profile computation, meshed network optimisation, active network management or spinning reserve management
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]

Abstract

The invention provides a method and a system for predicting generated power based on typical correlation analysis, comprising the following steps: acquiring historical predicted weather and historical actual weather; training to obtain a typical correlation analysis model based on the historical predicted weather and the historical actual weather; acquiring historical typical prediction weather and historical actual measurement power; training to obtain a power prediction model based on the historical typical predicted meteorological and the historical actual measured power; acquiring a predicted weather of a period to be predicted; inputting the predicted weather into the typical correlation analysis model, and outputting the typical predicted weather by the model; inputting the typical predicted meteorological into the power prediction model, and outputting predicted power by the model; according to the invention, the model migration problem is solved by introducing the typical correlation analysis into the power prediction, the dependence of the model on the prediction weather accuracy is reduced, and the accuracy and the robustness of the power prediction are improved.

Description

Power generation power prediction method and system based on typical correlation analysis
Technical Field
The invention relates to the technical field of electric power, in particular to a method and a system for predicting generated power based on typical correlation analysis.
Background
The rapid increase of the grid-connected capacity of wind power and photovoltaic power generation in the power system enables the connection between new energy power generation and the system to be more and more intimate, so that the influence of fluctuation and intermittent generation of new energy on the power quality, safe and stable operation and economic benefit of the power system is considered, and the new energy power prediction has important practical significance. In the prior art, a wind power prediction model is established through wind farm history actual measurement weather and output power data, and a result of Numerical Weather Prediction (NWP) is used as a weather source input to predict wind farm output, so that the wind farm output prediction method is one of the current common wind power prediction methods. However, in such cases, there is a problem in model migration, i.e., actual weather is used for training, and predicted weather is used for real-time prediction. The distribution differences between the predicted and measured weather may lead to reduced performance of the model in real-time prediction. Another common prediction method is to directly construct the corresponding relationship between the predicted weather and the input and output of the actually measured power, and the situation has no model migration problem theoretically, but the algorithm depends on the accuracy of the predicted weather to a great extent. The biased predicted weather can lead to false correspondence learned by the model, and the robustness of the model is reduced.
In view of the above, the invention provides a power generation power prediction method and system based on typical correlation analysis, which introduces the typical correlation analysis into power prediction, fully excavates the correlation between the predicted weather and the actual weather, uses the excavated information to power prediction, solves the problem of model migration in the power prediction, reduces the dependence of the power prediction model on the accuracy of the predicted weather, and improves the accuracy and the robustness of the power prediction.
Disclosure of Invention
The invention aims to provide a generated power prediction method based on typical correlation analysis, which comprises the following steps: acquiring historical predicted weather and historical actual weather; training to obtain a typical correlation analysis model based on the historical predicted weather and the historical actual weather; acquiring historical typical prediction weather and historical actual measurement power; training to obtain a power prediction model based on the historical typical predicted meteorological and the historical actual measured power; acquiring a predicted weather of a period to be predicted; inputting the predicted weather into the typical correlation analysis model, and outputting the typical predicted weather by the model; and inputting the typical predicted meteorological into the power prediction model, and outputting predicted power by the model.
Further, for wind power generation, the historical predicted meteorological, the historical actual meteorological, and the predicted meteorological include wind speed, wind direction, and temperature at a wind hub.
Further, for photovoltaic power generation, the historical predicted weather and the predicted weather include a total irradiance and temperature; the historical actual weather includes short wave radiation, long wave radiation, and temperature.
Further, the method further comprises the following steps: dividing the historical predicted weather, the historical actual weather and the historical actual power according to a time sequence to obtain a training set, a verification set and a test set; the training set is used for training the typical correlation analysis model and the power prediction model; the validation set is used for adjusting super parameters in the power prediction model; the test set is used for testing the typical correlation analysis model and the power prediction model; normalizing the historical predicted weather, the historical actual weather and the historical measured power in the training set, the verification set and the test set respectively; wherein, the normalized expression is:
Figure BDA0004045496730000021
Figure BDA0004045496730000031
wherein, when i epsilon 1, 2 and i=1, X 1 Representing a historical actual meteorological matrix, X when i=2 2 The method comprises the steps of representing a historical prediction meteorological matrix, wherein the number of rows of the matrix is the number of samples, and the number of columns is the number of meteorological indexes; y represents a history actual measurement power matrix (vector), the number of lines is the number of samples, the number of columns is 1,
Figure BDA0004045496730000032
representing the historical actual meteorological matrix or the historical forecast meteorological matrix after normalization processing,
Figure BDA0004045496730000033
x represents i Column minimum of>
Figure BDA0004045496730000034
X represents i Is the column maximum value of (2); />
Figure BDA0004045496730000035
Represents the history measured power after normalization processing, Y min Represents the column minimum value of Y max Represents the column maximum value of Y.
Further, the training obtains a typical correlation analysis model, which comprises the following steps: inputting the historical predicted weather and the historical actual weather into an initial typical correlation analysis model; the initial typical correlation model carries out linear combination on the historical predicted weather and the historical actual weather, and the direction of optimizing and solving the combination is the combination direction of the maximum pearson correlation of the random variable after the combination of the historical predicted weather and the historical actual weather; determining the combination direction to obtain a trained typical correlation analysis model; based on the combined direction, a typical correlation variable and a typical correlation coefficient are obtained.
Further, the training to obtain the power prediction model includes: inputting the historical typical weather into an initial power prediction model; constructing a loss function based on the output of the initial power prediction model and the historical measured power; and adjusting the super-parameters of the initial power prediction model based on the loss function to obtain a trained power prediction model.
The invention aims to provide a generating power prediction system based on typical correlation analysis, which comprises a first acquisition module, a first training module, a second acquisition module, a second training module, a third acquisition module, a predicted weather determination module and a predicted power determination module; the first acquisition module is used for acquiring historical predicted weather and historical actual weather; the first training module is used for training to obtain a typical correlation analysis model based on the historical predicted weather and the historical actual weather; the second acquisition module is used for acquiring historical typical prediction weather and historical actual measurement power; the second training module is used for training to obtain a power prediction model based on the historical typical predicted weather and the historical actual measured power; the third acquisition module acquires the predicted weather of the period to be predicted; the predicted weather determination module inputs the predicted weather into the typical correlation analysis model, and the model outputs typical predicted weather; the predicted power determination module inputs the typical predicted weather into the power prediction model, which outputs a predicted power.
Further, the device also comprises a preprocessing module; the preprocessing module is used for dividing the historical predicted weather, the historical actual weather and the historical actual power according to time sequence to obtain a training set, a verification set and a test set; the training set is used for training the typical correlation analysis model and the power prediction model; the validation set is used for adjusting super parameters in the power prediction model; the test set is used for testing the typical correlation analysis model and the power prediction model; normalizing the historical predicted weather, the historical actual weather and the historical measured power in the training set, the verification set and the test set respectively; wherein, the normalized expression is:
Figure BDA0004045496730000041
/>
Figure BDA0004045496730000042
wherein, when i epsilon 1, 2 and i=1, X 1 Representing a historical actual meteorological matrix, X when i=2 2 The method comprises the steps of representing a historical prediction meteorological matrix, wherein the number of rows of the matrix is the number of samples, and the number of columns is the number of meteorological indexes; y represents a history actual measurement power matrix (vector), the number of lines is the number of samples, the number of columns is 1,
Figure BDA0004045496730000043
representing the historical actual meteorological matrix or the historical forecast meteorological matrix after normalization processing,
Figure BDA0004045496730000044
x represents i Column minimum of>
Figure BDA0004045496730000045
X represents i Is the column maximum value of (2); />
Figure BDA0004045496730000046
Represents the history measured power after normalization processing, Y min Represents the column minimum value of Y max Represents the column maximum value of Y.
Further, the training by the first training module to obtain a typical correlation analysis model includes: inputting the historical predicted weather and the historical actual weather into an initial typical correlation analysis model; the initial typical correlation model carries out linear combination on the historical predicted weather and the historical actual weather, and the direction of optimizing and solving the combination is the combination direction of the maximum pearson correlation of the random variable after the combination of the historical predicted weather and the historical actual weather; determining the combination direction to obtain a trained typical correlation analysis model; based on the combined direction, a typical correlation variable and a typical correlation coefficient are obtained.
Further, the second training module trains to obtain a power prediction model, which comprises: inputting the historical typical weather into an initial power prediction model; constructing a loss function based on the output of the initial power prediction model and the historical measured power; and adjusting the super-parameters of the initial power prediction model based on the loss function to obtain a trained power prediction model.
The technical scheme of the embodiment of the invention has at least the following advantages and beneficial effects:
the invention introduces typical correlation analysis into power prediction, overcomes the problem of model migration in power prediction, and reduces the dependence of the power prediction model on the prediction weather accuracy, thereby improving the accuracy and the robustness of power prediction.
Drawings
FIG. 1 is an exemplary flow chart of a method for generating power prediction based on a canonical correlation analysis according to some embodiments of the invention;
FIG. 2 is an exemplary schematic diagram of a generated power prediction system based on a typical correlation analysis according to some embodiments of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
FIG. 1 is an exemplary flow chart of a method for generating power prediction based on a representative correlation analysis according to some embodiments of the present invention. In some embodiments, the process 100 may be performed by the system 200.
As shown in fig. 1, the process 100 may include the following:
step 110, a historical predicted weather and a historical actual weather are obtained.
Historical predicted weather may refer to weather that predicts weather prior to the current time. Historical actual weather may refer to weather that is weather before the current time. The historical predicted weather may correspond to a time of the historical actual weather. For example, the predicted weather and the actual weather at time t1 may be a pair of a historical predicted weather and a historical actual weather. In some embodiments, the historical predicted weather and the historical actual weather may be obtained in a variety of possible ways. In some embodiments, the predicted and actual weather may be obtained once at intervals (e.g., 15mi intervals) as historical predicted and actual weather. In some embodiments, the predicted weather and the actual weather indicators may be consistent, for wind power generation, the historical predicted weather, the historical actual weather, and the predicted weather include wind speed, wind direction, and temperature at the wind hub, etc. In some embodiments, the predicted weather and the actual weather indicators may be inconsistent, for photovoltaic power generation, the historical predicted weather and the predicted weather including total irradiance and temperature; the historical actual weather includes short wave radiation, long wave radiation, temperature, etc. The historical actual meteorological matrix is marked as X 1 The historical predictive weather matrix is denoted as X 2 ，X 1 ,X 2 The number of behavior samples n of (1) is listed asNumber of indexes.
Step 120, training to obtain a typical correlation analysis model based on the historical predicted weather and the historical actual weather.
The typical correlation analysis model is used for acquiring the correlation between the predicted weather and the actual weather. In some embodiments, training results in a canonical correlation analysis model, including:
the historical predicted weather and the historical actual weather are input into an initial typical correlation analysis model. Wherein the initial canonical correlation analysis model is a CCA (canonical correlation analysis) model.
The initial typical correlation model carries out linear combination on the historical predicted weather and the historical actual weather, and the direction of optimizing and solving the combination is the combination direction of the maximum pearson correlation of random variables after the historical predicted weather and the historical actual weather are combined.
Determining the combination direction to obtain a trained typical correlation analysis model; based on the combined direction, a typical correlation variable and a typical correlation coefficient can be obtained. Wherein, the typical correlation variable may refer to a random variable obtained by combining the predicted weather and the actual weather in the maximum pearson correlation direction, and their pearson correlation coefficients are regarded as typical correlation coefficients.
Step 130, obtaining historical typical predicted weather and historical measured power.
Historical typical predicted weather may refer to typical data of predicted weather in relation to historical actual weather. The historic measured power may refer to the power data of the new energy generation actually measured before the current time. In some embodiments, the historical typical predicted weather may be obtained by inputting the historical predicted weather into a typical correlation analysis model. The historically measured power may be obtained by actual measurements. The measured power is the measured generated power of the system energy source with the same time scale (such as 15mi n interval) as the meteorological data. For example, the predicted weather, actual weather, and measured power at time t1 may be a set of historical predicted weather, historical actual weather, and historical measured power. The historically measured power matrix is denoted Y and is an n x 1 column vector.
In some embodiments, further comprising: dividing the historical predicted weather, the historical actual weather and the historical actual power according to the time sequence to obtain three independent data sets: training set, validation set and test set. Wherein the time periods of the data in the training set, the validation set and the test set do not overlap. The training set is used to train a typical correlation analysis model and a power prediction model. The validation set is used to adjust the hyper-parameters in the power prediction model. The test set is used for testing a typical correlation analysis model and a power prediction model which are obtained through training.
Respectively normalizing the historical predicted weather, the historical actual weather and the historical actual power in the training set, the verification set and the test set; wherein, the normalized expression is:
Figure BDA0004045496730000081
Figure BDA0004045496730000082
wherein, when i epsilon 1, 2 and i=1, X 1 Representing a historical actual meteorological matrix, X when i=2 2 The method comprises the steps of representing a historical prediction meteorological matrix, wherein the number of rows of the matrix is the number of samples, and the number of columns is the number of meteorological indexes; y represents a history actual measurement power matrix (vector), the number of rows is the number of samples, and the number of columns is 1.
Figure BDA0004045496730000083
Representing the historical actual meteorological matrix or the historical predicted meteorological matrix after normalization processing, and +.>
Figure BDA0004045496730000084
X represents i Column minimum of>
Figure BDA0004045496730000085
X represents i Is the column maximum value of (2); />
Figure BDA0004045496730000086
The representation is normalized byHistory measured power after chemical treatment, Y min Represents the column minimum value of Y max Represents the column maximum value of Y.
And 140, training to obtain a power prediction model based on the historical typical predicted weather and the historical actual power.
The power prediction model may be used to predict measured power for a corresponding period of time based on typical predicted weather. In some embodiments, training results in a power prediction model, comprising:
historical typical weather is input into an initial power prediction model. The initial power prediction model may be an xgboost (distributed gradient enhancement library) based machine learning model.
A loss function is constructed based on the output of the initial power prediction model and the historical measured power.
And adjusting parameters of the initial power prediction model based on the loss function to obtain a trained power prediction model. And adjusting the super parameters of the model through the verification set, and taking the xgboost model with the best effect as a final power prediction model. The super parameters may include, among other things, the number of weak classifiers, the depth of the tree, etc. In some embodiments, training may be considered complete when the loss function converges or the number of iterations is greater than a threshold.
And 150, obtaining the predicted weather of the period to be predicted.
The period to be predicted may refer to a period of time during which weather needs to be predicted. In some embodiments, the predicted weather for the period to be predicted may be obtained in various possible ways. For example, future weather data may be predicted from historical weather data acquired by sensors. As another example, it may be acquired over a network.
Step 160, inputting the predicted weather into a model of a typical correlation analysis, and outputting the typical predicted weather by the model.
Features of the typical predicted weather are obtained using a typical correlation analysis model, and according to the typical correlation analysis principle, there is more correlation between the typical predicted weather and the typical actual weather than the original features.
Step 170, inputting the typical predicted weather into a power prediction model, and outputting the predicted power by the model.
Considering that the invention is compared with other two commonly used power prediction modes, namely, a p-p mode, a power prediction model is obtained based on historical prediction weather and actual measurement power generation training, and a mode of predicting power is obtained by inputting the prediction weather into the power prediction model; and the r-p mode is used for obtaining a power prediction model based on the historical actual weather and the actual measured power generation, and obtaining a predicted power mode by inputting the predicted weather into the power prediction model. The evaluation index may be rmspe, and the expression of the calculation formula may be:
Figure BDA0004045496730000101
wherein a value of rmspe closer to 1 indicates more accurate model prediction, n indicates the number of samples, y k ,
Figure BDA0004045496730000103
The k-th actual value of the output and the model predicted value of the output are respectively represented. The present invention and the comparative 2 power prediction modes are briefly described below,
and p-p, training an xgboost model by using the historical predicted weather and the historical actual measured power, and inputting the predicted weather on a test set to obtain the predicted power.
r-p: and training an xgboost model by using the historical actual measured weather and the historical actual measured power, and inputting the predicted weather on a test set to obtain the predicted power.
cca-xgboost (i.e., the power prediction mode of the present invention): obtaining a typical correlation analysis model by using historical actual weather and historical forecast weather training, and obtaining historical typical forecast weather; and training the xgboost model by using the historical typical predicted weather and the historical actual measured power to obtain a power prediction model, and inputting the future typical predicted weather into the power prediction model to obtain the predicted power.
The three prediction models are tested by using the data of a wind power plant in a certain province, and the three prediction models are consistent in other training details (such as data preprocessing, partial parameter adjustment of an xgboost model and the like), and the following table shows the test results in a certain month:
TABLE 1 rmspe values of three Power prediction models applied to wind Power prediction
Figure BDA0004045496730000102
Figure BDA0004045496730000111
The above table shows to some extent that the method and model for predicting power according to the present invention have advantages over the other two prediction modes. In engineering applications, xgboost may be replaced with other machine learning models.
FIG. 2 is an exemplary schematic diagram of a generated power prediction system based on a typical correlation analysis according to some embodiments of the present invention. As shown in FIG. 2, the system 200 includes a first acquisition module 210, a first training module 220, a second acquisition module 230, a second training module 240, a third acquisition module 250, a predicted weather determination module 260, and a predicted power determination module 270.
The first acquisition module 210 is configured to acquire a historical predicted weather and a historical actual weather. For more on the first acquisition module 210, see FIG. 1 and its associated description.
The first training module 220 is configured to train to obtain a typical correlation analysis model based on the historical predicted weather and the historical actual weather. Wherein, the training by the first training module 220 to obtain a typical correlation analysis model includes: inputting the historical predicted weather and the historical actual weather into an initial typical correlation analysis model; the initial typical correlation model carries out linear combination on the historical predicted weather and the historical actual weather, and the direction of optimizing and solving the combination is the combination direction of the maximum pearson correlation of the random variable after the combination of the historical predicted weather and the historical actual weather; determining the combination direction to obtain a trained typical correlation analysis model; based on the combined direction, a typical correlation variable and a typical correlation coefficient are obtained. For more details on the first training module 220, see FIG. 1 and its associated description.
The second acquisition module 230 is configured to acquire a historical typical predicted meteorological and a historical measured power. For more details on the second acquisition module 230, see FIG. 1 and its associated description.
The second training module 240 is configured to train to obtain a power prediction model based on the historical typical predicted weather and the historical measured power. Wherein, the second training module 240 trains to obtain a power prediction model, including: inputting the historical typical weather into an initial power prediction model; constructing a loss function based on the output of the initial power prediction model and the historical measured power; and adjusting the super-parameters of the initial power prediction model based on the loss function to obtain a trained power prediction model. For more details on the second training module 240, see FIG. 1 and its associated description.
The third acquisition module 250 acquires the predicted weather for the period to be predicted. For more details on the third acquisition module 250, see FIG. 1 and its associated description.
The predicted weather determination module 260 inputs the predicted weather into a model of a typical correlation analysis that outputs a typical predicted weather. For more details on the predictive weather determination module 260, see FIG. 1 and its associated description.
The predicted power determination module 270 inputs the typical predicted weather into a power prediction model, which outputs the predicted power. For more details on the predicted power determination module 270, see FIG. 1 and its associated description.
In some embodiments, the system 200 further includes a preprocessing module, where the preprocessing module is configured to divide the historical predicted weather, the historical actual weather, and the historical actual power according to the time sequence to obtain a training set, a verification set, and a test set; respectively normalizing the historical predicted weather, the historical actual weather and the historical actual power in the training set, the verification set and the test set; wherein, the normalized expression is:
Figure BDA0004045496730000121
Figure BDA0004045496730000122
wherein, when i epsilon 1, 2 and i=1, X 1 Representing a historical actual meteorological matrix, X when i=2 2 The method comprises the steps of representing a historical prediction meteorological matrix, wherein the number of rows of the matrix is the number of samples, and the number of columns is the number of meteorological indexes; y represents a history actual measurement power matrix (vector), the number of rows is the number of samples, and the number of columns is 1.
Figure BDA0004045496730000131
Representing the historical actual meteorological matrix or the historical forecast meteorological matrix after normalization processing,
Figure BDA0004045496730000132
x represents i Column minimum of>
Figure BDA0004045496730000133
X represents i Is the column maximum value of (2); />
Figure BDA0004045496730000134
Represents the history measured power after normalization processing, Y min Represents the column minimum value of Y max Represents the column maximum value of Y.
The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A generated power prediction method based on canonical correlation analysis, comprising:
acquiring historical predicted weather and historical actual weather;
training to obtain a typical correlation analysis model based on the historical predicted weather and the historical actual weather;
acquiring historical typical prediction weather and historical actual measurement power;
training to obtain a power prediction model based on the historical typical predicted meteorological and the historical actual measured power;
acquiring a predicted weather of a period to be predicted;
inputting the predicted weather into the typical correlation analysis model, and outputting the typical predicted weather by the model;
and inputting the typical predicted meteorological into the power prediction model, and outputting predicted power by the model.
2. The method of claim 1, wherein for wind power generation, the historical predicted weather, the historical actual weather, and the predicted weather include wind speed, wind direction, and temperature at a wind hub.
3. The representative correlation analysis-based generated power prediction method according to claim 1, wherein for photovoltaic generation, the historical predicted weather and the predicted weather include total irradiance and temperature; the historical actual weather includes short wave radiation, long wave radiation, and temperature.
4. The generated power prediction method based on canonical correlation analysis according to claim 1, further comprising:
dividing the historical predicted weather, the historical actual weather and the historical actual power according to a time sequence to obtain a training set, a verification set and a test set; the training set is used for training the typical correlation analysis model and the power prediction model; the validation set is used for adjusting super parameters in the power prediction model; the test set is used for testing the typical correlation analysis model and the power prediction model;
normalizing the historical predicted weather, the historical actual weather and the historical measured power in the training set, the verification set and the test set respectively; wherein, the normalized expression is:
Figure FDA0004045496720000021
Figure FDA0004045496720000022
wherein, when i epsilon 1, 2 and i=1, X 1 Representing a historical actual meteorological matrix, X when i=2 2 The method comprises the steps of representing a historical prediction meteorological matrix, wherein the number of rows of the matrix is the number of samples, and the number of columns is the number of meteorological indexes; y represents a history actual measurement power matrix (vector), the number of lines is the number of samples, the number of columns is 1,
Figure FDA0004045496720000023
representing the historical actual meteorological matrix or the historical forecast meteorological matrix after normalization processing,
Figure FDA0004045496720000024
x represents i Column minimum of>
Figure FDA0004045496720000025
X represents i Is the column maximum value of (2); />
Figure FDA0004045496720000026
Represents the history measured power after normalization processing, Y min Represents the column minimum value of Y max Represents the column maximum value of Y.
5. The method for predicting generated power based on canonical correlation analysis of claim 1, wherein the training results in a canonical correlation analysis model, comprising:
inputting the historical predicted weather and the historical actual weather into an initial typical correlation analysis model;
the initial typical correlation model carries out linear combination on the historical predicted weather and the historical actual weather, and the direction of optimizing and solving the combination is the combination direction of the maximum pearson correlation of the random variable after the combination of the historical predicted weather and the historical actual weather;
determining the combination direction to obtain a trained typical correlation analysis model; based on the combined direction, a typical correlation variable and a typical correlation coefficient are obtained.
6. The method for predicting generated power based on canonical correlation analysis of claim 1, wherein the training results in a power prediction model comprising:
inputting the historical typical weather into an initial power prediction model;
constructing a loss function based on the output of the initial power prediction model and the historical measured power;
and adjusting the super-parameters of the initial power prediction model based on the loss function to obtain a trained power prediction model.
7. The power generation power prediction system based on the typical correlation analysis is characterized by comprising a first acquisition module, a first training module, a second acquisition module, a second training module, a third acquisition module, a predicted weather determination module and a predicted power determination module;
the first acquisition module is used for acquiring historical predicted weather and historical actual weather;
the first training module is used for training to obtain a typical correlation analysis model based on the historical predicted weather and the historical actual weather;
the second acquisition module is used for acquiring historical typical prediction weather and historical actual measurement power;
the second training module is used for training to obtain a power prediction model based on the historical typical predicted weather and the historical actual measured power;
the third acquisition module acquires the predicted weather of the period to be predicted;
the predicted weather determination module inputs the predicted weather into the typical correlation analysis model, and the model outputs typical predicted weather;
the predicted power determination module inputs the typical predicted weather into the power prediction model, which outputs a predicted power.
8. The canonical correlation analysis based generated power prediction system of claim 7, further comprising a preprocessing module;
the preprocessing module is used for dividing the historical predicted weather, the historical actual weather and the historical actual power according to time sequence to obtain a training set, a verification set and a test set;
the training set is used for training the typical correlation analysis model and the power prediction model; the validation set is used for adjusting super parameters in the power prediction model; the test set is used for testing the typical correlation analysis model and the power prediction model; normalizing the historical predicted weather, the historical actual weather and the historical measured power in the training set, the verification set and the test set respectively; wherein, the normalized expression is:
Figure FDA0004045496720000041
Figure FDA0004045496720000051
wherein, when i epsilon 1, 2 and i=1, X 1 Representing a historical actual meteorological matrix, X when i=2 2 The method comprises the steps of representing a historical prediction meteorological matrix, wherein the number of rows of the matrix is the number of samples, and the number of columns is the number of meteorological indexes; y represents a history actual measurement power matrix (vector), the number of lines is the number of samples, the number of columns is 1,
Figure FDA0004045496720000052
representing the historical actual meteorological matrix or the historical predicted meteorological matrix after normalization processing, and +.>
Figure FDA0004045496720000053
X represents i Column minimum of>
Figure FDA0004045496720000054
X represents i Is the column maximum value of (2); />
Figure FDA0004045496720000055
Represents the history measured power after normalization processing, Y min Represents the column minimum value of Y max Represents the column maximum value of Y.
9. The canonical correlation analysis-based generated power prediction system of claim 7, wherein the first training module trains to derive a canonical correlation analysis model, comprising:
inputting the historical predicted weather and the historical actual weather into an initial typical correlation analysis model;
the initial typical correlation model carries out linear combination on the historical predicted weather and the historical actual weather, and the direction of optimizing and solving the combination is the combination direction of the maximum pearson correlation of the random variable after the combination of the historical predicted weather and the historical actual weather;
determining the combination direction to obtain a trained typical correlation analysis model; based on the combined direction, a typical correlation variable and a typical correlation coefficient are obtained.
10. The canonical correlation analysis based generated power prediction system of claim 7, wherein the second training module trains to obtain a power prediction model, comprising:
inputting the historical typical weather into an initial power prediction model;
constructing a loss function based on the output of the initial power prediction model and the historical measured power;
and adjusting the super-parameters of the initial power prediction model based on the loss function to obtain a trained power prediction model.
CN202310027000.XA 2023-01-09 2023-01-09 Power generation power prediction method and system based on typical correlation analysis Pending CN116090216A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310027000.XA CN116090216A (en) 2023-01-09 2023-01-09 Power generation power prediction method and system based on typical correlation analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310027000.XA CN116090216A (en) 2023-01-09 2023-01-09 Power generation power prediction method and system based on typical correlation analysis

Publications (1)

Publication Number Publication Date
CN116090216A true CN116090216A (en) 2023-05-09

Family

ID=86207809

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310027000.XA Pending CN116090216A (en) 2023-01-09 2023-01-09 Power generation power prediction method and system based on typical correlation analysis

Country Status (1)

Country Link
CN (1) CN116090216A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116742622A (en) * 2023-08-09 2023-09-12 山东理工职业学院 Photovoltaic power generation-based power generation amount prediction method and system

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116742622A (en) * 2023-08-09 2023-09-12 山东理工职业学院 Photovoltaic power generation-based power generation amount prediction method and system
CN116742622B (en) * 2023-08-09 2023-11-03 山东理工职业学院 Photovoltaic power generation-based power generation amount prediction method and system

Similar Documents

Publication Publication Date Title
KR102052289B1 (en) Method and apparatus for detecting equivalent load of wind turbine
CN111369070B (en) Multimode fusion photovoltaic power prediction method based on envelope clustering
CN111967688B (en) Power load prediction method based on Kalman filter and convolutional neural network
Guo et al. A new fault diagnosis classifier for rolling bearing united multi-scale permutation entropy optimize VMD and cuckoo search SVM
CN104573876A (en) Wind power plant short-period wind speed prediction method based on time sequence long memory model
CN116150897A (en) Machine tool spindle performance evaluation method and system based on digital twin
CN112861429B (en) Wind turbine engine room transfer function calculation method
CN116090216A (en) Power generation power prediction method and system based on typical correlation analysis
CN116306798A (en) Ultra-short time wind speed prediction method and system
CN115689055A (en) Short-term solar irradiance prediction method and device
CN115358437A (en) Power supply load prediction method based on convolutional neural network
CN110555566B (en) B-spline quantile regression-based photoelectric probability density prediction method
CN110414734B (en) Method for forecasting and evaluating wind resource utilization rate
CN116663727A (en) Photovoltaic power prediction method and system
CN116629406A (en) Power consumption carbon emission prediction method based on improved Prophet model
CN112581311B (en) Method and system for predicting long-term output fluctuation characteristics of aggregated multiple wind power plants
CN115577854A (en) Quantile regression wind speed interval prediction method based on EEMD-RBF combination
CN115828441A (en) Fan output scene generation and reduction method based on condition generation countermeasure network
CN111027816B (en) Photovoltaic power generation efficiency calculation method based on data envelope analysis
CN113256018B (en) Wind power ultra-short term probability prediction method based on conditional quantile regression model
Zhang et al. Fault Diagnosis of Wind Turbine Gearbox Based on Multisensor Data Fusion
Duan et al. Study on performance evaluation and prediction of mixed-flow hydraulic turbine units under variable operating conditions
CN117634652B (en) Dam deformation interpretable prediction method based on machine learning
Wang et al. Degradation Trend Prediction of Hydropower Units Based on a Comprehensive Deterioration Index and LSTM. Energies 2022, 15, 6273
CN117688367B (en) Wind power generation ultra-short term power prediction method and device based on instant learning

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