CN115062856A - New energy plant station medium-term output prediction method and system based on power grid private cloud - Google Patents
New energy plant station medium-term output prediction method and system based on power grid private cloud Download PDFInfo
- Publication number
- CN115062856A CN115062856A CN202210743417.1A CN202210743417A CN115062856A CN 115062856 A CN115062856 A CN 115062856A CN 202210743417 A CN202210743417 A CN 202210743417A CN 115062856 A CN115062856 A CN 115062856A
- Authority
- CN
- China
- Prior art keywords
- data
- power grid
- private cloud
- new energy
- output
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000012549 training Methods 0.000 claims abstract description 89
- 238000004364 calculation method Methods 0.000 claims abstract description 15
- 238000012937 correction Methods 0.000 claims description 33
- 238000012545 processing Methods 0.000 claims description 28
- 238000010606 normalization Methods 0.000 claims description 23
- 230000002354 daily effect Effects 0.000 claims description 19
- 238000004140 cleaning Methods 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 16
- 238000012423 maintenance Methods 0.000 claims description 15
- 230000010354 integration Effects 0.000 claims description 14
- 230000002159 abnormal effect Effects 0.000 claims description 10
- 238000010801 machine learning Methods 0.000 claims description 9
- 230000003203 everyday effect Effects 0.000 claims description 8
- 230000000007 visual effect Effects 0.000 abstract description 5
- 238000005065 mining Methods 0.000 abstract description 3
- 230000008569 process Effects 0.000 abstract description 3
- 238000011161 development Methods 0.000 description 4
- 238000010248 power generation Methods 0.000 description 4
- 230000002596 correlated effect Effects 0.000 description 3
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 229910052799 carbon Inorganic materials 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000009440 infrastructure construction Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000006467 substitution reaction Methods 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
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- 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
-
- 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
-
- 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
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/46—Controlling of the sharing of output between the generators, converters, or transformers
- H02J3/466—Scheduling the operation of the generators, e.g. connecting or disconnecting generators to meet a given demand
-
- 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
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/40—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation wherein a plurality of decentralised, dispersed or local energy generation technologies are operated simultaneously
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Software Systems (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Power Engineering (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Educational Administration (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Public Health (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a new energy plant station medium-term output prediction method and system based on a power grid private cloud. The method provided by the invention integrates and cleans data based on a power grid private cloud platform, so that the data availability is enhanced, the private cloud platform is adopted for deployment, the local city bureau is not required to invest equipment, the cloud platform multi-core GPU, super calculation and other computing resources can be shared, and 24-hour uninterrupted training is realized; the grid traversal method is adopted to realize automatic searching of optimal solution for algorithm parameters, and a complicated manual parameter adjusting process is not needed; the prediction error can be corrected in time, so that the error outside the model is effectively reduced; the output prediction of a power grid private cloud platform is relied on, data intercommunication and visual display can be carried out on the power grid private cloud platform and a three-region business system of a power grid city bureau, and follow-up potential mining and deep application of big data are facilitated.
Description
Technical Field
The invention belongs to the technical field of power system control, and particularly relates to a new energy plant station medium-term output prediction method and system based on a power grid private cloud.
Background
With the national 'double carbon' target, clean energy becomes a necessary trend for future energy development. In recent years, new energy plants such as photovoltaic plants, wind power plants and the like are put into production at an accelerated speed, and the new energy output accounts for more and more in the total output of a power grid. The new energy output prediction can provide powerful reference for power grid auxiliary decision-making and power generation planning, and the current power grid new energy prediction is performed by adopting an artificial experience method and combining daily meteorological data and historical similarity points; and secondly, intelligent algorithms such as a neural network and gradient descent are adopted, a computer is enabled to automatically learn the incidence relation between the new energy output and the characteristic data to form an algorithm model, and the algorithm model is utilized to predict the middle-short term power generation output of the new energy plant station.
The output prediction of new energy is always a difficult point in power prediction, and the accuracy is not high in practical power grid application, and the main reasons are as follows: firstly, the training model cannot reach an ideal state due to low precision and low relevance of training data; secondly, the algorithm selection and the parameter setting are unreasonable, the machine learning is converged too fast, so that overfitting is caused, the model is well represented in a training set and a testing set, but the actual prediction effect is poor; thirdly, the historical data amount is huge, the requirement of model training on computing resources is high, and the infrastructure construction of part of local cities is not enough to support; fourthly, many unexpected factors such as temporary power failure, accident maintenance and the like exist in the actual operation of the power grid, and the algorithm model cannot predict the changes.
Disclosure of Invention
In view of the above, the invention aims to solve the problem that the existing new energy output prediction method is not high in accuracy in actual power grid application.
In order to solve the technical problems, the invention provides the following technical scheme:
in a first aspect, the invention provides a new energy plant station medium-term output prediction method based on a power grid private cloud, which comprises the following steps:
acquiring associated data for predicting new energy output by using a power grid private cloud, and integrating and cleaning the data to form a feature set;
extracting new features from different time points in the time data based on the feature set, and expressing the maintenance plan features by using Boolean values to complete feature set updating;
calculating the solar height and declination angle of the new energy plant station at each moment by using the time and longitude and latitude information, adding a feature set as a new feature, completing secondary feature reconstruction, and performing normalization processing on the feature set;
selecting characteristics from the characteristic set after normalization processing, and performing online model training by using the power grid private cloud, wherein the online model is used for outputting an output predicted value according to characteristic input of a predicted time point;
traversing and dereferencing the algorithm hyper-parameters of the online model by adopting a grid traversal method, and taking the optimal parameters for training;
setting daily timing scheduling time, and inputting characteristic values of different time points at the set time to obtain a daily output prediction result;
calculating an error curve of a daily output prediction result and an actual output;
and judging the single-day prediction accuracy according to the error curve, carrying out error correction according to the single-day prediction accuracy, generating correction values at different time points, and adding the correction values into the next output prediction result.
Further, acquiring associated data for new energy output prediction by using a power grid private cloud, integrating and cleaning the data, and forming a feature set, wherein the feature set specifically comprises the following steps:
acquiring historical meteorological data, forecast meteorological data and a plant station maintenance plan by using a power grid private cloud, and matching according to data time points;
carrying out statistics on the matched data to a data table to form a feature set;
and calling the feature set by using a tool of the private cloud platform of the power grid, and eliminating data abnormal points by adopting a clustering algorithm to complete data integration and cleaning.
Further, selecting features from the feature set after normalization processing, and performing online model training by using the power grid private cloud, specifically comprising:
performing Pearson correlation coefficient calculation on all the features in the feature set after the normalization processing to obtain a correlation coefficient between each feature and the new energy output, and reserving the features of which the correlation coefficients meet set conditions to form a selected new feature set;
dividing the features in the new feature set into a training set and a testing set, wherein the training set is used for training the online model, and the testing set is used for verifying the prediction accuracy after the online model is trained;
and deploying an xgboost algorithm package by using a tool of the power grid private cloud, calling an algorithm to perform machine learning training on data of the training set, and locally storing the trained online model in the power grid private cloud.
Further, the calculation formula of the single-day prediction accuracy and the correction amount is specifically as follows:
Further, the method also comprises the following steps:
and merging the actual output data and the characteristic data of the previous day into a training set, updating the training set every day, and updating and training the online model based on the updated training set.
In a second aspect, the present invention provides a new energy plant station medium term output prediction system based on a power grid private cloud, including:
the data processing unit is used for acquiring the associated data for predicting the output of the new energy by utilizing the private cloud of the power grid, and performing data integration and cleaning to form a feature set; the system is also used for extracting new features from different time points in the time data based on the feature set, expressing the maintenance plan features by using Boolean values and finishing feature set updating; the system is also used for calculating the solar altitude and declination angle of the new energy plant station at each moment by using the time and longitude and latitude information, adding a feature set as a new feature, completing secondary feature reconstruction, and performing normalization processing on the feature set;
the output prediction unit is used for selecting characteristics from the characteristic set after normalization processing, performing online model training by using the power grid private cloud, and outputting an output prediction value according to characteristic input of a prediction time point by using the online model; the method is also used for traversing and taking values of the algorithm hyper-parameters of the online model by adopting a grid traversal method, and taking the optimal parameters for training; the system is also used for setting daily timing scheduling time, and inputting characteristic values of different time points at the set time to obtain a daily output prediction result;
the error correction unit is used for calculating an error curve of a daily output prediction result and actual output; and the correction value is added into the next output prediction result.
Further, in the data processing unit, the power grid private cloud is used for obtaining the associated data for predicting the new energy output, and data integration and cleaning are performed to form a feature set, which specifically comprises the following steps:
acquiring historical meteorological data, forecast meteorological data and a plant station maintenance plan by using a power grid private cloud, and matching according to data time points;
carrying out statistics on the matched data to a data table to form a feature set;
and calling the feature set by using a tool of the private cloud platform of the power grid, and eliminating data abnormal points by adopting a clustering algorithm to complete data integration and cleaning.
Further, in the output prediction unit, selecting features from the feature set after normalization processing, and performing online model training by using the private cloud of the power grid, specifically including:
performing Pearson correlation coefficient calculation on all the features in the feature set after the normalization processing to obtain a correlation coefficient between each feature and the new energy output, and reserving the features of which the correlation coefficients meet set conditions to form a selected new feature set;
dividing the features in the new feature set into a training set and a testing set, wherein the training set is used for training the online model, and the testing set is used for verifying the prediction accuracy after the online model is trained;
and deploying an xgboost algorithm package by using a tool of the power grid private cloud, calling an algorithm to perform machine learning training on data of the training set, and locally storing the trained online model in the power grid private cloud.
Further, in the error correction unit, the calculation formula of the single-day prediction accuracy and the correction amount is specifically as follows:
Further, the method also comprises the following steps: a model updating unit;
and the model updating unit is used for merging the actual output data and the characteristic data of the previous day into a training set, updating the training set every day and carrying out updating training on the online model based on the updated training set.
In conclusion, the invention provides a new energy plant station medium-term output prediction method and system based on a power grid private cloud. The method provided by the invention integrates and cleans data based on a power grid private cloud platform, so that the data availability is enhanced, the private cloud platform is adopted for deployment, the local city bureau is not required to invest equipment, the cloud platform multi-core GPU, super calculation and other computing resources can be shared, and 24-hour uninterrupted training is realized; the grid traversal method is adopted to realize automatic searching of optimal solution for algorithm parameters, and a complicated manual parameter adjusting process is not needed; the prediction error can be corrected in time, so that the error outside the model is effectively reduced; the output prediction of a power grid private cloud platform is relied on, data intercommunication and visual display can be carried out on the power grid private cloud platform and a three-region business system of a power grid city bureau, and follow-up potential mining and deep application of big data are facilitated.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without inventive exercise.
Fig. 1 is a schematic flow chart of a new energy plant station medium-term output prediction method based on a power grid private cloud according to an embodiment of the present invention;
fig. 2 is a structural block diagram of a new energy plant station medium term output prediction system based on a power grid private cloud according to an embodiment of the present invention.
Detailed Description
In order to make the objects, features and advantages of the present invention more obvious and understandable, the technical solutions in 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 embodiments described below are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
With the national 'double carbon' target, clean energy becomes a necessary trend for future energy development. In recent years, new energy plants such as photovoltaic plants, wind power plants and the like are put into production at an accelerated speed, and the new energy output accounts for more and more in the total output of a power grid. The new energy output prediction can provide powerful reference for power grid auxiliary decision-making and power generation planning, and the current power grid new energy prediction is performed by adopting an artificial experience method and combining daily meteorological data and historical similarity points; and secondly, intelligent algorithms such as a neural network and gradient descent are adopted, a computer is enabled to automatically learn the incidence relation between the new energy output and the characteristic data to form an algorithm model, and the algorithm model is utilized to predict the middle-short term power generation output of the new energy plant station.
The output prediction of new energy is always a difficult point in power prediction, and the accuracy is not high in practical power grid application, and the main reasons are as follows: firstly, the training model cannot reach an ideal state due to low precision and low relevance of training data; secondly, the algorithm selection and the parameter setting are unreasonable, the machine learning is converged too fast, so that overfitting is caused, the model is well represented in a training set and a testing set, but the actual prediction effect is poor; thirdly, the historical data volume is huge, the requirement of model training on computing resources is high, and part of infrastructure construction of local municipalities is not enough to support; fourthly, many unexpected factors such as temporary power failure, accident maintenance and the like exist in the actual operation of the power grid, and the algorithm model cannot predict the changes.
Aiming at the obvious problems existing in the current new energy output prediction, the invention provides a new energy station medium-term output prediction method and system based on a power grid private cloud.
The embodiment of the new energy plant station medium-term output prediction method based on the power grid private cloud is described in detail below.
Referring to fig. 1, the embodiment provides a new energy plant station medium-term output prediction method based on a power grid private cloud, including the following steps:
(1) data integration and cleaning
Specifically, the data integration and cleaning are to match partial associated data used for output prediction on a power grid private cloud platform, such as historical meteorological data, predicted meteorological data, plant station maintenance plans and the like, according to data time points, automatically eliminating time points with partial data missing by an algorithm, integrating the matched data into a data table (namely a characteristic set), and uploading the data to a Maxcomputer table space of the cloud platform; and calling the feature set through a cloud platform DSW development tool, and eliminating data abnormal points by adopting a clustering algorithm to complete data integration and cleaning.
The clustering algorithm is a commonly used abnormal value detection algorithm, for example, a K-means algorithm is used, the algorithm divides feature quantity data into a plurality of clusters, and the distance between a sample farthest from a cluster center and the cluster center in each cluster is set as a threshold of the cluster. And when new data comes, removing the abnormal data by judging whether the distance between the point change and the cluster center exceeds a threshold value or not, and if the distance exceeds the threshold value, removing the abnormal data, so that the purity of the data is kept.
(2) Feature extraction
On the basis of the feature set after the abnormal points are cleaned, the month, day, hour and minute in the time data are extracted as new features, the maintenance plan features are represented by Boolean values (planned to be 1 and unplanned to be 0), and the feature set updating is completed.
(3) Quadratic feature reconstruction
On the basis of the existing feature set, new feature construction is carried out: calculating the solar height and declination angle of the area where the plant is located at each moment by using the specific time and the detailed longitude and latitude, and adding a feature set as a new feature; and the feature set is subjected to normalization processing, so that subsequent training is facilitated.
Wherein the solar altitude and declination angle are calculated as follows:
1. declination angle δ is also called solar declination, and is the angle between the equatorial plane of the earth and the line connecting the sun and the earth's center.
Wherein n is the number of the date in one year, such as 3 months and 22 days, and n is 81.
2. The solar altitude α, also called solar altitude angle, is the angle between the incident direction of sunlight to a certain location on the earth and the ground plane
α＝arcsin(sinδsinβ+cosδcosβcosγ)……(2)
Wherein, delta is the solar declination angle, and beta is the new energy station latitude; gamma is the solar time angle.
γ＝-180°+(15n)°……(3)
Wherein n is the number of hours (0-24).
Because different feature quantities have different sizes, such as temperature 0-40 and altitude 0-3000, if the original values are directly used, the calculation results will be greatly different when performing weight-related calculation, so data normalization is performed to eliminate the influence of absolute value size, and each dimension of feature quantity is treated equally.
Set of data X ═ X1, X2, … … Xn]Wherein the maximum value X max Minimum value X min
X Normalization ＝(X-X min )/(X max -X min )。
(4) Feature relevance computation
And calculating the Pearson correlation coefficient of all the features in the feature set to obtain the correlation coefficient of each feature and the target (new energy output). Pearson correlation coefficient, which characterizes the correlation between two sets of data.
The calculation method comprises the following steps: characteristic quantity 1[ X1, X2,X3...Xn]Target amount [ Y1, Y2, Y3... Yn]Average values are respectively X av ,Y av 。
Product of standard deviation:
pearson correlation coefficient:
wherein 0.8-1.0 is extremely strong correlation; 0.6-0.8 are strongly correlated; 0.4-0.6 are moderately related; 0.2-0.4 are weakly correlated; 0.0-0.2 are very weakly or no correlated.
(5) Feature selection
And eliminating the features with the relevance lower than a set value (0.3 in the embodiment) to form a latest selected feature set.
(6) Training set and test set partitioning
And randomly dividing the feature set data into a training set and a testing set according to a 4:1 ratio. The training set is used for algorithm model training, and the testing set is used for verifying the prediction accuracy after model training is completed.
(7) On-line model training
The method comprises the steps of deploying an xgboost algorithm package by using a cloud platform DSW development tool, calling an algorithm to perform machine learning training on training set data, storing a model after training in a cloud platform local storage, and conveniently calling at any time.
(8) Algorithm hyper-parameter automatic tuning
By adopting a grid traversal method, traversal values are carried out on important parameters of the algorithm, such as a loss function, a learning rate, a regularization term, a maximum depth, child node weights, father node weights and the like, model training is carried out on each parameter value combination, the algorithm is comprehensively researched and judged, and the parameter setting with the least overfitting condition is automatically selected. This step requires a lot of time and computing resources, and the cloud platform solves this problem well.
An example of a trellis traversal method is as follows:
for example, three algorithm parameters, X, Y, Z, need to be adjusted.
The value to be selected X: 0.2, 0.5, 0.7;
Y：0.1、0.3；
Z：0.02、0.13、0.24。
and 3, 2, 3 and 18 possibilities are total, and then 18 possibilities are traversed, model training is carried out on each value, and the case with the lowest overfitting degree is selected as a determination parameter.
(9) Predictive program timing scheduling
After the parameter tuning is completed, the model is determined on the current day. When the day is 15:00 in the afternoon, an algorithm model is automatically called, the characteristic values (year, month, day and hour, forecast meteorological data, declination angle, solar altitude, maintenance plan and the like) of 96 time points in the next day are input, and the model automatically obtains an output prediction result.
Since the stored training model includes the mathematical relationship between the feature quantity learned by the machine and the target value, the predicted value can be output as long as the model is called and the feature quantity at the predicted time point is input.
(10) Visual display of prediction results
Uploading the daily prediction result to a test result table of a Maxcomputer table space, calling result table data, performing visual display on a three-region interface in a local city, and displaying a daily prediction result and an actual output error curve in detail.
(11) Instantaneous error correction
If the accuracy of the actual prediction result of three consecutive days is lower than 60%, the prediction error is automatically judged, correction values at all times (96 points) are generated, and the correction values are added into the prediction result of the next day to improve the error outside the model.
The calculation formula of the single-day prediction accuracy and the correction amount is as follows:
Will be delta k And (4) superposing the prediction result of the next day as prediction correction.
(12) Training set updating
The latest feature data is incorporated into the training set. Uploading the prediction result every day, merging the actual output data and the characteristic data of the previous day into a training set after waiting for 24:00 of the next day, updating the training set in a rolling mode every day, and executing the steps 6-11 again every day.
It should be noted that the machine learning algorithm needs a large amount of training data, such as installed capacity, geographic longitude and latitude, historical meteorological data, predicted meteorological data, historical power data, plant maintenance plan, and the like, and the data can be uniformly obtained from a cloud platform big data interface. The meteorological data are from accurate weather forecast of various meteorological offices, and comprise historical wind speed, wind direction, irradiance, humidity and temperature of meteorological grids where the plant station is located and meteorological prediction data for 24 hours in the future; the installed capacity and the geography latitude and longitude are from a power grid asset management system; the plant station overhaul plan is from a power grid operation management platform and comprises the working time and the influence capacity of the unit overhaul plan.
The embodiment provides a new energy plant station medium-term output prediction method based on a power grid private cloud, data are integrated and cleaned based on the power grid private cloud platform, so that the data availability is enhanced, the data are deployed by depending on the private cloud platform, equipment investment of a city bureau is not needed, computing resources such as multi-core GPU (graphic processing unit), super-computation and the like of the cloud platform can be shared, and 24-hour uninterrupted training is realized; the grid traversal method is adopted to realize automatic searching of optimal solution for algorithm parameters, and a complicated manual parameter adjusting process is not needed; the prediction error can be corrected in time, so that the error outside the model is effectively reduced; the output prediction of a power grid private cloud platform is relied on, data intercommunication and visual display can be carried out on the power grid private cloud platform and a three-region business system of a power grid city bureau, and follow-up potential mining and deep application of big data are facilitated.
The embodiment of the new energy plant station medium-term output prediction method based on the power grid private cloud is described in detail above, and the embodiment of the new energy plant station medium-term output prediction system based on the power grid private cloud is described in detail below.
Referring to fig. 2, the present embodiment provides a new energy plant station medium term output prediction system based on a power grid private cloud, including: the device comprises a data processing unit, an output prediction unit, an error correction unit and a model updating unit.
In this embodiment, the data processing unit is configured to acquire associated data for predicting new energy output by using a power grid private cloud, and perform data integration and cleaning to form a feature set; the system is also used for extracting new features from different time points in the time data based on the feature set, expressing the maintenance plan features by using Boolean values and finishing feature set updating; and the method is also used for calculating the solar altitude and declination angle of the new energy plant station at each moment by using the time and longitude and latitude information, adding the feature set as a new feature, completing secondary feature reconstruction, and performing normalization processing on the feature set.
The method comprises the following steps of obtaining relevant data for new energy output prediction by utilizing a power grid private cloud, integrating and cleaning the data, forming a feature set, and specifically comprising the following steps of: acquiring historical meteorological data, forecast meteorological data and a plant station maintenance plan by using a power grid private cloud, and matching according to data time points; carrying out statistics on the matched data to a data table to form a feature set; and (3) calling the feature set by using a tool of a private cloud platform of the power grid, and eliminating abnormal data points by adopting a clustering algorithm to complete data integration and cleaning.
In this embodiment, the output prediction unit is configured to select a feature from the feature set after the normalization processing, perform online model training using a power grid private cloud, and output an output prediction value according to a feature input at a prediction time point by using an online model; the method is also used for traversing and taking values of the algorithm hyper-parameters of the online model by adopting a grid traversal method, and taking the optimal parameters for training; and the method is also used for setting daily timing scheduling time, and inputting characteristic values of different time points at the set time to obtain a daily output prediction result.
The method comprises the following steps of selecting features from feature sets after normalization processing, and performing online model training by using a power grid private cloud, wherein the method specifically comprises the following steps: performing Pearson correlation coefficient calculation on all the features in the feature set after the normalization processing to obtain a correlation coefficient between each feature and the new energy output, and reserving the features of which the correlation coefficients meet set conditions to form a selected new feature set; dividing the features in the new feature set into a training set and a testing set, wherein the training set is used for training the online model, and the testing set is used for verifying the prediction accuracy after the online model is trained; and deploying an xgboost algorithm package by using a tool of the power grid private cloud, calling an algorithm to perform machine learning training on data of the training set, and locally storing the trained online model in the power grid private cloud.
In this embodiment, the error correction unit is configured to calculate an error curve between the daily output prediction result and the actual output; and the correction value is added into the next output prediction result.
The calculation formula of the single-day prediction accuracy and the correction value is as follows:
In addition, the prediction system comprises a model updating unit. And the model updating unit is used for merging the actual output data and the characteristic data of the previous day into a training set, updating the training set every day, and updating and training the online model based on the updated training set.
It should be noted that, the prediction system provided in this embodiment is used to implement the prediction method provided in the foregoing embodiment, and the specific settings of each unit are subject to complete implementation of the method, which is not described herein again.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. The new energy plant station medium-term output prediction method based on the power grid private cloud is characterized by comprising the following steps of:
acquiring associated data for predicting new energy output by using a power grid private cloud, and integrating and cleaning the data to form a feature set;
extracting new features from different time points in the time data based on the feature set, and expressing the maintenance plan features by using Boolean values to complete feature set updating;
calculating the solar height and declination angle of the new energy plant station at each moment by using the time and longitude and latitude information, adding a feature set as a new feature, completing secondary feature reconstruction, and performing normalization processing on the feature set;
selecting features from the feature set after normalization processing, and performing online model training by using the power grid private cloud, wherein the online model is used for outputting an output predicted value according to feature input of a predicted time point;
traversing and taking values of the algorithm hyper-parameters of the online model by adopting a grid traversal method, and taking the optimal parameters for training;
setting daily timing scheduling time, and inputting characteristic values of different time points at the set time to obtain a daily output prediction result;
calculating an error curve of a daily output prediction result and an actual output;
and judging the single-day prediction accuracy according to the error curve, carrying out error correction according to the single-day prediction accuracy, generating correction values at different time points, and adding the correction values into the next output prediction result.
2. The new energy plant mid-term output prediction method based on the power grid private cloud as claimed in claim 1, wherein the power grid private cloud is used for obtaining the associated data for new energy output prediction, and performing data integration and cleaning to form a feature set, and specifically comprises:
acquiring historical meteorological data, forecast meteorological data and a plant station maintenance plan by using a power grid private cloud, and matching according to data time points;
carrying out statistics on the matched data to a data table to form a feature set;
and calling the feature set by using a tool of a private cloud platform of the power grid, and eliminating data abnormal points by adopting a clustering algorithm to complete data integration and cleaning.
3. The new energy plant medium-term output prediction method based on the power grid private cloud as claimed in claim 1, wherein the method comprises the steps of selecting features from the feature set after normalization processing, and performing online model training by using the power grid private cloud, and specifically comprises the following steps:
performing Pearson correlation coefficient calculation on all the features in the feature set after normalization processing to obtain a correlation coefficient between each feature and new energy output, and reserving the features of which the correlation coefficients meet set conditions to form a selected new feature set;
dividing the features in the new feature set into a training set and a test set, wherein the training set is used for training the online model, and the test set is used for verifying the prediction accuracy after the online model is trained;
and deploying an xgboost algorithm package by using the tool of the power grid private cloud, calling an algorithm to perform machine learning training on the data of the training set, and locally storing the trained on-line model in the power grid private cloud.
4. The new energy plant medium-term output prediction method based on the power grid private cloud as claimed in claim 1, wherein the calculation formula of the single-day prediction accuracy and the correction value is as follows:
5. The new energy plant mid-term output prediction method based on the power grid private cloud as claimed in claim 1, further comprising:
and merging the actual output data and the characteristic data of the previous day into the training set, updating the training set every day, and updating and training the online model based on the updated training set.
6. New energy plant station medium-term output prediction system based on power grid private cloud is characterized by comprising the following steps:
the data processing unit is used for acquiring the associated data for predicting the output of the new energy by utilizing the private cloud of the power grid, and performing data integration and cleaning to form a feature set; the system is also used for extracting new features from different time points in time data based on the feature set, expressing the maintenance plan features by using Boolean values and finishing feature set updating; the system is also used for calculating the solar altitude and declination angle of the new energy plant station at each moment by using the time and longitude and latitude information, adding a feature set as a new feature, completing secondary feature reconstruction, and performing normalization processing on the feature set;
the output prediction unit is used for selecting characteristics from the characteristic set after normalization processing and utilizing the power grid private cloud to carry out on-line model training, and the on-line model is used for outputting an output prediction value according to characteristic input of a prediction time point; traversing and taking values of the algorithm hyper-parameters of the online model by adopting a grid traversal method, and taking the optimal parameters for training; setting daily timing scheduling time, and inputting characteristic values of different time points at the set time to obtain a daily output prediction result;
the error correction unit is used for calculating an error curve of a daily output prediction result and actual output; and the system is also used for judging the single-day prediction accuracy according to the error curve, carrying out error correction according to the single-day prediction accuracy, generating correction values at different time points, and adding the correction values into the next output prediction result.
7. The new energy plant mid-term output prediction system based on the power grid private cloud as claimed in claim 6, wherein in the data processing unit, the power grid private cloud is used to obtain the associated data for new energy output prediction, and perform data integration and cleaning to form a feature set, specifically comprising:
acquiring historical meteorological data, predicted meteorological data and a plant station overhaul plan by using a private cloud of a power grid, and matching according to data time points;
carrying out statistics on the matched data to a data table to form a feature set;
and calling the feature set by using a tool of a private cloud platform of the power grid, and eliminating data abnormal points by adopting a clustering algorithm to complete data integration and cleaning.
8. The new energy plant mid-term output prediction system based on the power grid private cloud as claimed in claim 6, wherein in the output prediction unit, features are selected from the feature set after normalization processing, and online model training is performed by using the power grid private cloud, specifically comprising:
performing Pearson correlation coefficient calculation on all the features in the feature set after normalization processing to obtain a correlation coefficient between each feature and new energy output, and reserving the features of which the correlation coefficients meet set conditions to form a selected new feature set;
dividing the features in the new feature set into a training set and a test set, wherein the training set is used for training the online model, and the test set is used for verifying the prediction accuracy after the online model is trained;
and deploying an xgboost algorithm package by using the tool of the power grid private cloud, calling an algorithm to perform machine learning training on the data of the training set, and locally storing the trained on-line model in the power grid private cloud.
9. The system according to claim 6, wherein the error correction unit calculates the single-day prediction accuracy and the correction amount according to the following formula:
10. The new energy plant mid-station contribution prediction system based on the grid private cloud of claim 6, further comprising: a model updating unit;
the model updating unit is used for merging the actual output data and the characteristic data of the previous day into the training set, updating the training set every day, and updating and training the online model based on the updated training set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210743417.1A CN115062856A (en) | 2022-06-28 | 2022-06-28 | New energy plant station medium-term output prediction method and system based on power grid private cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210743417.1A CN115062856A (en) | 2022-06-28 | 2022-06-28 | New energy plant station medium-term output prediction method and system based on power grid private cloud |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115062856A true CN115062856A (en) | 2022-09-16 |
Family
ID=83203517
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210743417.1A Pending CN115062856A (en) | 2022-06-28 | 2022-06-28 | New energy plant station medium-term output prediction method and system based on power grid private cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115062856A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116796926A (en) * | 2023-06-30 | 2023-09-22 | 国家电网有限公司华东分部 | Low-voltage side new energy resource aggregation capacity assessment method and device, medium and equipment |
| CN117411078A (en) * | 2023-09-22 | 2024-01-16 | 华中科技大学 | New energy grid-connected system output prediction method considering privacy protection |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108986470A (en) * | 2018-08-20 | 2018-12-11 | 华南理工大学 | The Travel Time Estimation Method of particle swarm algorithm optimization LSTM neural network |
| CN111177644A (en) * | 2019-12-20 | 2020-05-19 | 中国移动通信集团江苏有限公司 | Model parameter optimization method, device, equipment and storage medium |
| AU2020104000A4 (en) * | 2020-12-10 | 2021-02-18 | Guangxi University | Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model |
| CN113722375A (en) * | 2021-08-10 | 2021-11-30 | 国网浙江省电力有限公司电力科学研究院 | Double-layer hybrid photovoltaic prediction method based on multi-algorithm optimization and computer equipment |
-
2022
- 2022-06-28 CN CN202210743417.1A patent/CN115062856A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108986470A (en) * | 2018-08-20 | 2018-12-11 | 华南理工大学 | The Travel Time Estimation Method of particle swarm algorithm optimization LSTM neural network |
| CN111177644A (en) * | 2019-12-20 | 2020-05-19 | 中国移动通信集团江苏有限公司 | Model parameter optimization method, device, equipment and storage medium |
| AU2020104000A4 (en) * | 2020-12-10 | 2021-02-18 | Guangxi University | Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model |
| CN113722375A (en) * | 2021-08-10 | 2021-11-30 | 国网浙江省电力有限公司电力科学研究院 | Double-layer hybrid photovoltaic prediction method based on multi-algorithm optimization and computer equipment |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116796926A (en) * | 2023-06-30 | 2023-09-22 | 国家电网有限公司华东分部 | Low-voltage side new energy resource aggregation capacity assessment method and device, medium and equipment |
| CN116796926B (en) * | 2023-06-30 | 2024-03-15 | 国家电网有限公司华东分部 | Low-voltage side new energy resource aggregation capacity assessment method and device, medium and equipment |
| CN117411078A (en) * | 2023-09-22 | 2024-01-16 | 华中科技大学 | New energy grid-connected system output prediction method considering privacy protection |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021238505A1 (en) | Federated learning-based regional photovoltaic power probability prediction method, and cooperative regulation and control system | |
| US20220373984A1 (en) | Hybrid photovoltaic power prediction method and system based on multi-source data fusion | |
| Antonanzas et al. | Review of photovoltaic power forecasting | |
| Chang et al. | An improved neural network-based approach for short-term wind speed and power forecast | |
| Carneiro et al. | Review on photovoltaic power and solar resource forecasting: current status and trends | |
| CN115062856A (en) | New energy plant station medium-term output prediction method and system based on power grid private cloud | |
| Carriere et al. | A novel approach for seamless probabilistic photovoltaic power forecasting covering multiple time frames | |
| CN108388962B (en) | Wind power prediction system and method | |
| Kim et al. | Designing solar power generation output forecasting methods using time series algorithms | |
| CN110766200A (en) | Method for predicting generating power of wind turbine generator based on K-means mean clustering | |
| Wang et al. | Deep learning based irradiance mapping model for solar PV power forecasting using sky image | |
| CN112215428B (en) | Photovoltaic power generation power prediction method and system based on error correction and fuzzy logic | |
| CN103117546A (en) | Ultrashort-term slide prediction method for wind power | |
| CN114021420A (en) | Distributed photovoltaic ultra-short-term power generation prediction method and system | |
| CN111626473A (en) | Two-stage photovoltaic power prediction method considering error correction | |
| CN112994092B (en) | Independent wind-solar storage micro-grid system size planning method based on power prediction | |
| KR20140021179A (en) | Method and apparatus for predicting daily solar radiation level | |
| CN113128786B (en) | Wind, light and water complementary capacity and site selection optimization method based on space grid search | |
| CN115965132A (en) | Distributed photovoltaic digital twin system power prediction method based on GA-BP neural network | |
| CN115829126A (en) | Photovoltaic power generation power prediction method based on multi-view self-adaptive feature fusion | |
| CN113112085A (en) | New energy station power generation load prediction method based on BP neural network | |
| Li et al. | Wind power probabilistic forecasting based on wind correction using weather research and forecasting model | |
| Alharbi et al. | Short-term wind speed and temperature forecasting model based on gated recurrent unit neural networks | |
| CN113984198A (en) | Short wave radiation prediction method and system based on convolutional neural network | |
| CN117277304A (en) | Photovoltaic power generation ultra-short-term power prediction method and system considering sunrise and sunset time |
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 |