CN116523540A - Ultra-short-term electricity price prediction method based on self-adaptive LGBM - Google Patents
Ultra-short-term electricity price prediction method based on self-adaptive LGBM Download PDFInfo
- Publication number
- CN116523540A CN116523540A CN202310401621.XA CN202310401621A CN116523540A CN 116523540 A CN116523540 A CN 116523540A CN 202310401621 A CN202310401621 A CN 202310401621A CN 116523540 A CN116523540 A CN 116523540A
- Authority
- CN
- China
- Prior art keywords
- electricity price
- day
- prediction
- model
- ahead
- 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
- 230000005611 electricity Effects 0.000 title claims abstract description 187
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012549 training Methods 0.000 claims abstract description 31
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 22
- 238000012795 verification Methods 0.000 claims abstract description 15
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 238000012360 testing method Methods 0.000 claims abstract description 5
- 230000036316 preload Effects 0.000 claims description 19
- 230000003044 adaptive effect Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 claims description 2
- 238000004870 electrical engineering Methods 0.000 abstract description 2
- 238000012706 support-vector machine Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000007405 data analysis Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 230000008030 elimination Effects 0.000 description 2
- 238000003379 elimination reaction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 238000012300 Sequence Analysis Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000012843 least square support vector machine Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- YHXISWVBGDMDLQ-UHFFFAOYSA-N moclobemide Chemical compound C1=CC(Cl)=CC=C1C(=O)NCCN1CCOCC1 YHXISWVBGDMDLQ-UHFFFAOYSA-N 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 230000005612 types of electricity Effects 0.000 description 1
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
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Finance (AREA)
- Marketing (AREA)
- Theoretical Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (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)
Abstract
The invention provides an ultra-short-term electricity price prediction method based on a self-adaptive LGBM, and relates to the technical field of electrical engineering. The method comprises the following steps: collecting and preprocessing historical electricity price data within a preset time length to form a sample library; splitting a sample library into a training set and a verification set, and training a day-ahead electricity price prediction model by using the training set by adopting an LGBM algorithm; s5: initializing a day-ahead electricity price prediction model, and training the day-ahead electricity price prediction model by using a training set by adopting an LGBM algorithm; testing the prediction precision of the trained day-ahead electricity price prediction model by using a verification set, and recording the prediction precision of the day-ahead electricity price prediction model; selecting the highest-precision power price prediction model from the trained day-ahead power price prediction model as an optimal model; and acquiring real-time electricity price data to form a prediction model feature vector, inputting the prediction model feature vector into an optimal model, and outputting an electricity price prediction result within a preset time length of the real-time electricity price data. The method can accurately and timely predict the day-ahead electricity price.
Description
Technical Field
The invention relates to the technical field of electrical engineering, in particular to an ultra-short-term electricity price prediction method based on a self-adaptive LGBM.
Background
In the electricity market, electricity prices are an indispensable factor because it affects the behavior of power generation enterprises, power supply enterprises, and other power buyers. Electricity price prediction is the process of predicting future electricity market prices. In particular, it needs to use historical data, including information on power demand, production costs, power markets, weather information, social activity conditions, etc., to predict future electricity price trends. Electricity price predictions may be used for a variety of purposes, including providing future electricity price predictions for electricity producers and sellers so that they can rationally plan production and sales; helping a power management mechanism to perform power resource allocation; helping the customer to select the best power usage strategy. In recent years, with the rapid development of smart grids, the number of power consumers is increased, and renewable energy sources are accessed on a large scale, electricity prices are beginning to be affected by more and more factors, which makes electricity price prediction more difficult. Therefore, how to extract and mine useful information from electricity prices and influence factors thereof is important for accurately and timely predicting.
Electricity price prediction typically relies on machine learning and data analysis techniques, which can help analyze large amounts of data and determine predictive models. The accuracy of electricity price predictions is affected by many factors, including power market fluctuations, politics, social and economic factors, and the like. The difficulty of electricity price prediction is mainly represented by:
(1) Complex market factors: the electricity market is affected by a variety of factors, such as weather, economic situation, political situation, technological progress, etc.
(2) Uncertainty: electricity prices are affected by many uncertainty factors, such as market fluctuations, policy changes, etc.
(3) Data problems: the need to collect and analyze large amounts of data is also a challenge for data accuracy and integrity. The data quantity affecting the electricity price is large, contains a large amount of useless information, and is easy to cause low accuracy. In addition, there is a high time cost when using big data analysis. In the past, researchers perform data processing on the characteristics, and the prediction precision and the time cost are affected by omitting too many useless samples.
(4) Model selection: electricity price prediction requires selection of an appropriate prediction model, but selecting a model is also a challenge. The existing prediction model is generally based on a single sample set formed by features, and is combined with a data processing model, so that excessive data is easily extracted, and the prediction accuracy is poor.
The difficulty of electricity price prediction can be overcome by methods of continuously updating and perfecting a prediction model, introducing new data technology, actively coping with market fluctuation and the like. At present, the prediction method mainly comprises algorithms such as ARIMA, holt-winter and the like of a time sequence analysis class, regression analysis of a machine learning class, random forests, support vector machines, neural networks (multi-layer perceptrons, long-short-term memory networks, generation countermeasure networks and the like), deep learning (convolutional neural networks, cyclic neural networks, attention mechanisms and the like). Among them, the Support Vector Machine (SVM) has a high generalization ability and a good robustness, and thus is widely used for price prediction. Based on the support vector machine, a frame consisting of time series segmentation, recursive feature elimination and minimum redundancy maximum correlation is proposed by the learner. Also scholars have proposed a method of predicting peak and valley prices combining nonlinear regression and SVM. In addition, a least square support vector machine combining a Radial Basis Function (RBF) and a general support vector machine kernel function is also provided for predicting electricity prices.
Data processing has a great influence on prediction accuracy. A scholars proposed a highly efficient sparse self-encoder to extract features and predict electricity prices using a nonlinear autoregressive network. Processing the input data using an improved wavelet transform and filtering the input data using a new feature selection has also been reported; adopting XG-boost, decision tree, recursion feature elimination, random forest and other methods to select and extract features; features were selected using Gray Correlation Analysis (GCA) and data was denoised using a deep neural network with stacked denoising self-encoders. The choice of electricity price prediction algorithm depends on various factors such as the characteristics of the prediction data, the prediction target, the prediction accuracy and the like. It is generally necessary to flexibly select an appropriate algorithm according to actual requirements, and to continuously perfect and optimize the prediction model. In general, the predicted average of the electricity prices at home and abroad is to be improved.
Disclosure of Invention
The invention aims to provide an ultra-short-term electricity price prediction method based on an adaptive LGBM, which can accurately and timely predict the day-ahead electricity price.
Embodiments of the invention may be implemented as follows:
the invention provides an ultra-short-term electricity price prediction method based on a self-adaptive LGBM, which comprises the following steps:
s1: collecting and preprocessing historical electricity price data within a preset time length to form an information sample;
s2: carrying out feature engineering treatment on the information sample to form effective model features;
s3: forming a sample library according to the effective model characteristics;
s4: splitting a sample library into a training set and a verification set according to time sequence and preset proportion;
s5: initializing a day-ahead electricity price prediction model, and training the day-ahead electricity price prediction model by using a training set by adopting an LGBM algorithm;
s6: testing the prediction precision of the trained day-ahead electricity price prediction model by using a verification set, and recording the prediction precision of the day-ahead electricity price prediction model;
s7: judging whether the iteration number i in the training process reaches the maximum iteration number N, if not, returning to the step S5, and if so, executing the step S8;
s8: selecting the highest-precision power price prediction model from the trained day-ahead power price prediction model as an optimal model;
s9: collecting real-time electricity price data;
s10: forming a prediction model feature vector from the real-time electricity price data in the step S9;
s11: inputting the feature vector of the prediction model in the step S10 into the optimal model in the step S8, and outputting the electricity price prediction result within the preset time length of the real-time electricity price data;
the method further comprises the steps of:
after S9 is completed, the real-time electricity price data acquired in S9 is used as historical electricity price data to update the information sample formed in S1, S1-S8 are continuously executed, the optimal model is updated, and the updated optimal model is used for predicting the next electricity price.
In an alternative embodiment, in S1, the historical electricity price data includes a class B space day preload predictor, a tuning day preload predictor, a west electricity day preload predictor, a class a day preload predictor, a local electricity day preload predictor, a day front link power predictor, and node electricity price history information.
In an alternative embodiment, S1 includes:
reconstructing the historical electricity price data to form an information sample, including: calculating corresponding hour, weekend and working day information according to the time stamp; the time, B-type space daily load predicted value, the unified daily load predicted value, the western electricity daily load predicted value, the A-type daily load predicted value, the local electricity daily load predicted value, the daily link power predicted value and the node electricity price historical information respectively take 3-day historical data, and each data has 96 x 3 = 288 dimensions.
The ultra-short-term electricity price prediction method based on the self-adaptive LGBM provided by the embodiment of the invention has the beneficial effects that:
1. the method comprises the steps of collecting historical electricity price data in a specific preset time length, training a daily electricity price prediction model by adopting an LGBM algorithm, and training the model in a short time, so that the training efficiency of the model is improved, the characteristics and the relation of the data can be accurately captured, and higher prediction precision is obtained;
2. in the process of collecting real-time electricity price data and predicting the electricity value by using the model, a sample library and the model are updated based on the real-time electricity price data, so that the model is beneficial to keeping stable prediction precision.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of an ultra-short-term electricity price prediction method based on an adaptive LGBM according to an embodiment of the present invention;
FIG. 2 is historical electricity price data collected from the Guangdong electricity market;
FIG. 3 is a node electricity price curve for the Guangdong electricity market;
FIG. 4 is a graph of node electricity price predictions versus true values for the Guangdong electricity market.
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.
Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that: like reference numerals and letters refer to like items in the following figures. Thus, once an item is defined in one drawing, no further definition or explanation thereof is necessary in the subsequent drawings.
It should be noted that the features of the embodiments of the present invention may be combined with each other without conflict.
Referring to fig. 1, the present embodiment provides an ultra-short term electricity price prediction method (hereinafter referred to as "method") based on an adaptive LGBM, the method includes the following steps:
s1: and collecting and preprocessing historical electricity price data within a preset time period to form an information sample.
The historical electricity price data comprises information closely related to electricity price, and the preprocessing comprises cleaning bad data, namely removing error data and supplementing missing data. The preprocessed data is stored in a file in a form of a table.
The method provided by the embodiment focuses on the prediction of the day-ahead node electricity price in Guangdong province, and researches find that the following data are main factors influencing the electricity price, and in a normal case, the influence weight of the following data on the electricity price exceeds 98%, that is, the electricity price can be accurately predicted based on the following data as long as the influence relationship or the corresponding relationship of the following data on the electricity price is mastered.
Thus, the historical electricity rate data collected by the present application includes: class B space day-ahead load prediction value, tuning day-ahead load prediction value, west electricity day-ahead load prediction value, class a day-ahead load prediction value, local electricity day-ahead load prediction value, day-ahead link power prediction value, and node electricity price history information. The A type and the B type are two different types of generator sets in Guangdong province. All information is stored at 15 minute intervals. Therefore, the preset time length can be 3 days, the information which mainly influences the electricity price is collected as comprehensively as possible, and the information is used for constructing a model, so that the model can accurately predict the electricity price.
The preprocessing further comprises the step of reconstructing historical electricity price data to form an information sample. The specific reconstruction mode is as follows:
1) Time: calculating corresponding hour, weekend and working day information according to the time stamp; each type of information takes 3 days of historical data, and 96×3=288 dimensions.
2) Class B space daily preload prediction value: a total of 96×3=288 dimensions were taken of 3 days of history data.
3) Tuning the daily preload predictive value: a total of 96×3=288 dimensions were taken of 3 days of history data.
4) Predicted value of daily load of west electricity: a total of 96×3=288 dimensions were taken of 3 days of history data.
5) Class a daily preload prediction value: a total of 96×3=288 dimensions were taken of 3 days of history data.
6) Local electric daily preload predictive value: a total of 96×3=288 dimensions were taken of 3 days of history data.
7) Predicted value of link power before day: a total of 96×3=288 dimensions were taken of 3 days of history data.
8) Node electricity price history information: a total of 96×3=288 dimensions were taken of 3 days of history data.
By adopting the mode to construct the information sample, particularly, each type of data takes 3 days of historical data and forms 288 dimensions, not only is each type of collected data accurately distinguished, but also the 3 days of historical data is rich enough from the viewpoint of electricity price prediction to play a main decisive role on the current electricity price, if the longer historical data is taken, the operand is too large, and the stability of the model is also influenced.
S2: and carrying out feature engineering processing on the information sample to form effective model features.
Specifically, in order to train a high-precision day-ahead electricity price prediction model, feature engineering processing is required to be performed on the preprocessed historical electricity price data to form effective model features, and then the effective model features are used as input information of the day-ahead electricity price prediction model, so that the day-ahead electricity price prediction model is trained.
S3: and forming a sample library according to the effective model characteristics.
Specifically, a sample library is formed for each information sample based on the feature engineering process in S2. The input information of the sample library is shown in table 1 below:
table 1 input information
The output information of the sample library is shown in table 2 below:
TABLE 2 output information
| Node electricity price predictive value |
| [Price_forecast] |
In this way, the formed sample library is provided with clear and definite input information (comprising a time stamp sequence, B-type space day-ahead load prediction, unified day-ahead load prediction, western electric day-ahead load prediction, A-type day-ahead load prediction, local electric day-ahead load prediction, day-ahead tie line power prediction and node electricity price historical data), the original true value of the input information is already obtained in S1, and the output information of the sample library only comprises the node electricity price prediction value, namely, the node electricity price prediction value can be directly output as long as the input information is directly input according to the day-ahead electricity price prediction model trained by the sample library, so that the efficient prediction is realized. Then, only a sample library and a special algorithm are needed to train and verify the model so as to improve the accuracy of model prediction, and a prediction model with high efficiency and accuracy can be obtained.
S4: splitting the sample library into a training set and a verification set according to time sequence and preset proportion.
Specifically, the sample library can be split into a training set and a verification set according to time sequence and a certain proportion, for example, the first 80% of information samples of the sample library are used as the training set, and the last 20% of information samples are used as the verification set, wherein the information samples of the training set have the ratio of 4 times that of the verification set, so that the requirement of verification is met, the model can be fully trained, and the prediction precision of the model is ensured.
S5: initializing a daily electricity price prediction model, and training the daily electricity price prediction model by using a training set by adopting an LGBM algorithm.
The method provided by the embodiment adopts an LGBM (full name: light Gradient Boosting Machine) algorithm to train the day-ahead electricity price prediction model. The LGBM algorithm is a gradient hoisting algorithm based on a decision tree, and is used for two-classification and multi-classification problems and can also be used for regression problems. The LGBM algorithm has the main characteristics that:
(1) And (3) quick: the LGBM algorithm can train out the model in a short time by pruning the tree, adopting parallel algorithm and other technologies.
(2) High efficiency: the LGBM algorithm uses the technologies of pre-sequencing, feature splitting and the like, so that the training efficiency of the model is improved.
(3) The accuracy is as follows: the LGBM algorithm uses a gradient lifting algorithm, and can accurately capture the characteristics and the relation of data, so that higher prediction accuracy is obtained.
The LGBM algorithm is widely applied to various data mining and machine learning fields, and particularly has been widely accepted in the big data environment.
In order to obtain a more accurate and more robust prediction model, the method provided by the embodiment automatically searches an optimal parameter set in a high-dimensional LGBM parameter space, firstly sets maximum super-parameter searching times, max_iteration, and secondly randomly selects a group of parameters from the following parameter sets in each searching process, and trains a day-ahead electricity price prediction model by using a training set. The parameter sets are as follows:
N_estimator＝[70,100,150,200,300,500,1000,1250,1500,2000,3000]
Num_leaves＝[31,50,70,90,110]
Max_depth＝[5,6,7,8,9]
Bagging_fraction＝[0.7,0.8,0.9,1.0]
Min_child_samples＝[18,20,22]
Min_child_weight＝[0.001,0.002]
Reg_alpha＝[0.0,0.001,0.01,0.1,1.0]
Reg_lamda＝[0.0,0.001,0.01,0.1,1.0]
Learning_rate＝[0.005,0.01,0.05,0.075,0.1,0.2,0.3]
the selection of the parameter set range is derived from a large number of test results, and parameters with excellent performance can be provided for different types of electricity price prediction models.
S6: and testing the prediction precision of the trained day-ahead electricity price prediction model by using the verification set, and recording the prediction precision of the day-ahead electricity price prediction model.
S7: judging whether the Iteration number i in the training process reaches the maximum Iteration number N (Max_iteration). If not, returning to S5. If yes, S8 is executed.
S8: and selecting the model with the highest precision from the trained day-ahead electricity price prediction models as an optimal model.
Specifically, for each day-ahead electricity price prediction model i, the prediction accuracy of the day-ahead electricity price prediction model i is verified by using a verification set. The model constructed in this embodiment is used for predicting electricity price, and in this embodiment, the historical electricity price data is mainly collected in S1, that is, the electricity price is predicted according to the historical electricity price data, and the influence of other factors on the electricity price needs to be eliminated, that is, the optimal model needs to be selected has higher stability, where the stability refers to that the data input by the model are the same or similar, the output prediction results are the same or similar, and unexpected prediction results with larger gaps cannot occur, so in this embodiment, the prediction accuracy uses Root Mean Square Error (RMSE) as a measurement index, the smaller the Root Mean Square Error (RMSE) is, the higher the stability of the model is, the better the performance of the model is, and the calculation formula of the Root Mean Square Error (RMSE) is as follows:
wherein RMSE is root mean square error; x is X obs Is a true value; x is X model Is a model predictive value; n is the total number of samples.
And recording the prediction precision of the model obtained after training different super parameter combinations, and selecting the model with the highest precision from the prediction precision as the most optimal model, namely, inputting the model into the real-time application module.
The optimal model can be used for obtaining the current price prediction result of the node in the future. Specifically, the information sample in the step S1 is sequentially subjected to the step S2 and the step S3 to form a sample library containing effective model features, wherein the effective model features are used as input information of a day-ahead electricity price prediction model, and an optimal model output in a training process is called to obtain a day-ahead node electricity price prediction result.
S9: and collecting real-time electricity price data.
The real-time electricity rate data here contains the same type of data as the historical electricity rate data in S1, and may also be referred to as historical electricity rate data for the prediction result.
Specifically, the real-time electricity price data comprises class B space day preload, general day preload, western electricity day preload, class A day preload, local electricity day preload, day front link power and node electricity price information.
S10: and (3) forming a prediction model feature vector from the real-time electricity price data in the step S9.
That is, the real-time electricity rate data in S9 is processed into input information for inputting the optimal model in S8.
S11: and (3) inputting the feature vector of the prediction model in the step (S10) into the optimal model in the step (S8), and outputting the electricity price prediction result within the preset time length of the real-time electricity price data.
Specifically, the historical electricity price data collected in S1 is 3-day historical data, and in the process of model training, the electricity price of any node in the 3-day historical data can be output, that is, the electricity price of the desired node can be predicted based on any one or more groups of historical data before the output predicted electricity price, for example, the electricity price on the 2 nd or 3 rd day can be predicted based on the 1 st day historical data in 3 days.
Similarly, based on the real-time electricity price data collected in S9, the predicted electricity price within 3 days can be output in S11. That is, in S1, historical electricity price data in a preset time period is collected and preprocessed, and in S11, the electricity price prediction result in the preset time period after the real-time electricity price data can be output, where the preset time period may be 3 days.
In addition, after S9 is completed, the real-time electricity price data collected in S9 is also required to be used as historical electricity price data for updating the information sample formed in S1, that is, the information sample is returned to S1, and S1 to S8 are continuously executed, the optimal model is updated, and the updated optimal model is used for the prediction of the next electricity price.
The model update period may be user defined, for example: and (3) taking the real-time electricity price data acquired in the step (S9) as historical electricity price data every two weeks, returning to the step (S1), updating the information sample, and continuing to execute the step (S1-S8) to obtain an updated optimal model, so that the model is updated regularly, and the latest model is called after the electricity price prediction is started every time, so that the prediction accuracy is improved, and the real-time application of the model is realized.
Examples:
taking the prediction of the day-ahead node electricity prices of the Guangdong electricity market as an example, the collected historical electricity price data is shown in fig. 2, and the node electricity prices of the Guangdong electricity market are shown in fig. 3. The node electricity price prediction result of the Guangdong electricity market by adopting the ultra-short-term electricity price prediction method based on the self-adaptive LGBM provided by the embodiment is shown in fig. 4, so that the error between the prediction value and the true value is smaller, and the accuracy of the prediction result is higher.
The ultra-short-term electricity price prediction method based on the self-adaptive LGBM provided by the embodiment comprises the following beneficial effects:
the LGBM algorithm is adopted to train the current price prediction model in the day, so that the model can be trained in a short time, the training efficiency of the model is improved, and the gradient lifting algorithm is adopted, so that the characteristics and the relation of data can be accurately captured, and the high prediction precision is obtained.
The present invention is not limited to the above embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (6)
1. An ultra-short term electricity price prediction method based on an adaptive LGBM, which is characterized by comprising the following steps:
s1: collecting and preprocessing historical electricity price data within a preset time length to form an information sample;
s2: performing feature engineering processing on the information sample to form effective model features;
s3: forming a sample library according to the effective model characteristics;
s4: splitting the sample library into a training set and a verification set according to time sequence and preset proportion;
s5: initializing a day-ahead electricity price prediction model, and training the day-ahead electricity price prediction model by using a training set by adopting an LGBM algorithm;
s6: testing the prediction precision of the trained day-ahead electricity price prediction model by using the verification set, and recording the prediction precision of the day-ahead electricity price prediction model;
s7: judging whether the iteration number i in the training process reaches the maximum iteration number N, if not, returning to the step S5, and if so, executing the step S8;
s8: selecting the highest-precision model from the trained day-ahead electricity price prediction model as an optimal model;
s9: collecting real-time electricity price data;
s10: forming a prediction model feature vector from the real-time electricity price data in the step S9;
s11: inputting the feature vector of the prediction model in the step S10 into the optimal model in the step S8, and outputting the electricity price prediction result within the preset time of the real-time electricity price data;
the method further comprises the steps of:
after the step S9 is completed, the real-time electricity price data acquired in the step S9 is used as historical electricity price data to update the information sample formed in the step S1, the steps S1-S8 are continuously executed, the optimal model is updated, and the updated optimal model is used for predicting the next electricity price.
2. The adaptive LGBM-based ultra-short-term electricity price prediction method according to claim 1, wherein in S1 the historical electricity price data includes a class B space day-ahead load prediction value, a tuning day-ahead load prediction value, a west electricity day-ahead load prediction value, a class a day-ahead load prediction value, a local electricity day-ahead load prediction value, a day-ahead link power prediction value, and node electricity price history information.
3. The adaptive LGBM-based ultra-short-term electricity price prediction method according to claim 2, wherein S1 includes:
reconstructing the historical electricity price data to form the information sample, wherein the method comprises the following steps of: calculating corresponding hour, weekend and working day information according to the time stamp; the time, B-type space daily load predicted value, the unified daily load predicted value, the western electricity daily load predicted value, the A-type daily load predicted value, the local electricity daily load predicted value, the daily link power predicted value and the node electricity price historical information respectively take 3-day historical data, and each data has 96 x 3 = 288 dimensions.
4. The adaptive LGBM-based ultra-short-term electricity price prediction method according to claim 3, wherein S3 includes:
and (3) forming a sample library for each information sample based on the characteristic engineering processing in the step (S2), wherein the input information of the sample library comprises a time stamp sequence, a class B space day preload prediction, a unified day preload prediction, a western electricity day preload prediction, a class A day preload prediction, a local electricity day preload prediction, a day front tie line power prediction and node electricity price historical data, and the output information of the sample library is a node electricity price predicted value.
5. The adaptive LGBM-based ultra-short-term electricity price prediction method according to claim 1, wherein S4 includes:
splitting the sample library into the training set and the verification set according to time sequence and proportion, wherein the first 80% of information samples of the sample library are used as the training set, and the last 20% of information samples are used as the verification set.
6. The adaptive LGBM-based ultra-short-term electricity price prediction method according to claim 1, wherein S8 includes:
the prediction precision of the day-ahead electricity price prediction model takes root mean square error as a measurement index, and the calculation formula of the root mean square error is as follows:
wherein RMSE is root mean square error; x is X obs Is a true value; x is X model Is a model predictive value; n is the total number of samples.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310401621.XA CN116523540A (en) | 2023-04-08 | 2023-04-08 | Ultra-short-term electricity price prediction method based on self-adaptive LGBM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310401621.XA CN116523540A (en) | 2023-04-08 | 2023-04-08 | Ultra-short-term electricity price prediction method based on self-adaptive LGBM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116523540A true CN116523540A (en) | 2023-08-01 |
Family
ID=87398640
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310401621.XA Pending CN116523540A (en) | 2023-04-08 | 2023-04-08 | Ultra-short-term electricity price prediction method based on self-adaptive LGBM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116523540A (en) |
-
2023
- 2023-04-08 CN CN202310401621.XA patent/CN116523540A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jahangir et al. | A novel electricity price forecasting approach based on dimension reduction strategy and rough artificial neural networks | |
| CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
| CN112529282A (en) | Wind power plant cluster short-term power prediction method based on space-time graph convolutional neural network | |
| Esteves et al. | Long term electricity forecast: a systematic review | |
| CN105303262A (en) | Short period load prediction method based on kernel principle component analysis and random forest | |
| CN111680841B (en) | Short-term load prediction method, system and terminal equipment based on principal component analysis | |
| Shao et al. | Hybrid model for short-term wind power forecasting based on singular spectrum analysis and a temporal convolutional attention network with an adaptive receptive field | |
| CN111738522B (en) | Photovoltaic power generation power prediction method, storage medium and terminal equipment | |
| CN113988477A (en) | Photovoltaic power short-term prediction method and device based on machine learning and storage medium | |
| Massaoudi et al. | Short-term electric load forecasting based on data-driven deep learning techniques | |
| Wang et al. | Feature selection for probabilistic load forecasting via sparse penalized quantile regression | |
| CN111967675A (en) | Photovoltaic power generation amount prediction method and prediction device | |
| CN112651576A (en) | Long-term wind power prediction method and device | |
| CN116663746A (en) | Power load prediction method and device, computer equipment and storage medium | |
| CN115238854A (en) | Short-term load prediction method based on TCN-LSTM-AM | |
| CN114444821A (en) | Integrated learning load prediction method, system and medium for power internet of things | |
| Kim et al. | Tutorial on time series prediction using 1D-CNN and BiLSTM: A case example of peak electricity demand and system marginal price prediction | |
| Usman et al. | Data analytics for short term price and load forecasting in smart grids using enhanced recurrent neural network | |
| CN114118401A (en) | Neural network-based power distribution network flow prediction method, system, device and storage medium | |
| CN113837434A (en) | Solar photovoltaic power generation prediction method and device, electronic equipment and storage medium | |
| CN114021425A (en) | Power system operation data modeling and feature selection method and device, electronic equipment and storage medium | |
| CN117277312A (en) | Gray correlation analysis-based power load influence factor method and equipment | |
| CN114862032B (en) | XGBoost-LSTM-based power grid load prediction method and device | |
| CN113837486B (en) | RNN-RBM-based distribution network feeder long-term load prediction method | |
| CN116523540A (en) | Ultra-short-term electricity price prediction method based on self-adaptive LGBM |
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 |