CN115099544A - Smart power grid load prediction method based on signal denoising - Google Patents

Smart power grid load prediction method based on signal denoising Download PDF

Info

Publication number
CN115099544A
CN115099544A CN202211038120.1A CN202211038120A CN115099544A CN 115099544 A CN115099544 A CN 115099544A CN 202211038120 A CN202211038120 A CN 202211038120A CN 115099544 A CN115099544 A CN 115099544A
Authority
CN
China
Prior art keywords
period
trend
random
time period
interested
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.)
Granted
Application number
CN202211038120.1A
Other languages
Chinese (zh)
Other versions
CN115099544B (en
Inventor
徐鹏飞
缪晓明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu Huawei Photoelectric Technology Co ltd
Original Assignee
Jiangsu Huawei Photoelectric Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jiangsu Huawei Photoelectric Technology Co ltd filed Critical Jiangsu Huawei Photoelectric Technology Co ltd
Priority to CN202211038120.1A priority Critical patent/CN115099544B/en
Publication of CN115099544A publication Critical patent/CN115099544A/en
Application granted granted Critical
Publication of CN115099544B publication Critical patent/CN115099544B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9537—Spatial or temporal dependent retrieval, e.g. spatiotemporal queries
    • 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
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas 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
    • H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
    • 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

Abstract

The invention relates to the technical field of smart power grids, in particular to a smart power grid load prediction method based on signal denoising, which comprises the following steps: determining various trend items and various non-trend items corresponding to load time sequence data according to the load time sequence data of a first set time period in the past of the power grid to be predicted; acquiring each specific interested time interval and each random interested time interval, and determining a noise degree index value and a noise threshold corresponding to each non-trend item in the two interested time intervals; screening out noise data in each non-trend item corresponding to the load time sequence data, and updating the screened noise data to obtain each non-trend item corresponding to the updated load time sequence data; and determining predicted load time sequence data of a future second set time period of the power grid to be predicted according to the trend items corresponding to the load time sequence data and the non-trend items corresponding to the updated load time sequence data. The method and the device effectively improve the accuracy of the predicted load time sequence data.

Description

Smart power grid load prediction method based on signal denoising
Technical Field
The invention relates to the technical field of smart power grids, in particular to a smart power grid load prediction method based on signal denoising.
Background
The power grid load prediction has important significance on the operation of a power system, and can be used as an important basis for power generation arrangement, energy scheduling, equipment maintenance and overhaul, and building or expanding a power plant. The accurate power grid load prediction can improve the efficiency of work arrangement of a power plant, optimize supply and demand planning and reduce energy waste, and has great significance to the economy and ecology of China, so that the prediction of power grid load data is very important.
Most of the traditional power grid load prediction is to predict future loads based on single-value loads, and the analysis of short-term uncertainty, long-term trend and periodicity of the power grid loads is neglected, so that the accuracy of the finally obtained power grid load prediction is low.
Disclosure of Invention
In order to solve the problem that the existing power grid load prediction is inaccurate, the invention aims to provide a smart power grid load prediction method based on signal denoising.
The invention provides a smart power grid load prediction method based on signal denoising, which comprises the following steps:
acquiring load time sequence data of a to-be-predicted power grid in a first set time period in the past, and determining trend items and non-trend items corresponding to the load time sequence data of the to-be-predicted power grid in the first set time period in the past according to the load time sequence data of the to-be-predicted power grid in the first set time period in the past;
acquiring each specific interesting time period and each random interesting time period corresponding to load time sequence data of a first set time period in the past of a power grid to be predicted;
determining noise degree index values and noise thresholds corresponding to non-trend items in each specific interesting time period and each random interesting time period according to the non-trend items in each specific interesting time period and each random interesting time period corresponding to load time sequence data of a first set time period in the past of a power grid to be predicted;
screening out noise data in each non-trend item corresponding to load time sequence data of a first set time period in the past of the power grid to be predicted according to noise degree index values and noise thresholds corresponding to the non-trend items in each specific interested time period and each random interested time period, and updating the screened noise data to obtain each non-trend item corresponding to the load time sequence data of the first set time period in the past of the power grid to be predicted after updating;
and determining predicted load time sequence data of a future second set time period of the power grid to be predicted according to the non-trend items corresponding to the load time sequence data of the past first set time period of the power grid to be predicted after updating and the trend items corresponding to the load time sequence data of the past first set time period of the power grid to be predicted.
Further, the step of determining the noise degree index value and the noise threshold corresponding to each non-trend item in each specific interested period and each random interested period comprises:
determining a plurality of related non-trend items corresponding to the non-trend items in each specific interesting period and each random interesting period according to the non-trend items in each specific interesting period and each random interesting period corresponding to the load time series data of the past first set time period of the power grid to be predicted, wherein the related non-trend items are the non-trend items adjacent to the non-trend items in the specific interesting period or the random interesting period;
determining noise degree index values and power grid load level indexes corresponding to the non-trend items in the specific interesting periods and the random interesting periods according to the non-trend items in the specific interesting periods and the random interesting periods and the plurality of relevant non-trend items corresponding to the non-trend items;
determining specific interested period noise threshold components of the non-trend items in the specific interested periods according to the time sequence date of the non-trend items in the specific interested periods and the time sequence date at the end point of the specific interested periods;
determining a noise threshold corresponding to each non-trend item in each specific interested period according to the specific interested period noise threshold component and the power grid load level index of each non-trend item in each specific interested period;
determining random interested period noise threshold components of the non-trend items in the random interested periods according to the time sequence date of the non-trend items in the random interested periods;
and determining the noise threshold corresponding to each non-trend item in each random interesting period according to the random interesting period noise threshold component of each non-trend item in each random interesting period and the power grid load level index.
Further, the step of determining the noise degree index value corresponding to each non-trend term in each specific interested period and each random interested period comprises:
determining the absolute value of the difference between each non-trend item in each specific interested time period and each random interested time period and the corresponding multiple related non-trend items according to the multiple related non-trend items corresponding to the non-trend items in each specific interested time period and each random interested time period;
screening a plurality of target difference absolute values from the absolute values of the differences between each non-trend item and a plurality of relevant non-trend items corresponding to the non-trend item in each specific interesting period and each random interesting period, wherein the target difference absolute value is the smaller difference absolute value in the absolute values of the differences between each non-trend item and a plurality of relevant non-trend items corresponding to the non-trend item;
and determining the noise degree index values corresponding to the non-trend items in each specific interested period and each random interested period according to the plurality of target difference absolute values corresponding to the non-trend items in each specific interested period and each random interested period.
Further, the step of determining a period-of-interest-specific noise threshold component for each non-trend term within each particular period of interest includes:
determining the closeness degree of each non-trend item in each specific interested period and the corresponding specific interested period endpoint according to the time sequence date of each non-trend item in each specific interested period and the time sequence date of each specific interested period endpoint;
and determining the specific interested period noise threshold component of each non-trend item in each specific interested period according to the proximity of each non-trend item in each specific interested period to the corresponding specific interested period endpoint.
Further, the calculation formula for determining the proximity of each non-trend term in each specific interested period to the corresponding specific interested period endpoint is as follows:
Figure 268822DEST_PATH_IMAGE002
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE003
for the first in each particular time period of interestkThe proximity of an individual non-trending item to its corresponding endpoint for a particular time period of interest,ifor the first in each particular time period of interestkThe chronological date of the individual non-trending items,
Figure 355595DEST_PATH_IMAGE004
for the first in each particular time period of interestkThe minimum chronological date at the end of a particular time period of interest for an individual non-trending item,
Figure DEST_PATH_IMAGE005
for the first in each particular time period of interestkThe maximum chronological date at the end of a particular time period of interest for each non-trending term, min () is the minimum function.
Further, the calculation formula for determining the specific interested period noise threshold component of each non-trend term in each specific interested period is as follows:
Figure DEST_PATH_IMAGE007
wherein, the first and the second end of the pipe are connected with each other,
Figure 328230DEST_PATH_IMAGE008
for the first in each particular time period of interestkThe period of particular interest noise threshold component of each non-trending term,
Figure 846543DEST_PATH_IMAGE003
for the first in each particular time period of interestkThe proximity of each non-trending term to its corresponding endpoint for a particular time period of interest.
Further, the step of determining the random interest period noise threshold component of each non-trend term in each random interest period comprises:
determining a quadratic function corresponding to each non-trend item in each random interested time period according to each non-trend item in each random interested time period and the time sequence date of each non-trend item;
determining the extreme value time sequence date corresponding to each non-trend item in each random interesting time period according to the quadratic function corresponding to each non-trend item in each random interesting time period;
and determining the random interested period noise threshold component of each non-trend item in each random interested period according to the time sequence date of each non-trend item in each random interested period and the extreme value time sequence date corresponding to each non-trend item.
Further, the calculation formula for determining the random interested period noise threshold component of each non-trend term in each random interested period is as follows:
Figure 675959DEST_PATH_IMAGE010
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE011
for the first in each random period of interestkThe random period of interest noise threshold component of each non-trending term,ifor the first in each particular time period of interestkThe chronological date of the individual non-trending items,i * for the first in each random period of interestkAnd the extreme value time sequence date corresponding to each non-trend item.
Further, the calculation formula for determining the noise threshold corresponding to each non-trend term in each specific interested time period is as follows:
Figure DEST_PATH_IMAGE013
wherein the content of the first and second substances,
Figure 922132DEST_PATH_IMAGE014
for the first in each particular time period of interestkThe noise threshold corresponding to each of the non-trend terms,
Figure DEST_PATH_IMAGE015
andnin order to be a hyper-parameter,
Figure 229617DEST_PATH_IMAGE016
for the first in each particular time period of interestkThe grid load level indicator of each non-trend term,
Figure DEST_PATH_IMAGE017
for the intra-domain factor of a particular time period of interest,
Figure 808628DEST_PATH_IMAGE008
for the first in each particular time period of interestkA period-of-interest-specific noise threshold component of the individual non-trend terms;
the calculation formula for determining the noise threshold corresponding to each non-trend term in each random interesting period is as follows:
Figure DEST_PATH_IMAGE019
wherein the content of the first and second substances,
Figure 543366DEST_PATH_IMAGE020
for the first in each random interesting periodkThe noise threshold corresponding to each non-trend term,
Figure 135890DEST_PATH_IMAGE015
andnin order to be a hyper-parameter,
Figure DEST_PATH_IMAGE021
for the first in each random period of interestkThe grid load level indicator of each non-trend term,
Figure 981486DEST_PATH_IMAGE022
is a random in-domain factor for the period of interest,
Figure 461009DEST_PATH_IMAGE011
for the first in each random period of interestkRandom periods of interest noise threshold components of the individual non-trending terms.
Further, the step of determining the predicted load time sequence data of the future second set time period of the power grid to be predicted comprises the following steps:
inputting each non-trend item corresponding to the load time sequence data of the updated power grid to be predicted in the past first set time period into a Gaussian process regression function, so as to obtain a predicted non-trend item of the power grid to be predicted in the future second set time period;
determining a predicted trend item of a future second set time period of the power grid to be predicted according to trend items corresponding to load time sequence data of the past first set time period of the power grid to be predicted;
and determining predicted load time sequence data of the future second set time period of the power grid to be predicted according to the predicted non-trend item and the predicted trend item of the future second set time period of the power grid to be predicted.
The invention has the following beneficial effects:
the method comprises the steps of obtaining trend items and non-trend items corresponding to load time sequence data of a first set time period in the past of a power grid to be predicted, determining noise degree index values and noise thresholds corresponding to the non-trend items, obtaining updated non-trend items according to the noise degree index values and the noise thresholds corresponding to the non-trend items, and determining predicted load time sequence data of a second set time period in the future of the power grid to be predicted according to the updated non-trend items and the trend items corresponding to the load time sequence data of the first set time period in the past of the power grid to be predicted.
According to the method, the load time sequence data of the past set time period of the power grid to be predicted are decomposed into the trend item and the non-trend item, the change characteristics of the load time sequence data of the past set time period of the power grid to be predicted are accurately reflected, and the accuracy of the prediction result is improved. In order to further improve the accuracy of the predicted data, the noise data in each non-trend item is identified by using the data characteristics of each non-trend item corresponding to the load time sequence data of the past set time period of the power grid to be predicted, and the noise data in each non-trend item is updated, so that the accuracy of the power grid load prediction is effectively improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a smart grid load prediction method based on signal denoising according to the present invention.
Detailed Description
To further explain the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description of the embodiments, structures, features and effects of the technical solutions according to the present invention will be given with reference to the accompanying drawings and preferred embodiments. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The embodiment provides a smart grid load prediction method based on signal denoising, as shown in fig. 1, the method includes the following steps:
(1) the method comprises the steps of obtaining load time sequence data of a power grid to be predicted in the past of a first set time period, and determining trend items and non-trend items corresponding to the load time sequence data of the power grid to be predicted in the past of the first set time period according to the load time sequence data of the power grid to be predicted in the past of the first set time period.
In this embodiment, the past of a certain area is acquiredNThe method comprises the steps of acquiring power grid load data every four hours at sampling frequency of the daily power grid load data every day, calculating the mean value of the power grid load data at different time points in one day, taking the mean value as the load time sequence data of the power grid to be predicted, and representing the power grid load level of the power grid to be predicted in one day by the power grid load mean value every day, so that the length of the power grid load data is 365, wherein the sampling frequency of the daily power grid load data is that the power grid load data are acquired every four hoursNLoad timing data of (1). For example, on a daily basis
Figure DEST_PATH_IMAGE023
…
Figure 114451DEST_PATH_IMAGE024
And (3) collecting power grid load data in real time, calculating the average value of the power grid load data corresponding to 7 time points in each day, and taking the power grid load average value as the load time sequence data of each day.
It should be noted that, due to the influence of economic development and population growth, the grid load data shows a tendency of increasing in a long term, that is, the grid load data shows a relatively slow but long-term continuous rising phenomenon as time progresses. Therefore, in order to more clearly understand the development trend of the power grid load data, after the load time series data of the past first set time period of the power grid to be predicted is obtained, the load time series data of the past first set time period is decomposed into trend items and non-trend items.
In the present embodiment, use is made ofkThe fitting is carried out by the order least square method,
Figure DEST_PATH_IMAGE025
load time sequence data of past first set time period of power grid to be predicted
Figure 413845DEST_PATH_IMAGE026
Trend item
Figure DEST_PATH_IMAGE027
Solving to obtain each trend term corresponding to the load time sequence data of the past first set time period of the power grid to be predicted, wherein the calculation formula is as follows:
Figure DEST_PATH_IMAGE029
wherein the content of the first and second substances,
Figure 984504DEST_PATH_IMAGE027
the time sequence date corresponding to the load time sequence data of the past first set time period of the power grid to be predicted isiThe trend term of (1)
Figure 115271DEST_PATH_IMAGE030
），
Figure DEST_PATH_IMAGE031
In order to optimize the parameter vector,
Figure 145806DEST_PATH_IMAGE032
the order of the least square method corresponding to the load time sequence data of the past first set time period of the power grid to be predicted. By usingkThe process of solving the fitting by the order least squares is prior art and is not within the scope of the present invention, and will not be elaborated herein.
Load time sequence data of past first set time period of power grid to be predicted
Figure 198076DEST_PATH_IMAGE026
Trend items corresponding to load time sequence data of past first set time period of power grid to be predicted
Figure 165901DEST_PATH_IMAGE027
The difference is taken from the first and the second,
Figure DEST_PATH_IMAGE033
obtaining a non-trend project corresponding to the load time sequence data of the past first set time period of the power grid to be predicted
Figure 354437DEST_PATH_IMAGE034
. To this end, the present embodiment obtains trend items and non-trend items corresponding to load time series data of a past first set time period of the power grid to be predicted.
(2) And acquiring each specific interesting time period and each random interesting time period corresponding to the load time sequence data of the past first set time period of the power grid to be predicted.
First, it should be noted that the interested time period can be divided into two types, one type is a specific interested time period, and the other type is a random interested time period. The periods of particular interest refer to seasons or holidays of fixed chronological dates each year, e.g., summer, winter and national day, the periods of chronological dates corresponding to summer being approximately [170, 240], the periods of chronological dates corresponding to national day holidays being approximately [274, 280 ]; the random interesting period refers to a peak period and a trough period in the load timing data of the set time period.
Through the data development states of all non-trend items corresponding to the load time sequence data of the past first set time period of the power grid to be predicted, all random interesting time periods corresponding to the load time sequence data of the past first set time period can be directly determined. And then, obtaining each specific interesting period corresponding to the load time sequence data of the past first set time period of the power grid to be predicted through each season of each year and the time sequence dates corresponding to a plurality of festivals and holidays. Finally, through each specific interested time period and each random interested time period corresponding to the load time sequence data of the past first set time period of the power grid to be predicted, each non-trend item in each specific interested time period and each non-trend item in each random interested time period can be obtained.
(3) And determining noise degree index values and noise thresholds corresponding to the non-trend items in each specific interested time period and each random interested time period according to the non-trend items in each specific interested time period and each random interested time period corresponding to the load time sequence data of the past first set time period of the power grid to be predicted.
Firstly, it should be noted that under the influence of some uncertain factors, each non-trend item corresponding to the load time series data of the past first set time period of the power grid to be predicted
Figure 133037DEST_PATH_IMAGE034
There is a tendency for some noisy data, such as a power outage in a certain area over a certain day or a sudden natural disaster. While the noise data generated under the influence of the season, the holiday, and the peaks and valleys are the useful noise data, the useful noise data has a sudden change with respect to other data in the vicinity, which may be caused by the influence of the periodic change, the season, or the holiday, and not a sudden occurrence of an unexpected event, so the useful noise data in the period of interest is useful for improving the accuracy of the predicted load timing data. The more likely the noise data is to be the desired noise data, the more likely the noise data is to be the desired noise data. In the peak period and the valley period, the closer the load time series data is to the peak period extreme value or the valley period extreme value, the higher the probability that it is the beneficial noise data. The step of determining the noise degree index value and the noise threshold corresponding to each non-trend item in each specific interested period and each random interested period comprises the following steps:
(3-1) determining a plurality of related non-trend items corresponding to the non-trend items in each specific interesting period and each random interesting period according to each specific interesting period and each non-trend item in each random interesting period corresponding to the load time sequence data of the past first set time period of the power grid to be predicted, wherein the related non-trend items are adjacent non-trend items in the specific interesting period or the random interesting period.
In this embodiment, the related non-trend items are non-trend items adjacent to the non-trend items in the specific interesting period or the random interesting period, and according to the non-trend items in the specific interesting period and the non-trend items in the random interesting period, a plurality of non-trend items adjacent to the time sequence date of the non-trend items are obtained, that is, the front part of each non-trend item close to the time sequence date of the non-trend item is selected from two sides of each non-trend itemHA non-trend term, in this embodimentHTo 10, each non-trend term is assigned a value of 2HThe non-trend items are used as a plurality of related non-trend items corresponding to each non-trend item.
(3-2) according to each non-trend item and a plurality of relevant non-trend items corresponding to the non-trend item in each specific interesting period and each random interesting period, determining a noise degree index value and a power grid load level index corresponding to each non-trend item in each specific interesting period and each random interesting period, wherein the steps comprise:
(3-2-1) determining noise degree index values corresponding to the non-trend items in each specific interesting period and each random interesting period according to the plurality of relevant non-trend items corresponding to the non-trend items in each specific interesting period and the plurality of relevant non-trend items corresponding to the non-trend items in each random interesting period, wherein the noise degree index values comprise the following steps:
(3-2-1-1) determining the absolute value of the difference value between each non-trend item in each specific interesting period and each random interesting period and the corresponding multiple related non-trend items according to the multiple related non-trend items corresponding to the non-trend items in each specific interesting period and each random interesting period.
In this embodiment, each non-trend term in each specific interesting period and each non-trend term in each random interesting period obtained in step (3-1) have a plurality of related non-trend terms corresponding thereto, and the absolute value of the difference between each non-trend term and its corresponding related non-trend term is calculated, so as to obtain the absolute value of the difference between each non-trend term and its corresponding related non-trend term in each specific interesting period and each random interesting period. The process of calculating the absolute value of the difference between the non-trending terms is prior art and is not within the scope of the present invention and will not be elaborated upon herein.
(3-2-1-2) screening a plurality of target difference absolute values from the absolute values of the differences between each non-trend term and a plurality of related non-trend terms corresponding to the non-trend term in each specific interested time period and each random interested time period, wherein the target difference absolute value is the smaller difference absolute value of the absolute values of the differences between each non-trend term and a plurality of related non-trend terms corresponding to the non-trend term.
In this embodiment, the target absolute difference value is a smaller absolute difference value among the absolute differences between each non-trend item and its corresponding multiple related non-trend items, the absolute differences between each non-trend item and its corresponding multiple related non-trend items in each specific interested time period and each random interested time period are sorted in the order from small to large, and the top is selected from the sorted sequence of absolute difference valueshAn absolute value of the difference, the difference is calculatedhThe absolute value of each difference is used as the absolute value of a plurality of target differences, in this embodimenthIs 5.
(3-2-1-3) calculating the average value of the absolute values of a plurality of target difference values corresponding to the non-trend items in each specific interesting period and each random interesting period, and taking the average value as the noise degree index value corresponding to the corresponding non-trend items in each specific interesting period and each random interesting period.
In this embodiment, according to the 5 target absolute difference values corresponding to each non-trend term in each specific interesting period and each random interesting period, an average value of the 5 target absolute difference values corresponding to each non-trend term in each specific interesting period and each random interesting period is calculated, and the average value of the 5 target absolute difference values corresponding to each non-trend term is used as the noise degree index value. Thus, the noise degree index values corresponding to the non-trend items in each specific interested period and each random interested period are obtained by the embodiment. It should be noted that the noise level index value refers to the possibility that the non-trend term is noise data, and each non-trend term has its corresponding noise level index value.
And (3-2-2) determining the power grid load level indexes corresponding to the non-trend items in each specific interesting period and each random interesting period according to the non-trend items and the corresponding related non-trend items in each specific interesting period and each random interesting period.
In this embodiment, according to each non-trend item and its corresponding multiple related non-trend items in each specific interesting period and each random interesting period, an average value of each non-trend item and its corresponding multiple related non-trend items in each specific interesting period and each random interesting period is calculated, the average value corresponding to each non-trend item in each specific interesting period is used as a power grid load level index corresponding to each non-trend item in each specific interesting period, and the average value corresponding to each non-trend item in each random interesting period is used as a power grid load level index corresponding to each non-trend item in each random interesting period. For example, the number of the related non-trend terms corresponding to each non-trend term in a certain interested time period is 20, and the grid load level index corresponding to each non-trend term in the interested time period is determined by calculating the average value of each non-trend term and the 20 related non-trend terms corresponding to each non-trend term, that is, by calculating the average value of 21 non-trend terms corresponding to each non-trend term.
It should be noted that the grid load level indicator refers to an average value of non-trend terms corresponding to a certain number of consecutive time series dates, and the grid load level indicator is used for subsequently determining a noise threshold corresponding to the non-trend term, for example, the higher the grid load level indicator corresponding to a certain non-trend term is, the higher the noise threshold corresponding to the non-trend term is.
(3-3) determining a period-of-interest-specific noise threshold component of each non-trend item in each period of specific interest based on the time series date of each non-trend item in each period of specific interest and the time series date at the end point of each period of specific interest, the steps comprising:
(3-3-1) determining the closeness degree of each non-trend item in each specific interesting period and the corresponding specific interesting period end point according to the time sequence date of each non-trend item in each specific interesting period and the time sequence date at each specific interesting period end point.
This embodiment is to determine the first time period of interestkThe closeness degree of each non-trend item to the corresponding specific interested period endpoint is taken as an example, according to the first time in each specific interested periodkThe time sequence date of each non-trend item and the time sequence date at the end point of each specific interesting period, and the first time sequence in each specific interesting period is calculatedkThe closeness of each non-trend term to its corresponding endpoint for a particular time period of interest is calculated by the formula:
Figure DEST_PATH_IMAGE035
wherein, the first and the second end of the pipe are connected with each other,
Figure 104228DEST_PATH_IMAGE003
for the first in each particular time period of interestkThe proximity of each non-trending term to its corresponding particular time period of interest endpoint,ifor the first in each particular time period of interestkThe chronological date of the individual non-trending items,
Figure 626476DEST_PATH_IMAGE004
for the first in each particular time period of interestkThe minimum chronological date at the end of a particular time period of interest for an individual non-trending item,
Figure 184365DEST_PATH_IMAGE005
for the first in each particular time period of interestkThe maximum chronological date at the end of a particular time period of interest for each non-trending term, min () is the minimum function.
Reference is made to the first in each particular time period of interestkDetermining the proximity of each non-trend term to its corresponding particular time period of interest endpoint to obtain the proximity of each particular time period of interestThe proximity of each non-trending term within a time period to its corresponding particular time period of interest endpoint.
It should be noted that, since the useful noise data is generally closer to the non-trend term data at the end point of a specific interesting period, the closer any non-trend term in the specific interesting period is to the end point of the specific interesting period, that is, the smaller the distance between the time-series date of any non-trend term and the time-series date at the end point of the specific interesting period is, the higher the probability that the non-trend term is useful noise data is, the lower the probability that the non-trend term is noise data is.
(3-3-2) determining the specific interesting period noise threshold component of each non-trend item in each specific interesting period according to the closeness degree of each non-trend item in each specific interesting period to the corresponding specific interesting period endpoint.
In the present embodiment, to determine the first time period of interestkTaking the noise threshold component of the specific interesting period of the non-trend term as an example, the noise threshold component is obtained according to the step (3-3-1) in the specific interesting periodkThe degree of closeness of each non-trend item to the corresponding specific interested period end point is calculated, and the first interested period in each specific interested period is calculatedkThe noise threshold component of a specific interested period of the non-trend term is calculated by the following formula:
Figure 133867DEST_PATH_IMAGE007
wherein the content of the first and second substances,
Figure 160729DEST_PATH_IMAGE008
for the first in each particular time period of interestkThe period of particular interest noise threshold component of each non-trending term,
Figure 486668DEST_PATH_IMAGE003
for the first in each particular time period of interestkThe proximity of each non-trending term to its corresponding endpoint for a particular time period of interest.
Reference is made to the first in each particular time period of interestkAnd determining the specific interested period noise threshold component of each non-trend item in each specific interested period. It should be noted that the proximity of any non-trend term within a specific time period of interest to its corresponding specific time period of interest endpoint
Figure 400528DEST_PATH_IMAGE003
The larger the noise threshold component for a particular time period of interest of the non-trending term
Figure 520931DEST_PATH_IMAGE008
The smaller the likelihood, the more likely the non-trending term is to be noisy data.
And (3-4) determining the noise threshold corresponding to each non-trend item in each specific interested period according to the specific interested period noise threshold component of each non-trend item in each specific interested period and the power grid load level index.
In the present embodiment, to determine the first time period of interestkTaking the noise threshold corresponding to each non-trend term as an example, according to the first time interval in each specific interested periodkThe power grid load level index of each non-trend item and the noise threshold component of the specific interested time period are calculated, and the first time period in each specific interested time period is calculatedkThe noise threshold corresponding to each non-trend term is calculated by the formula:
Figure 35089DEST_PATH_IMAGE013
wherein, the first and the second end of the pipe are connected with each other,
Figure 899140DEST_PATH_IMAGE014
for the first in each particular time period of interestkThe noise threshold corresponding to each of the non-trend terms,
Figure 166042DEST_PATH_IMAGE015
andnin order to be a hyper-parameter,
Figure 926187DEST_PATH_IMAGE036
，
Figure DEST_PATH_IMAGE037
，
Figure 144286DEST_PATH_IMAGE016
for the first in each particular time period of interestkThe grid load level indicator of each non-trend term,
Figure 812027DEST_PATH_IMAGE017
is the in-domain factor for a particular time period of interest,
Figure 684169DEST_PATH_IMAGE008
for the first in each particular time period of interestkThe period of specific interest noise threshold component of the individual non-trend terms.
Reference is made to the first in each particular time period of interestkThe noise threshold corresponding to each non-trend term in each specific interesting period can be obtained through the step of determining the noise threshold corresponding to each non-trend term. It should be noted that, the grid load level index of any non-trend item in each specific interesting periodhLarger, specific time period of interest noise threshold component
Figure 411953DEST_PATH_IMAGE038
The larger the non-trend term, the larger the noise threshold.
(3-5) determining the random interesting period noise threshold component of each non-trend item in each random interesting period according to the time sequence date of each non-trend item in each random interesting period, wherein the steps comprise:
and (3-5-1) determining a quadratic function corresponding to each non-trend item in each random interesting period according to each non-trend item in each random interesting period and the time sequence date of each non-trend item.
Since the distribution of the data of each non-trend item in the random interesting time period is relatively discrete, in order to obtain a more accurate extremum time sequence date in the subsequent process, the embodiment performs unitary quadratic function fitting on each non-trend item in each random interesting time period and 2 non-trend items adjacent to each non-trend item through the time sequence dates of each non-trend item and each non-trend item in each random interesting time period, so as to obtain a quadratic function corresponding to each non-trend item in each random interesting time period. The process of fitting the quadratic function is prior art and is not within the scope of the present invention, and will not be described in detail here.
And (3-5-2) determining the extreme value time sequence date corresponding to each non-trend item in each random interested time period according to the quadratic function corresponding to each non-trend item in each random interested time period.
In this embodiment, the extreme point of the quadratic function corresponding to each non-trend term is determined through the quadratic function corresponding to each non-trend term in each random interesting time period, if the quadratic function corresponding to a certain non-trend term is a peak time period, the non-trend term has a maximum value, and the time sequence date corresponding to the maximum value is an extreme value time sequence date; if the quadratic function corresponding to a certain non-trend item is a trough time period, the non-trend item has a minimum value, and the time sequence date corresponding to the minimum value is an extreme value time sequence date. For example, for non-trending terms within a random period of interest
Figure DEST_PATH_IMAGE039
The non-trend term of each time series date in the time series date is detected in the peak period and the valley period if
Figure 556495DEST_PATH_IMAGE040
And is provided with
Figure DEST_PATH_IMAGE041
Then, then
Figure 981923DEST_PATH_IMAGE042
At the peak time period, if
Figure DEST_PATH_IMAGE043
And is provided with
Figure 708570DEST_PATH_IMAGE044
Then, then
Figure 607256DEST_PATH_IMAGE042
In a valley period.
And (3-5-3) determining the random interesting period noise threshold component of each non-trend item in each random interesting period according to the time sequence date of each non-trend item in each random interesting period and the extreme value time sequence date corresponding to each non-trend item.
In this embodiment, to determine the first time period of random interestkThe random interested period noise threshold component of each non-trend term is taken as an example, and the first random interested period noise threshold component is passed throughkThe extreme value time sequence date and time sequence date corresponding to each non-trend item calculate the first time segment of random interestkThe random interested period noise threshold component of each non-trend term is calculated by the following formula:
Figure 301412DEST_PATH_IMAGE010
wherein the content of the first and second substances,
Figure 310956DEST_PATH_IMAGE011
for the first in each random period of interestkThe random period of interest noise threshold component of each non-trending term,ifor the first in each particular time period of interestkThe chronological date of the individual non-trending items,i * for the first in each random period of interestkAnd the extreme value time sequence date corresponding to each non-trend item.
Reference to the first of each random period of interestkAnd a step of determining the random interested period noise threshold component of each non-trend item, so as to obtain the random interested period noise threshold component of each non-trend item in each random interested period. It should be noted that, since the useful noise data is usually closer to the non-trend term corresponding to the extreme value time sequence date in the random interesting period, any one non-trend term in the random interesting periodThe time sequence date of (1) and the distance of the extreme value time sequence date corresponding to the non-trend item
Figure DEST_PATH_IMAGE045
The smaller the probability that the non-trend term is the beneficial noise data, the smaller the probability that the non-trend term is the noise data, i.e., the larger the random interesting period noise threshold component corresponding to the non-trend term.
And (3-6) determining the noise threshold corresponding to each non-trend item in each random interesting period according to the random interesting period noise threshold component of each non-trend item in each random interesting period and the power grid load level index.
In this embodiment, to determine the first in each random period of interestkTaking the noise threshold corresponding to each non-trend term as an example, passing the first time interval of random interestkThe random interested period noise threshold component and the power grid load level index of each non-trend item calculate the first in each random interested periodkThe noise threshold corresponding to each non-trend term is calculated by the following formula:
Figure 843175DEST_PATH_IMAGE046
wherein the content of the first and second substances,
Figure 647183DEST_PATH_IMAGE020
for the first in each random period of interestkThe noise threshold corresponding to each non-trend term,
Figure 844946DEST_PATH_IMAGE015
andnin order to be a hyper-parameter,
Figure 658181DEST_PATH_IMAGE036
，
Figure 874268DEST_PATH_IMAGE037
，
Figure 849177DEST_PATH_IMAGE021
for the first in each random period of interestkThe grid load level indicator of each non-trend term,
Figure 534236DEST_PATH_IMAGE022
is the intra-domain factor of the random period of interest,
Figure 885583DEST_PATH_IMAGE011
for the first in each random interesting periodkRandom periods of interest noise threshold components of the individual non-trending terms.
Reference to the first of each random period of interestkDetermining noise threshold corresponding to each non-trend term to obtain the first random interested periodkThe noise threshold corresponding to each non-trend term.
(4) And screening noise data in each non-trend item corresponding to the load time sequence data of the past first set time period of the power grid to be predicted according to the noise degree index value and the noise threshold corresponding to each non-trend item in each specific interested time period and each random interested time period, and updating the screened noise data to obtain each non-trend item corresponding to the load time sequence data of the past first set time period of the power grid to be predicted after updating.
In this embodiment, if the noise degree index value corresponding to any one non-trend term in each specific interested period is greater than the corresponding noise threshold, or the noise degree index value corresponding to any one non-trend term in each random interested period is greater than the noise threshold, the non-trend term is used as the noise data in each non-trend term corresponding to the load time series data of the first set time period in the past of the prediction power grid, and the noise data is marked. And updating the marked non-trend items corresponding to the load time sequence data of the power grid to be predicted in the past first set time period, namely replacing the marked non-trend items by the average values at the end points of the interested time period corresponding to the marked non-trend items, so that the non-trend items corresponding to the load time sequence data of the power grid to be predicted in the past first set time period are smoother, and the non-trend items corresponding to the load time sequence data of the power grid to be predicted in the past first set time period after updating are obtained.
It should be noted that, by updating each non-trend item corresponding to the load time sequence data of the past first set time period of the power grid to be predicted, an unexpected situation of the power grid to be detected in the prediction process is avoided, and the accuracy of the predicted load time sequence data of the future second set time period of the power grid to be predicted is effectively improved.
(5) Determining predicted load time sequence data of a future second set time period of the power grid to be predicted according to each non-trend item corresponding to the load time sequence data of the past first set time period of the power grid to be predicted after updating and each trend item corresponding to the load time sequence data of the past first set time period of the power grid to be predicted, wherein the steps comprise:
and (5-1) inputting each non-trend term corresponding to the load time sequence data of the updated to-be-predicted power grid in the past first set time period into a Gaussian process regression function, so as to obtain a predicted non-trend term of the to-be-predicted power grid in the future second set time period.
In this embodiment, a sample set is constructed according to each non-trend item corresponding to the updated load time series data of the past first set time period of the power grid to be predicted
Figure DEST_PATH_IMAGE047
Wherein, in the step (A),
Figure 129745DEST_PATH_IMAGE048
is the time-series date of the time-series date,
Figure DEST_PATH_IMAGE049
for non-trend terms, a Gaussian regression function is used to deduce the functional relationship of the non-trend terms
Figure 524823DEST_PATH_IMAGE050
So as to obtain a predicted non-trend item of a future second set time period of the power grid to be predicted
Figure DEST_PATH_IMAGE051
。
It should be noted that, in the following description,
Figure 103703DEST_PATH_IMAGE052
and
Figure 6544DEST_PATH_IMAGE049
the functional relationship of (a) can be described as:
Figure DEST_PATH_IMAGE053
wherein, in the step (A),
Figure 416797DEST_PATH_IMAGE054
is a noise disturbance of a gaussian distribution,
Figure 186038DEST_PATH_IMAGE054
has a mean value of 0 and a variance of
Figure DEST_PATH_IMAGE055
. The process of using gaussian regression functions to determine predictive non-trending terms is prior art and is not within the scope of the present invention and will not be described in detail herein.
And (5-2) determining a predicted trend item of a future second set time period of the power grid to be predicted according to trend items corresponding to load time sequence data of the past first set time period of the power grid to be predicted.
In the embodiment, trend items corresponding to load time series data of a past first set time period of the power grid to be predicted, which are obtained in the step (1), are obtained
Figure 314531DEST_PATH_IMAGE056
Determining an optimal parameter vector
Figure DEST_PATH_IMAGE057
And further determining a prediction trend term of a future second set time period of the power grid to be predicted, wherein the calculation formula is as follows:
Figure DEST_PATH_IMAGE059
wherein the content of the first and second substances,
Figure 899359DEST_PATH_IMAGE060
for the first in a second future set period of time of the grid to be predictedwThe item of the predicted trend of the day,
Figure 695276DEST_PATH_IMAGE031
in order to optimize the parameter vector,Nthe number of years corresponding to the first set time period in the past,Nthe magnitude of the numerical value can be determined by an implementer according to the specific load time sequence data acquisition condition. It should be noted that the process of determining the optimal parameter vector is prior art and is not within the scope of the present invention, and is not described in detail herein.
And (5-3) determining the predicted load time sequence data of the future second set time period of the power grid to be predicted according to the predicted non-trend item and the predicted trend item of the future second set time period of the power grid to be predicted.
In this embodiment, the predicted load time series data of the future second set time period of the power grid to be predicted is calculated by the predicted non-trend term of the future second set time period of the power grid to be predicted obtained in the step (5-1) and the predicted trend term of the future second set time period of the power grid to be predicted obtained in the step (5-2), and the calculation formula is as follows:
Figure 697736DEST_PATH_IMAGE062
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE063
for the predicted load time series data of the future second set time period of the power grid to be predicted,
Figure 313525DEST_PATH_IMAGE064
for the forecast trend item of the future second set time period of the power grid to be forecasted,
Figure 544787DEST_PATH_IMAGE051
and the predicted non-trend item of the future second set time period of the power grid to be predicted is obtained.
Therefore, the embodiment obtains the future second set time period of the power grid to be predicteduPredicted load time series data of a day, which may be expressed as
Figure DEST_PATH_IMAGE065
。
According to the method, through the load time sequence data development rule of the power grid to be predicted in the past first set time period, the noise data in each non-trend item corresponding to the load time sequence data of the power grid to be predicted in the past first set time period are updated, and the predicted load time sequence data of the power grid to be predicted in the future second set time period are determined according to the predicted non-trend item and the predicted trend item of the power grid to be predicted in the past first set time period, so that the accuracy of the predicted load time sequence data is effectively improved.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should 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; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.

Claims (10)

1. A smart grid load prediction method based on signal denoising is characterized by comprising the following steps:
acquiring load time sequence data of a to-be-predicted power grid in a first set time period in the past, and determining trend items and non-trend items corresponding to the load time sequence data of the to-be-predicted power grid in the first set time period in the past according to the load time sequence data of the to-be-predicted power grid in the first set time period in the past;
acquiring each specific interesting time period and each random interesting time period corresponding to load time sequence data of a first set time period in the past of a power grid to be predicted;
determining noise degree index values and noise thresholds corresponding to non-trend items in each specific interesting time period and each random interesting time period according to the non-trend items in each specific interesting time period and each random interesting time period corresponding to load time sequence data of a first set time period in the past of a power grid to be predicted;
screening out noise data in each non-trend item corresponding to load time sequence data of a first set time period in the past of the power grid to be predicted according to noise degree index values and noise thresholds corresponding to the non-trend items in each specific interested time period and each random interested time period, and updating the screened noise data to obtain each non-trend item corresponding to the load time sequence data of the first set time period in the past of the power grid to be predicted after updating;
and determining predicted load time sequence data of a future second set time period of the power grid to be predicted according to the updated non-trend items corresponding to the load time sequence data of the past first set time period of the power grid to be predicted and the trend items corresponding to the load time sequence data of the past first set time period of the power grid to be predicted.
2. The method for predicting the load of the smart grid based on signal denoising as claimed in claim 1, wherein the step of determining the noise degree index value and the noise threshold corresponding to each non-trend term in each specific interested period and each random interested period comprises:
determining a plurality of related non-trend items corresponding to the non-trend items in each specific interested period and each random interested period according to the non-trend items in each specific interested period and each random interested period corresponding to the load time series data of the past first set time period of the power grid to be predicted, wherein the related non-trend items are the non-trend items adjacent to the non-trend items in the specific interested period or the random interested period;
determining noise degree index values and power grid load level indexes corresponding to the non-trend items in each specific interesting period and each random interesting period according to the non-trend items in each specific interesting period and each random interesting period and a plurality of relevant non-trend items corresponding to the non-trend items;
determining specific interested period noise threshold components of the non-trend items in the specific interested periods according to the time sequence date of the non-trend items in the specific interested periods and the time sequence date at the end point of the specific interested periods;
determining a noise threshold corresponding to each non-trend item in each specific interested period according to the specific interested period noise threshold component and the power grid load level index of each non-trend item in each specific interested period;
determining random interested period noise threshold components of the non-trend items in the random interested periods according to the time sequence date of the non-trend items in the random interested periods;
and determining the noise threshold corresponding to each non-trend item in each random interesting period according to the random interesting period noise threshold component of each non-trend item in each random interesting period and the power grid load level index.
3. The method for predicting the load of the smart grid based on signal denoising as claimed in claim 2, wherein the step of determining the noise degree index value corresponding to each non-trend term in each specific interested time period and each random interested time period comprises:
determining the absolute value of the difference between each non-trend item in each specific interested time period and each random interested time period and the corresponding multiple related non-trend items according to the multiple related non-trend items corresponding to the non-trend items in each specific interested time period and each random interested time period;
screening a plurality of target difference absolute values from the absolute values of the differences between each non-trend item and a plurality of relevant non-trend items corresponding to the non-trend item in each specific interesting period and each random interesting period, wherein the target difference absolute value is the smaller difference absolute value in the absolute values of the differences between each non-trend item and a plurality of relevant non-trend items corresponding to the non-trend item;
and calculating the average value of the absolute values of the plurality of target difference values corresponding to the non-trend items in each specific interesting period and each random interesting period, and taking the average value as the noise degree index value corresponding to the corresponding non-trend items in each specific interesting period and each random interesting period.
4. The smart grid load prediction method based on signal denoising as claimed in claim 2, wherein the step of determining the period-of-interest-specific noise threshold component of each non-trend term in each period of specific interest includes:
determining the closeness degree of each non-trend item in each specific interested period and the corresponding specific interested period endpoint according to the time sequence date of each non-trend item in each specific interested period and the time sequence date of each specific interested period endpoint;
and determining the specific interested period noise threshold component of each non-trend item in each specific interested period according to the proximity of each non-trend item in each specific interested period to the corresponding specific interested period endpoint.
5. The smart grid load prediction method based on signal denoising as claimed in claim 4, wherein the calculation formula for determining the proximity degree of each non-trend term in each specific interesting period and the corresponding specific interesting period endpoint is as follows:
Figure DEST_PATH_IMAGE002
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE004
for the first in each particular time period of interestkThe non-trend item and the corresponding specific feelingThe proximity of the end points of the time period of interest,ifor the first in each particular time period of interestkThe chronological date of the individual non-trending items,
Figure DEST_PATH_IMAGE006
for the first in each particular time period of interestkThe minimum chronological date at the end of a particular time period of interest for an individual non-trending item,
Figure DEST_PATH_IMAGE008
for the first in each particular time period of interestkThe maximum chronological date at the end of a particular time period of interest for each non-trending item, min () is the minimum function.
6. The smart grid load prediction method based on signal denoising as claimed in claim 4, wherein a calculation formula for determining a specific interesting period noise threshold component of each non-trend term in each specific interesting period is:
Figure DEST_PATH_IMAGE010
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE012
for the first in each particular time period of interestkThe period of particular interest noise threshold component of each non-trending term,
Figure 900311DEST_PATH_IMAGE004
for the first in each particular time period of interestkThe proximity of each non-trending term to its corresponding endpoint for a particular time period of interest.
7. The method for smart grid load prediction based on signal denoising as claimed in claim 2, wherein the step of determining the random interested period noise threshold component of each non-trend term in each random interested period comprises:
determining a quadratic function corresponding to each non-trend item in each random interested time period according to each non-trend item in each random interested time period and the time sequence date of each non-trend item;
determining an extremum time sequence date corresponding to each non-trend item in each random interested time period according to the quadratic function corresponding to each non-trend item in each random interested time period;
and determining the random interested period noise threshold component of each non-trend item in each random interested period according to the time sequence date of each non-trend item in each random interested period and the extreme value time sequence date corresponding to each non-trend item.
8. The smart grid load prediction method based on signal denoising as claimed in claim 7, wherein a calculation formula for determining the random interesting period noise threshold component of each non-trend term in each random interesting period is:
Figure DEST_PATH_IMAGE014
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE016
for the first in each random period of interestkThe random period of interest noise threshold component of each non-trending term,ifor the first in each particular time period of interestkThe chronological date of the individual non-trending items,i * for the first in each random period of interestkAnd the extreme value time sequence date corresponding to each non-trend item.
9. The method for predicting the load of the smart grid based on signal denoising as claimed in claim 2, wherein the calculation formula for determining the noise threshold corresponding to each non-trend term in each specific interested time interval is as follows:
Figure DEST_PATH_IMAGE018
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE020
for the first in each particular time period of interestkThe noise threshold corresponding to each of the non-trend terms,
Figure DEST_PATH_IMAGE022
andnin order to be a hyper-parameter,
Figure DEST_PATH_IMAGE024
for the first in each particular time period of interestkThe grid load level indicator of each non-trend term,
Figure DEST_PATH_IMAGE026
is the in-domain factor for a particular time period of interest,
Figure 96412DEST_PATH_IMAGE012
for the first in each particular time period of interestkA period-of-interest-specific noise threshold component of the individual non-trending terms;
the calculation formula for determining the noise threshold corresponding to each non-trend term in each random interesting period is as follows:
Figure DEST_PATH_IMAGE028
wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE030
for the first in each random period of interestkThe noise threshold corresponding to each of the non-trend terms,
Figure 767827DEST_PATH_IMAGE022
andnin order to be a hyper-parameter,
Figure DEST_PATH_IMAGE032
for the first in each random period of interestkThe grid load level indicator of each non-trend term,
Figure DEST_PATH_IMAGE034
is the intra-domain factor of the random period of interest,
Figure 835140DEST_PATH_IMAGE016
for the first in each random interesting periodkRandom periods of interest noise threshold components of the individual non-trending terms.
10. The intelligent power grid load prediction method based on signal denoising as claimed in claim 1, wherein the step of determining predicted load time series data of a future second set time period of the power grid to be predicted comprises:
inputting each non-trend item corresponding to the load time sequence data of the updated power grid to be predicted in the past first set time period into a Gaussian process regression function, so as to obtain a predicted non-trend item of the power grid to be predicted in the future second set time period;
determining a predicted trend item of a future second set time period of the power grid to be predicted according to trend items corresponding to load time sequence data of the past first set time period of the power grid to be predicted;
and determining predicted load time sequence data of the future second set time period of the power grid to be predicted according to the predicted non-trend item and the predicted trend item of the future second set time period of the power grid to be predicted.
CN202211038120.1A 2022-08-29 2022-08-29 Smart power grid load prediction method based on signal denoising Active CN115099544B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211038120.1A CN115099544B (en) 2022-08-29 2022-08-29 Smart power grid load prediction method based on signal denoising

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211038120.1A CN115099544B (en) 2022-08-29 2022-08-29 Smart power grid load prediction method based on signal denoising

Publications (2)

Publication Number Publication Date
CN115099544A true CN115099544A (en) 2022-09-23
CN115099544B CN115099544B (en) 2022-10-25

Family

ID=83301678

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211038120.1A Active CN115099544B (en) 2022-08-29 2022-08-29 Smart power grid load prediction method based on signal denoising

Country Status (1)

Country Link
CN (1) CN115099544B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116400870A (en) * 2023-06-08 2023-07-07 西安品盛互联网技术有限公司 On-site construction on-line management system based on Internet of things

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104268660A (en) * 2014-10-13 2015-01-07 国家电网公司 Trend recognition method for electric power system predication-like data
CN109345027A (en) * 2018-10-25 2019-02-15 国网江苏省电力有限公司盐城供电分公司 Micro-capacitance sensor short-term load forecasting method based on independent component analysis and support vector machines
CN111080489A (en) * 2020-01-23 2020-04-28 广东工业大学 Load trend prediction method, system and equipment
CN113839845A (en) * 2021-09-18 2021-12-24 南方电网科学研究院有限责任公司 Secret distributed optimal scheduling method, system, computer equipment and medium

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104268660A (en) * 2014-10-13 2015-01-07 国家电网公司 Trend recognition method for electric power system predication-like data
CN109345027A (en) * 2018-10-25 2019-02-15 国网江苏省电力有限公司盐城供电分公司 Micro-capacitance sensor short-term load forecasting method based on independent component analysis and support vector machines
CN111080489A (en) * 2020-01-23 2020-04-28 广东工业大学 Load trend prediction method, system and equipment
CN113839845A (en) * 2021-09-18 2021-12-24 南方电网科学研究院有限责任公司 Secret distributed optimal scheduling method, system, computer equipment and medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116400870A (en) * 2023-06-08 2023-07-07 西安品盛互联网技术有限公司 On-site construction on-line management system based on Internet of things
CN116400870B (en) * 2023-06-08 2023-08-18 西安品盛互联网技术有限公司 On-site construction on-line management system based on Internet of things

Also Published As

Publication number Publication date
CN115099544B (en) 2022-10-25

Similar Documents

Publication Publication Date Title
CN109919353B (en) Distributed photovoltaic prediction method of ARIMA model based on spatial correlation
Dong et al. Wind power day-ahead prediction with cluster analysis of NWP
CN108009938B (en) System load clustering and load period pattern recognition method based on shape
Yildiz et al. Household electricity load forecasting using historical smart meter data with clustering and classification techniques
CN105117810A (en) Residential electricity consumption mid-term load prediction method under multistep electricity price mechanism
CN105260803A (en) Power consumption prediction method for system
CN110097220B (en) Method for predicting monthly electric quantity of wind power generation
CN111191193A (en) Long-term soil temperature and humidity high-precision prediction method based on autoregressive moving average model
CN111027775A (en) Step hydropower station generating capacity prediction method based on long-term and short-term memory network
CN115099544B (en) Smart power grid load prediction method based on signal denoising
Fan et al. Comparative study on load forecasting technologies for different geographical distributed loads
CN112347652B (en) Heavy rain high risk division method based on linear moment frequency analysis of hydrological region
Wang et al. Uncertainty in SPI calculation and its impact on drought assessment in different climate regions over China
CN111626006A (en) Quantitative identification method for multivariate evaluation on runoff change sensitivity
CN115660225A (en) Electricity load prediction management method and system based on ammeter communication module
Lourenço et al. Short term load forecasting using Gaussian process models
Kou et al. Prediction intervals for wind power forecasting: using sparse warped Gaussian process
Wang et al. Multiscale dynamic correlation analysis of wind-PV power station output based on TDIC
Grigoraş et al. Use of data from smart meters in optimal operation of distribution systems
CN116470491A (en) Photovoltaic power probability prediction method and system based on copula function
CN111414354A (en) Wind power plant data processing method and device and computer readable storage medium
CN115456440A (en) Method, device and system for evaluating power generation characteristics of distributed photovoltaic power station
Dan et al. Short-Term Forecasting of Wind Power Generation
CN101923605B (en) Wind pre-warning method for railway disaster prevention
Das et al. Estimation of long range correlations and FARIMA modelling of wind speed in Maharashtra

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
GR01 Patent grant
GR01 Patent grant