CN113888202A - Training method and application method of electricity price prediction model - Google Patents
Training method and application method of electricity price prediction model Download PDFInfo
- Publication number
- CN113888202A CN113888202A CN202111031042.8A CN202111031042A CN113888202A CN 113888202 A CN113888202 A CN 113888202A CN 202111031042 A CN202111031042 A CN 202111031042A CN 113888202 A CN113888202 A CN 113888202A
- Authority
- CN
- China
- Prior art keywords
- system load
- load rate
- real
- data
- price
- 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 76
- 238000012549 training Methods 0.000 title claims abstract description 40
- 230000005611 electricity Effects 0.000 title claims abstract description 34
- 230000011218 segmentation Effects 0.000 claims abstract description 34
- 238000012937 correction Methods 0.000 claims abstract description 31
- 238000010801 machine learning Methods 0.000 claims abstract description 15
- 238000005457 optimization Methods 0.000 claims description 31
- 238000012545 processing Methods 0.000 claims description 18
- 239000011664 nicotinic acid Substances 0.000 claims description 16
- 238000005096 rolling process Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 230000005540 biological transmission Effects 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000003066 decision tree Methods 0.000 description 8
- 230000007547 defect Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 2
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003592 biomimetic effect Effects 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
- G06Q30/0206—Price or cost determination based on market factors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Economics (AREA)
- Finance (AREA)
- Artificial Intelligence (AREA)
- Accounting & Taxation (AREA)
- Marketing (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Game Theory and Decision Science (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Primary Health Care (AREA)
- Tourism & Hospitality (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a training method and an application method of an electricity price prediction model, which comprises the steps of firstly utilizing historical data, calculating typical curves of real-time price and system load rate and segmentation points corresponding to the system load rate by adopting a 3sigma (3 sigma) criterion algorithm and a box line graph algorithm, and then selecting the historical data with strong correlation as the input of a machine learning algorithm by combining the characteristics of each section, thereby training the corresponding prediction model. And finally, calling the model and combining a system load rate sectional type correction algorithm to accurately predict the real-time price of a certain window period length in the future.
Description
Technical Field
The invention belongs to the field of power price prediction of a power market, and particularly relates to a training method and an application method of a power price prediction model.
Background
In the process of developing green electric energy transformation, wind power generation plays a very important role. However, because the output of the wind power generation has certain randomness and fluctuation, certain uncertainty is necessarily brought to the power balance of the power system, and meanwhile, great challenge is brought to the stable operation of the power system. In addition, the real-time adjustment of the electricity price is one of the most effective methods for realizing the adjustment of the electricity utilization habits of users and balancing the thermal power output and the wind power output in the power market. The method can realize the prediction of real-time price and can coordinate contradiction between grid connection and energy storage of wind power to a great extent.
At present, the electric power market mainly aims at the prediction of the price in the day-ahead, and research on real-time price, particularly the rolling prediction of the real-time price, also has certain hysteresis. Currently, widely used prediction methods include machine learning algorithms, time series neural network algorithms, and time-phased prediction algorithms. However, the electricity price of the unit is affected by various factors such as the load of the whole network, the output of new energy, the proportion of the outgoing power and the like, so that even the algorithm predicted according to the time interval is used for the price of the peak and valley section, the algorithm cannot be used for all situations.
Disclosure of Invention
The invention aims to overcome the defects of poor generalization capability and low parameter adjusting efficiency of time-interval prediction of electricity prices in the prior art, and provides a training method and an application method of an electricity price prediction model. And finally, calling the model and combining a system load rate sectional type correction algorithm to accurately predict the real-time price of a certain window period length in the future.
In order to solve the technical problem, the technical scheme of the invention is as follows:
a method of training a power price prediction model, the method comprising:
obtaining real-time price related data from a database;
respectively preprocessing the real-time price related data according to a preset 3sigma criterion algorithm and a preset least square method to obtain a typical curve of the real-time price and a typical curve of the system load rate;
processing the typical curve of the real-time price and the typical curve of the system load rate according to a preset box line graph algorithm to obtain a segmentation standard of the system load rate;
according to the segmentation standard of the system load rate, historical data with strong correlation is obtained after processing and is used as a training data set of a machine learning algorithm;
establishing a GBDT algorithm model, inputting the training data set to train the GBDT algorithm model, and obtaining an updated GBDT algorithm model;
and performing parameter optimization on the updated GBDT algorithm model by using a bionic algorithm to obtain a GBDT optimization model.
Firstly, determining a segmentation standard of the system load rate, wherein the segmentation standard can make up the defects that the time-interval prediction in the existing price prediction technology is too mechanized and cannot be completely suitable for all predicted daily conditions; the GBDT prediction model is established by combining the bionic parameter adjusting optimization algorithm, and compared with the traditional grid parameter adjusting algorithm, the bionic optimization algorithm well combines the advantages of a machine learning algorithm and an intelligent optimization algorithm, so that the parameter adjusting speed and efficiency are greatly improved; and finally, performing deviation correction on the prediction result by using a deviation correction algorithm to finally obtain a real-time price correction value. By correcting the prediction result, not only the latest real-time price data characteristic can be utilized to the maximum extent, but also the abnormal value and the deviation value can be corrected in real time, so that the error between the predicted value and the real value can be further reduced.
Further, the real-time price related data comprises: the data comprises date data, meteorological data, total network external power transmission data, system load rate data, total network electricity price data, total network installed capacity data and total network load data.
Further, the segmentation standard of the system load rate is to divide the system load rate into a low system load rate segment, a flat system load rate segment and a high system load rate segment.
Further, fitting the whole network historical data set according to a preset 3sigma criterion algorithm and a preset least square method respectively to obtain a typical curve of real-time price and system load rate.
Further, the processing according to the segmentation standard of the system load rate to obtain historical data with strong correlation as a training data set of the machine learning algorithm includes:
obtaining a segmentation result of the system load rate according to the segmentation standard of the system load rate;
extracting the data characteristics of each section of the system load rate according to the segmentation result of the system load rate;
and for the data characteristics of each section of the system load rate, combining the pre-stored historical data to perform corresponding selection, obtaining the historical data with strong correlation and using the historical data as a training data set of a machine learning algorithm.
An application method of a power price prediction model, wherein the power price prediction model is formed by adopting a training method of the power price prediction model in any one of the above manners; the application method comprises the following steps:
acquiring a data set of a forecast day;
determining a system load rate section to which the moment to be predicted belongs according to the data set;
and correspondingly selecting the system load rate section by combining with the pre-stored historical data to obtain the historical data with strong correlation and using the historical data as a test data set.
Calling a GBDT optimization model of the electricity price prediction model training method;
inputting the test data set into the GBDT optimization model, and performing rolling prediction to obtain a real-time price prediction value with a certain window period length;
performing deviation correction processing on the real-time price predicted value of the certain window period length to obtain a corrected real-time price;
further, the system load rate section includes: a high system load rate segment, a flat system load rate segment, and a low system load rate segment.
Further, the performing deviation correction processing on the predicted value of the real-time price specifically includes:
correcting the predicted value of the real-time price by adopting a deviation correction method for the high system load rate section and the flat system load rate section;
and correcting the predicted value of the real-time price by adopting a zero setting mode for the low system load rate section.
Further, after the deviation correction processing is performed on the predicted value of the real-time price to obtain a corrected real-time price, the method further includes:
and detecting whether the real-time prices of all the opening periods are predicted or not, and if not, determining the system load rate section to which the time to be predicted belongs according to the data set.
Further, the data set includes:
the system comprises forecast daily meteorological data, forecast daily whole-network external power transmission data, forecast daily system load rate data, forecast daily whole-network electricity price data, forecast daily whole-network installed capacity data and forecast daily whole-network load data.
Compared with the prior art, the invention has the advantages that:
a training method and an application method of an electricity price prediction model are divided into three parts: the first part firstly determines a segmentation standard of the system load rate, and the segmentation standard can make up the defects that the time-interval prediction in the existing price prediction technology is too mechanized and cannot be completely suitable for all predicted day conditions; a second part: the GBDT prediction model is established by combining the bionic parameter adjusting optimization algorithm, and compared with the traditional grid parameter adjusting algorithm, the bionic optimization algorithm well combines the advantages of a machine learning algorithm and an intelligent optimization algorithm, so that the parameter adjusting speed and efficiency are greatly improved; and a third part: and finally, performing deviation correction on the prediction result by using a deviation correction algorithm to finally obtain a real-time price correction value. By correcting the prediction result, not only the latest real-time price data characteristic can be utilized to the maximum extent, but also the abnormal value and the deviation value can be corrected in real time, so that the error between the predicted value and the real value can be further reduced.
The training method and the application method of the electricity price prediction model combine the advantages of various algorithms, can make up the defect of poor generalization capability in the existing price prediction technology, improve the training and parameter adjusting speed, and reduce the error between the predicted value and the true value.
Drawings
FIG. 1 is a flowchart illustrating a method for training a power price prediction model and a method for applying the same according to the present invention;
FIG. 2 is a schematic diagram of a GBDT and bionic optimization algorithm model of a training method and an application method of the electricity price prediction model of the invention;
FIG. 3 shows the 4-month 10-day prediction results of the training method and the application method of the electricity price prediction model according to the third embodiment;
FIG. 4 shows the 4-month and 11-day prediction results of the training method and the application method of the electricity price prediction model according to the third embodiment;
FIG. 5 shows the 4-month and 12-day prediction results of the training method and the application method of the electricity price prediction model according to the third embodiment;
FIG. 6 shows the 4-month and 13-day prediction results of the training method and the application method of the electricity price prediction model according to the third embodiment;
FIG. 7 shows the prediction results of the electricity price prediction model in the third embodiment, namely, the prediction results in 4 months and 14 days in the training method and the application method.
Detailed Description
The following describes embodiments of the present invention with reference to examples:
it should be noted that the structures, proportions, sizes, and other elements shown in the specification are included for the purpose of understanding and reading only, and are not intended to limit the scope of the invention, which is defined by the claims, and any modifications of the structures, changes in the proportions and adjustments of the sizes, without affecting the efficacy and attainment of the same.
In addition, the terms "upper", "lower", "left", "right", "middle" and "one" used in the present specification are for clarity of description, and are not intended to limit the scope of the present invention, and the relative relationship between the terms and the terms is not to be construed as a scope of the present invention.
The first embodiment is as follows:
as shown in fig. 1, a method for training a power rate prediction model includes:
obtaining real-time price related data from a database;
respectively preprocessing the real-time price related data according to a preset 3sigma criterion algorithm and a preset least square method to obtain a typical curve of the real-time price and a typical curve of the system load rate;
processing the typical curve of the real-time price and the typical curve of the system load rate according to a preset box line graph algorithm to obtain a segmentation standard of the system load rate;
according to the segmentation standard of the system load rate, historical data with strong correlation is obtained after processing and is used as a training data set of a machine learning algorithm;
establishing a GBDT algorithm model, inputting the training data set to train the GBDT algorithm model, and obtaining an updated GBDT algorithm model;
and performing parameter optimization on the updated GBDT algorithm model by using a bionic algorithm to obtain a GBDT optimization model.
Firstly, determining a segmentation standard of the system load rate, wherein the segmentation standard can make up the defects that the time-interval prediction in the existing price prediction technology is too mechanized and cannot be completely suitable for all predicted daily conditions; the GBDT prediction model is established by combining the bionic parameter adjusting optimization algorithm, and compared with the traditional grid parameter adjusting algorithm, the bionic optimization algorithm well combines the advantages of a machine learning algorithm and an intelligent optimization algorithm, so that the parameter adjusting speed and efficiency are greatly improved; and finally, performing deviation correction on the prediction result by using a deviation correction algorithm to finally obtain a real-time price correction value. By correcting the prediction result, not only the latest real-time price data characteristic can be utilized to the maximum extent, but also the abnormal value and the deviation value can be corrected in real time, so that the error between the predicted value and the real value can be further reduced.
The real-time price related data comprises: the data comprises date data, meteorological data, total network external power transmission data, system load rate data, total network electricity price data, total network installed capacity data and total network load data.
The segmentation standard of the system load rate is to divide the system load rate into a low system load rate section, a flat system load rate section and a high system load rate section.
And respectively fitting the whole network historical data set according to a preset 3sigma criterion algorithm and a preset least square method to obtain a typical curve of real-time price and system load rate.
The processing according to the segmentation standard of the system load rate to obtain historical data with strong correlation as a training data set of a machine learning algorithm comprises the following steps:
obtaining a segmentation result of the system load rate according to the segmentation standard of the system load rate;
extracting the data characteristics of each section of the system load rate according to the segmentation result of the system load rate;
and for the data characteristics of each section of the system load rate, combining the pre-stored historical data to perform corresponding selection, obtaining the historical data with strong correlation and using the historical data as a training data set of a machine learning algorithm.
An application method of a power price prediction model, wherein the power price prediction model is formed by adopting a training method of the power price prediction model in any one of the above manners; the application method comprises the following steps:
acquiring a data set of a forecast day;
determining a system load rate section to which the moment to be predicted belongs according to the data set;
and correspondingly selecting the system load rate section by combining with the pre-stored historical data to obtain the historical data with strong correlation and using the historical data as a test data set.
Calling a GBDT optimization model of the electricity price prediction model training method;
inputting the test data set into the GBDT optimization model, and performing rolling prediction to obtain a real-time price prediction value with a certain window period length;
performing deviation correction processing on the real-time price predicted value of the certain window period length to obtain a corrected real-time price;
the system load rate section includes: a high system load rate segment, a flat system load rate segment, and a low system load rate segment.
The performing deviation correction processing on the predicted value of the real-time price specifically includes:
correcting the predicted value of the real-time price by adopting a deviation correction method for the high system load rate section and the flat system load rate section;
and correcting the predicted value of the real-time price by adopting a zero setting mode for the low system load rate section.
After the deviation correction processing is performed on the predicted value of the real-time price to obtain the corrected real-time price, the method further comprises the following steps:
and detecting whether the real-time prices of all the opening periods are predicted or not, and if not, determining the system load rate section to which the time to be predicted belongs according to the data set.
The data set includes:
the system comprises forecast daily meteorological data, forecast daily whole-network external power transmission data, forecast daily system load rate data, forecast daily whole-network electricity price data, forecast daily whole-network installed capacity data and forecast daily whole-network load data.
Example two:
the second embodiment is applied to the first embodiment, a set of prediction models with stronger generalization capability and more efficient parameter adjustment is established by using historical data information of the whole network where the unit is located, including whole network electricity price data, whole network load data, whole network outsourced electricity data, whole network installed capacity, date related data (whether the data are holidays, week types and quarter types) and weather data, and the modeling process is shown in fig. 1. Firstly, historical data of the last week is used, typical curves of real-time price and system load rate are respectively fitted according to a 3sigma (3 sigma) criterion algorithm and a least square method, then two dividing points of the typical curves of the real-time price and two corresponding dividing points of the typical curves of the system load rate corresponding to the dividing points are calculated by using a box plot algorithm, and 96 time moments in a whole day are divided into three sections of low system load rate, flat system load rate and high system load rate. And then, acquiring feature data with the same distribution from the historical data by combining the data characteristics of each section, training a corresponding prediction model by a machine learning algorithm, performing parameter adjustment optimization on the model by using a bionic optimization algorithm, and finally accurately predicting the real-time price of a certain window period length in the future by applying the model and a system load rate sectional type correction algorithm.
3.1 System load Rate segmentation determination
Respectively obtaining real-time price data P corresponding to 96 moments of each day in the last week by utilizing historical dataij(i 1, 2.., 7, j 1, 2.. 96) and system load rate data Rij(i 1, 2., 7, j 1, 2., 96), and then respectively calculating the average value of the real-time price and the system load rate at each moment
And then combining a least square method formula (3) and a formula (4) to respectively obtain a real-time price typical curve and a system load rate typical curve. Finally, a box diagram algorithm is utilized to obtain a lower quartile Q typical of real-time electricity price1And upper quartile Q3And the corresponding system load rate values in the system load rate typical curve are respectively used as the upper bound value of the low system load rate section and the upper bound value of the peace system load rate section, so that the system load rate is divided into a low section, a level section and a high section.
Wherein:
minfP(x) Representing a fitted real-time typical price curve function;
minfR(x) Representing a fitted typical curve function of the system load rate;
3.2 building GBDT algorithm model and parameter-adjusting bionic optimization algorithm
As shown in fig. 2, when predicting the real-time price of a certain future window period length by using a rolling prediction (once prediction every 15 minutes) method based on the calculated system load rate segmentation points, firstly, the segment to which the current time belongs is determined according to the system load rate corresponding to each time in the future window period length, then, a training data set corresponding to the segment trend is selected from historical data according to the data characteristics of the segment, then, N decision trees with the depth of D are built by using all the data, and a corresponding weak learner is generated
And the weight θ learned by each weak learneri(i 1, 2.., N), the final sum and all weak learners form a strong learner algorithm model. In the GBDT algorithm model, a bionic optimization algorithm is used for optimizing and tuning parameters of the number of decision trees, the minimum sample number of the cotyledons of the decision trees, the maximum depth of the decision trees and the maximum characteristic ratio example of data, so that a real-time price prediction model with higher generalization capability is trained.
3.3 predicted data bias algorithm correction
The real-time price of a certain window period in the future can be predicted by utilizing a prediction model formed after the GBDT algorithm model and the bionic optimization algorithm are adjusted. However, the predicted result may still have a certain error, so according to the characteristics of the real-time prices of different system load rate sections and the relation between the real-time prices and the day-ahead prices, the predicted data can be corrected by using a deviation correction method in the high system load rate section and the flat system load rate section, and the predicted data can be corrected by using a zero setting mode in the low system load rate section.
And (3) correcting algorithms for high system load rate sections and flat system load rate sections:
and (3) correcting algorithms for low system load rate sections and flat system load rate sections:
wherein:
m represents a window period length;
Pi (day ahead)A day-ahead price representing the ith time ahead of the predicted time;
Pi (real time)Representing the real-time price at the ith moment ahead of the predicted moment;
p represents the mean value of the difference value between the day-ahead price and the real-time price of the forward M window period length of the prediction time;
Pi (prediction)A predicted value representing the real-time price at the ith moment after the predicted moment;
Pi (correction)A correction value representing the real-time price at the ith time after the predicted time;
example three:
as shown in fig. 3, 4, 5, 6, and 7, real-time price data of 4 window periods in the future, corresponding to each time from 10/4/10/2021/4/14/2021, is predicted by using a rolling prediction method (prediction is performed once every 15 minutes) based on the wind farm-related real-time electricity price data, the entire grid load data, the new energy load data, the outgoing power ratio data, and historical data such as weather and holiday days of a certain wind farm 2021 in shanxi.
As shown in tables 1 and 2, in order to verify the superiority and generalization ability of the segmentation and biomimetic optimization algorithm based on the system load rate, 3 different schemes are adopted to perform comparative analysis on the prediction results. Scheme 1: only using a system load rate segmentation and GBDT algorithm, and setting the number of decision trees as 100, the maximum depth of the decision trees as 3, the minimum sample number of the cotyledons as 2 and the maximum characteristic ratio as 1; scheme 2: adding a bionic optimization algorithm for adjusting parameters on the basis of the scheme 1, wherein the number of the decision trees, the maximum depth of the decision trees, the minimum sample number of the cotyledons and the maximum characteristic proportion are optimized and adjusted through the optimization algorithm; scheme 3: and correcting by using a deviation algorithm on the basis of the scheme 2.
In order to verify the performance of the model, because the real-time electricity price is 0 price, and the traditional MAPE algorithm cannot evaluate the prediction result of the real-time electricity price, the improved MAPE algorithm is adopted, and the algorithm formula is as follows:
wherein:
Pi (true)The real value of the real-time price corresponding to the ith time in 96 days is represented;
P′i (correction)A correction value representing a real-time price corresponding to an ith time of 96 times a day;
p represents the mean value of the real-time price correction values corresponding to 96 moments in a day;
in order to verify the advantages of the real-time electricity price prediction method based on the system load rate segmentation and the bionic optimization algorithm, real-time price data with the lengths of 4 window periods in the future corresponding to each moment of the wind power plant 2021 year 4, month 10 and 2021 year 4, month 14 are respectively predicted by using a rolling prediction (prediction is performed once every 15 minutes), and the predicted value of the last moment of the window periods is used as a comparison object.
TABLE 1 test protocol
TABLE 2 partial date real-time price forecast results
While the preferred embodiments of the present invention have been described in detail, the present invention is not limited to the above embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.
Many other changes and modifications can be made without departing from the spirit and scope of the invention. It is to be understood that the invention is not to be limited to the specific embodiments, but only by the scope of the appended claims.
Claims (10)
1. A method for training a power price prediction model, the method comprising:
obtaining real-time price related data from a database;
respectively preprocessing the real-time price related data according to a preset 3sigma criterion algorithm and a preset least square method to obtain a typical curve of the real-time price and a typical curve of the system load rate;
processing the typical curve of the real-time price and the typical curve of the system load rate according to a preset box line graph algorithm to obtain a segmentation standard of the system load rate;
according to the segmentation standard of the system load rate, historical data with strong correlation is obtained after processing and is used as a training data set of a machine learning algorithm;
establishing a GBDT algorithm model, inputting the training data set to train the GBDT algorithm model, and obtaining an updated GBDT algorithm model;
and performing parameter optimization on the updated GBDT algorithm model by using a bionic algorithm to obtain a GBDT optimization model.
2. The method according to claim 1, wherein the real-time price-related data comprises: the data comprises date data, meteorological data, total network external power transmission data, system load rate data, total network electricity price data, total network installed capacity data and total network load data.
3. The method according to claim 1, wherein the system load rate is segmented into a low system load rate segment, a flat system load rate segment and a high system load rate segment.
4. The real-time electricity price prediction method based on system load rate segmentation according to claim 1, characterized in that the whole network historical data sets are respectively fitted according to a preset 3sigma criterion algorithm and a preset least square method to obtain typical curves of real-time price and system load rate.
5. The method for predicting the electricity prices in real time based on the system load rate segmentation according to claim 1, wherein the step of obtaining historical data with strong correlation after processing according to the segmentation standard of the system load rate as a training data set of a machine learning algorithm comprises:
obtaining a segmentation result of the system load rate according to the segmentation standard of the system load rate;
extracting the data characteristics of each section of the system load rate according to the segmentation result of the system load rate;
and for the data characteristics of each section of the system load rate, combining the pre-stored historical data to perform corresponding selection, obtaining the historical data with strong correlation and using the historical data as a training data set of a machine learning algorithm.
6. An application method of a power price prediction model is characterized in that the power price prediction model is formed by training by using a training method of the power price prediction model according to any one of claims 1-5; the application method comprises the following steps:
acquiring a data set of a forecast day;
determining a system load rate section to which the moment to be predicted belongs according to the data set;
correspondingly selecting the system load rate section by combining with prestored historical data to obtain historical data with strong correlation and using the historical data as a test data set;
calling a GBDT optimization model of the electricity price prediction model training method;
inputting the test data set into the GBDT optimization model, and performing rolling prediction to obtain a real-time price prediction value with a certain window period length;
and carrying out deviation correction processing on the real-time price predicted value of the certain window period length to obtain the corrected real-time price.
7. The method of claim 6, wherein the system load rate section comprises: a high system load rate segment, a flat system load rate segment, and a low system load rate segment.
8. The method for applying the electricity price prediction model according to claim 7, wherein the performing deviation correction processing on the predicted value of the real-time price specifically includes:
correcting the predicted value of the real-time price by adopting a deviation correction method for the high system load rate section and the flat system load rate section;
and correcting the predicted value of the real-time price in a zero setting mode for the low system load rate section.
9. The method of claim 6, wherein after the deviation correction processing is performed on the predicted value of the real-time price to obtain the corrected real-time price, the method further comprises:
and detecting whether the real-time prices of all the opening periods are predicted or not, and if not, determining the system load rate section to which the time to be predicted belongs according to the data set.
10. The method of claim 5, wherein the data set comprises:
the system comprises forecast daily meteorological data, forecast daily whole-network external power transmission data, forecast daily system load rate data and forecast days.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111031042.8A CN113888202A (en) | 2021-09-03 | 2021-09-03 | Training method and application method of electricity price prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111031042.8A CN113888202A (en) | 2021-09-03 | 2021-09-03 | Training method and application method of electricity price prediction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113888202A true CN113888202A (en) | 2022-01-04 |
Family
ID=79012256
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111031042.8A Pending CN113888202A (en) | 2021-09-03 | 2021-09-03 | Training method and application method of electricity price prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113888202A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115102202A (en) * | 2022-07-25 | 2022-09-23 | 中国华能集团清洁能源技术研究院有限公司 | Energy storage control method based on rolling type real-time electricity price prediction |
| CN115587531A (en) * | 2022-09-23 | 2023-01-10 | 中国华能集团清洁能源技术研究院有限公司 | Sectional type day-ahead power quota prediction method and equipment based on whole network load rate |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105069525A (en) * | 2015-07-30 | 2015-11-18 | 广西大学 | All-weather 96-point daily load curve prediction and optimization correction system |
| CN105608512A (en) * | 2016-03-24 | 2016-05-25 | 东南大学 | Short-term load forecasting method |
| CN109711636A (en) * | 2019-01-09 | 2019-05-03 | 南京工业大学 | River water level prediction method based on chaotic firefly and gradient lifting tree model |
| CN111598612A (en) * | 2020-04-28 | 2020-08-28 | 西安理工大学 | Time-sharing electricity price making method |
| CN112862142A (en) * | 2019-11-28 | 2021-05-28 | 新奥数能科技有限公司 | Load and price prediction and correction method |
-
2021
- 2021-09-03 CN CN202111031042.8A patent/CN113888202A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105069525A (en) * | 2015-07-30 | 2015-11-18 | 广西大学 | All-weather 96-point daily load curve prediction and optimization correction system |
| CN105608512A (en) * | 2016-03-24 | 2016-05-25 | 东南大学 | Short-term load forecasting method |
| CN109711636A (en) * | 2019-01-09 | 2019-05-03 | 南京工业大学 | River water level prediction method based on chaotic firefly and gradient lifting tree model |
| CN112862142A (en) * | 2019-11-28 | 2021-05-28 | 新奥数能科技有限公司 | Load and price prediction and correction method |
| CN111598612A (en) * | 2020-04-28 | 2020-08-28 | 西安理工大学 | Time-sharing electricity price making method |
Non-Patent Citations (1)
| Title |
|---|
| 彭显刚等: "基于季节性负荷自适应划分及重要点分割的多分段短期负荷预测", 电网技术, 26 April 2020 (2020-04-26), pages 603 - 613 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115102202A (en) * | 2022-07-25 | 2022-09-23 | 中国华能集团清洁能源技术研究院有限公司 | Energy storage control method based on rolling type real-time electricity price prediction |
| CN115102202B (en) * | 2022-07-25 | 2022-11-29 | 中国华能集团清洁能源技术研究院有限公司 | Energy storage control method based on rolling type real-time electricity price prediction |
| CN115587531A (en) * | 2022-09-23 | 2023-01-10 | 中国华能集团清洁能源技术研究院有限公司 | Sectional type day-ahead power quota prediction method and equipment based on whole network load rate |
| CN115587531B (en) * | 2022-09-23 | 2024-04-30 | 中国华能集团清洁能源技术研究院有限公司 | Segmented solar power limit prediction method and device based on full-network load rate |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110909912B (en) | Park electric power system net load combination prediction method based on self-adaptive error feedback | |
| CN104598986B (en) | Methods of electric load forecasting based on big data | |
| CN111598225B (en) | Air conditioner cold load prediction method based on self-adaptive deep confidence network | |
| CN111191854A (en) | Photovoltaic power generation prediction model and method based on linear regression and neural network | |
| CN115965132A (en) | Distributed photovoltaic digital twin system power prediction method based on GA-BP neural network | |
| CN110837915B (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
| CN113888202A (en) | Training method and application method of electricity price prediction model | |
| CN111461921B (en) | Load modeling typical user database updating method based on machine learning | |
| CN111539844B (en) | Moon electricity quantity prediction method based on K-L information quantity and ARIMA error correction | |
| CN109636054A (en) | Solar energy power generating amount prediction technique based on classification and error combination prediction | |
| CN114444660A (en) | Short-term power load prediction method based on attention mechanism and LSTM | |
| CN107871157B (en) | Data prediction method, system and related device based on BP and PSO | |
| CN109657874A (en) | A kind of electric power Mid-long term load forecasting method based on long memory models in short-term | |
| CN112926772A (en) | Light energy prediction method based on LSTM-GPR hybrid model | |
| CN107895211A (en) | A kind of long-medium term power load forecasting method and system based on big data | |
| CN108694475B (en) | Short-time-scale photovoltaic cell power generation capacity prediction method based on hybrid model | |
| Kou et al. | Prediction intervals for wind power forecasting: using sparse warped Gaussian process | |
| CN115622056B (en) | Energy storage optimal configuration method and system based on linear weighting and selection method | |
| CN114004405B (en) | Photovoltaic power prediction method and system based on Elman neural network and satellite cloud image | |
| CN116485582A (en) | Heat supply optimization regulation and control method and device based on deep learning | |
| CN113642784B (en) | Wind power ultra-short-term prediction method considering fan state | |
| CN114759579A (en) | Power grid active power optimization control system, method and medium based on data driving | |
| CN112990749A (en) | Intelligent low-voltage local prejudgment method for transformer area based on edge calculation | |
| CN111461462A (en) | Daily load prediction method based on TrellisNet-L STM | |
| CN112801333B (en) | XGBoost-based power distribution network line summer peak load prediction method |
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 |