CN113326985B - Short-term load prediction method and device - Google Patents
Short-term load prediction method and device Download PDFInfo
- Publication number
- CN113326985B CN113326985B CN202110604166.4A CN202110604166A CN113326985B CN 113326985 B CN113326985 B CN 113326985B CN 202110604166 A CN202110604166 A CN 202110604166A CN 113326985 B CN113326985 B CN 113326985B
- Authority
- CN
- China
- Prior art keywords
- historical
- load data
- holiday
- load
- current
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000036316 preload Effects 0.000 claims abstract description 5
- 239000013598 vector Substances 0.000 claims description 33
- 238000012546 transfer Methods 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000009499 grossing Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 abstract description 5
- 230000002354 daily effect Effects 0.000 description 49
- 230000008569 process Effects 0.000 description 8
- 230000008859 change Effects 0.000 description 6
- 238000012216 screening Methods 0.000 description 5
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 238000012795 verification 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2462—Approximate or statistical queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Software Systems (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Probability & Statistics with Applications (AREA)
- Tourism & Hospitality (AREA)
- General Engineering & Computer Science (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Development Economics (AREA)
- Water Supply & Treatment (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Fuzzy Systems (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Computational Linguistics (AREA)
- Entrepreneurship & Innovation (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses a short-term load prediction method and a short-term load prediction device, which are used for solving the technical problems that prediction indexes selected by the existing load prediction technology are not representative, and the prediction means is lack of pertinence, so that the deviation of a prediction result is large. The method specifically comprises the following steps: acquiring short-term date attributes, wherein the short-term date attributes comprise a time attribute and a week attribute; if the current day is judged to be a holiday according to the short-term date attribute, predicting the current holiday load data according to the historical associated day load data, the historical holiday load data and the current holiday associated day load data based on a holiday preload similarity principle; and if the current day is judged to be a non-holiday according to the short-term date attribute, predicting the load data of the current day by adopting a preset step length prediction model according to the historical associated time point temperature and the historical associated time point load data.
Description
Technical Field
The present application relates to the field of load prediction technologies, and in particular, to a short-term load prediction method and apparatus.
Background
The method is basically assumed to be two days with similar influence factors such as external meteorological conditions and the like, the load characteristics and the curve shape are approximately similar, and the holiday prediction curve is obtained by selecting the meteorological similar day as a reference based on the assumption. Non-holidays generally assume that the load has a periodic change rule and a trend is stable, and are usually predicted by adopting a prediction method such as a support vector machine.
However, the existing short-term load prediction schemes have some technical problems, for example, for a specific holiday, the current holiday prediction is performed by directly referring to historical holidays of the same type, and in addition, the load fluctuation is influenced by various factors, and the load characteristics of similar days selected only according to weather sometimes do not have high similarity. These technical problems directly result in large load prediction deviation, so that the accuracy of the prediction result is low.
Disclosure of Invention
The application provides a short-term load prediction method and device, which are used for solving the technical problems that prediction indexes selected by the existing load prediction technology are not representative, and the prediction means is lack of pertinence, so that the deviation of a prediction result is large.
In view of the above, a first aspect of the present application provides a short-term load prediction method, including:
acquiring short-term date attributes, wherein the short-term date attributes comprise a time attribute and a week attribute;
if the current day is judged to be a holiday according to the short-term date attribute, predicting the current holiday load data according to the historical associated day load data, the historical holiday load data and the current holiday associated day load data based on a holiday before-load similarity principle;
and if the current day is judged to be a non-holiday according to the short-term date attribute, predicting the load data of the current day by adopting a preset step prediction model according to the historical associated time point temperature and the historical associated time point load data.
Preferably, if it is determined that the current day is a holiday according to the short-term date attribute, predicting the current holiday load data according to the historical associated day load data, the historical holiday load data, and the current holiday associated day load data based on a holiday association rule, includes:
if the current day is judged to be a holiday according to the short-term date attribute, acquiring historical associated daily load data and historical holiday load data, corresponding to historical holidays of the same type as the current day, of a plurality of historical years to obtain a historical load data set, wherein the historical associated daily load data are load data of different moments of preset days before the historical holiday;
according to a preset historical power transferring information table, historical associated daily load data and historical holiday load data corresponding to historical years of power transferring operation are removed, and a screened historical load data set is obtained;
if the screened historical load data set is a non-empty set, calculating annual load similarity according to the historical associated daily load data and the current holiday associated daily load data, and keeping the historical associated daily load data and the historical holiday load data of the year corresponding to the maximum annual load similarity;
calculating a historical time smooth value and a current time smooth value corresponding to different times according to the historical associated daily load data and the current holiday associated daily load data of the year corresponding to the maximum annual load similarity;
and predicting the current holiday load data through the historical time smooth value, the historical holiday load data and the current time smooth value.
Preferably, if the filtered historical load data set is a non-empty set, calculating annual load similarity according to the historical associated daily load data and the current holiday associated daily load data, and retaining the historical associated daily load data and the historical holiday load data of a year corresponding to the maximum annual load similarity, the method further includes:
and if the screened historical load data set is an empty set, acquiring historical holiday load data of the recent historical year at each moment of the historical holiday with the type different from that of the current day, and performing mean calculation to obtain the current holiday load data.
Preferably, the obtaining historical associated daily load data and historical holiday load data corresponding to historical holidays of a plurality of historical years and the current day type is the same to obtain a historical load data set, and then further includes:
and performing holiday alignment operation on the historical associated day load data and the historical holiday load data in the historical load data set according to the holiday type of the current day.
Preferably, if the current day is judged to be a non-holiday according to the short-term date attribute, predicting the load data of the current day according to the history associated time point temperature and the history associated time point load data by using a preset step prediction model, including:
if the current day is judged to be a non-holiday according to the short-term date attribute, acquiring historical association time point temperature and historical association time point load data in a preset time period before the current day;
calculating the correlation between the temperature and the load according to the historical associated time point temperature and the historical associated time point load data;
if the correlation is larger than a correlation threshold value, constructing a preset time sequence feature vector according to the historical correlation time point temperature and the historical correlation time point load data;
judging whether the load has a preset load fluctuation condition according to the historical associated time point load data, if so, adopting a preset short step prediction model to predict the load according to the preset time sequence characteristic vector to obtain the current day load data;
if not, a preset long-step prediction model is adopted to carry out load prediction according to the preset time sequence characteristic vector to obtain the current daily load data;
the preset step length prediction model comprises a preset short step length prediction model and a preset long step length prediction model.
Preferably, if the correlation is greater than a correlation threshold, constructing a preset time series feature vector according to the historical associated time point temperature and the historical associated time point load data, further including:
and if the correlation is smaller than the correlation threshold value, constructing the preset time sequence feature vector according to the load data of the historical associated time point.
Preferably, if the current day is judged to be a non-holiday according to the short-term date attribute, acquiring a history associated time point temperature and history associated time point load data in a preset time period before the current day, and then further comprising:
and performing preset data processing on the temperature at the historical associated time point and the load data at the historical associated time point, wherein the preset data processing comprises data filling, data replacement and data association establishment.
A second aspect of the present application provides a short-term load prediction apparatus, including:
the date acquisition module is used for acquiring short-term date attributes, wherein the short-term date attributes comprise a time attribute and a week attribute;
a holiday prediction module for predicting the current holiday load data according to the historical associated day load data, the historical holiday load data and the current holiday associated day load data based on a holiday before load similarity principle if the current day is judged to be a holiday according to the short-term date attribute;
and the non-holiday prediction module is used for adopting a preset step length prediction model to predict the load data of the current day according to the historical associated time point temperature and the historical associated time point load data if the current day is judged to be a non-holiday according to the short-term date attribute.
Preferably, the holiday prediction module comprises:
the first load obtaining submodule is used for obtaining historical associated day load data and historical holiday load data, corresponding to historical holidays of the same type as the current day, of a plurality of historical years to obtain a historical load data set if the current day is judged to be a holiday according to the short-term date attribute, and the historical associated day load data are load data of preset days before the historical holiday at different moments;
the load removing submodule is used for removing the historical associated daily load data and the historical holiday load data corresponding to the historical years of the power transferring operation according to a preset historical power transferring information table to obtain a screened historical load data set;
a first similarity calculation submodule, configured to calculate annual load similarity according to the historical associated daily load data and the current holiday associated daily load data if the filtered historical load data set is a non-empty set, and retain the historical associated daily load data and the historical holiday load data of a year corresponding to a maximum annual load similarity;
a smoothing value operator module, configured to calculate, according to the historical associated daily load data and the current holiday associated daily load data of the year corresponding to the maximum annual load similarity, a historical time smoothing value and a current time smoothing value corresponding to different times;
and the holiday prediction submodule is used for predicting the current holiday load data through the historical time smooth value, the historical holiday load data and the current time smooth value.
Preferably, the non-holiday prediction module comprises:
the second load obtaining submodule is used for obtaining the historical association time point temperature and the historical association time point load data in the preset time period before the current day if the current day is judged to be a non-holiday according to the short-term date attribute;
the second similarity calculation submodule is used for calculating the correlation between the temperature and the load according to the historical association time point temperature and the historical association time point load data;
the characteristic vector construction submodule is used for constructing a preset time sequence characteristic vector according to the historical association time point temperature and the historical association time point load data if the correlation is larger than a correlation threshold value;
the first non-holiday prediction submodule is used for judging whether a preset load fluctuation condition exists in the load according to the historical associated time point load data, if yes, a preset short step length prediction model is adopted to carry out load prediction according to the preset time sequence characteristic vector, and current day load data are obtained;
the second non-holiday prediction submodule is used for performing load prediction according to the preset time sequence characteristic vector by adopting a preset long-step prediction model if the current holiday load data is not the current holiday load data;
the preset step length prediction model comprises a preset short step length prediction model and a preset long step length prediction model.
According to the technical scheme, the embodiment of the application has the following advantages:
in the present application, a short-term load prediction method is provided, including: acquiring short-term date attributes, wherein the short-term date attributes comprise a time attribute and a week attribute; if the current day is judged to be a holiday according to the short-term date attribute, predicting the current holiday load data according to the historical associated day load data, the historical holiday load data and the current holiday associated day load data based on a holiday preload similarity principle; and if the current day is judged to be a non-holiday according to the short-term date attribute, predicting the load data of the current day by adopting a preset step length prediction model according to the historical associated time point temperature and the historical associated time point load data.
According to the short-term load prediction method, attribute division is carried out on the current day through the acquired short-term date attribute information, different load prediction schemes are selected according to different date attributes of the current day, specifically, historical holiday load data are considered in the holiday load prediction schemes, historical holiday load data and current holiday associated load data are used in the current holiday load data prediction process, and the load data are adopted for prediction and need to be based on a holiday before load similarity principle, namely the selected load data and the current holiday have certain similarity, the reliability of the prediction result cannot be influenced, and therefore the accuracy of the load prediction result is improved. In the scheme of predicting the load in the holiday instead, the preset step prediction model which can be set is adopted, and the load prediction is carried out by combining the relation between the temperature and the load, so that the prediction precision of the model can be guaranteed. Therefore, the method and the device can solve the technical problems that the prediction indexes selected by the existing load prediction technology are not representative, the prediction means is lack of pertinence, and the prediction result deviation is large.
Drawings
Fig. 1 is a schematic flowchart of a short-term load prediction method according to an embodiment of the present disclosure;
fig. 2 is another schematic flow chart of a short-term load prediction method according to an embodiment of the present disclosure;
fig. 3 is a schematic structural diagram of a short-term load prediction apparatus according to an embodiment of the present disclosure.
Detailed Description
In order to make those skilled in the art better understand the technical solutions of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present application without making any creative effort belong to the protection scope of the present application.
For easy understanding, referring to fig. 1, a first embodiment of a short-term load prediction method provided in the present application includes:
The short-term date attribute includes not only the time attribute of a specific year, month, day, time, etc., but also important week attribute, that is, the date and day of the week, so that an accurate attribute judgment can be made on the current day according to the short-term date attribute, for example, whether the current day is a holiday, a holiday name, a year, month, etc. to which the holiday belongs.
And 102, if the current day is judged to be a holiday according to the short-term date attribute, predicting the current holiday load data according to the historical associated day load data, the historical holiday load data and the current holiday associated day load data based on a holiday before load similarity principle.
The set list can be directly formed for each holiday of a year, and can be a holiday list, the list is closely related to time, and the attribute judgment result of the current day can be obtained by comparing the short-term date attribute with the holiday list. The specific holiday category comprises 7 legal holidays such as New year, spring festival, qingming festival, early noon festival, labor festival, mid-autumn festival, national day festival and the like, holiday setting can be added according to needs, and the holiday category can be set as a working day or a weekend.
The load similarity principle before holiday is set according to the characteristics of the acquired load data, and the acquired load data not only comprise the load data of historical holidays, but also comprise historical associated daily load data and current holiday associated daily load data; therefore, the selected historical associated daily load needs a certain similarity regulation with the current holiday associated daily load, and the load data of the selected historical associated daily load can be used for predicting the current daily load instead of the random historical associated daily load. The reliability of the prediction result can be ensured to the maximum extent by performing load prediction according to various closely related load data selected based on the section-false pre-load similarity principle.
The association day is a period of time close to the holiday, and the load fluctuation in the period of time has certain relevance with the holiday; the specific time length can be set according to the actual load fluctuation condition, and a preset day before holidays, such as 6 days, is generally selected; the historical associated day load data is load data of a preset day before the holiday in the historical year; the load data of the current holiday-associated day is the load data of a preset day before the current holiday of the current year, and the load data are generated and existed and can be obtained.
And 103, if the current day is judged to be a non-holiday according to the short-term date attribute, predicting the load data of the current day according to the historical associated time point temperature and the historical associated time point load data by adopting a preset step prediction model.
If the current day is a working day or weekend, the current day is defined as a non-holiday, the short-term load prediction mechanism is a prediction model at the moment, the prediction model can be used for adaptively setting step length, automatically sensing the change of the load and improving the robustness of model prediction. The selected indexes are historical associated time point temperature and historical associated time point load data; the historical associated time point temperature is the temperature values of different time points of similar dates in the historical year, and the temperature values can be obtained by selecting the time points; the load data at the historical association time point is associated with the acquiring process of the temperature at the historical association time point, and the load data and the temperature at the historical association time point are in one-to-one correspondence, namely, each time point can acquire one temperature value and one corresponding load data.
The short-term load prediction method provided by the embodiment of the application includes the steps that attribute division is carried out on the current day through acquired short-term date attribute information, different load prediction schemes are selected according to different date attributes of the current day, specifically, historical holiday load data are considered in the holiday load prediction schemes, historical holiday load data and current holiday associated load data are used in the current holiday load data prediction process, and the load data are adopted for prediction and need to be based on a holiday before-load similarity principle, namely the selected load data and the current holiday have certain similarity, the reliability of prediction results cannot be affected, and therefore the accuracy of the load prediction results is improved. In the scheme for predicting the loads in the holidays, the preset step length prediction model which can be set is adopted, the load prediction is carried out by combining the relation between the temperature and the load, and the prediction accuracy of the model can be guaranteed. Therefore, the method and the device for predicting the load can solve the technical problems that the prediction indexes selected by the existing load prediction technology are not representative, the prediction means is lack of pertinence, and the prediction result deviation is large.
The above is an embodiment of a short-term load prediction method provided by the present application, and the following is another embodiment of a short-term load prediction method provided by the present application.
For easy understanding, referring to fig. 2, the present application provides a second embodiment of a short-term load prediction method, including:
Acquiring date attributes of the year, month, day, hour, minute, second and day of week of the current day; meanwhile, the preset day-time attribute and the preset week attribute before the current day are also obtained, so that the subsequent time analysis is facilitated.
The historical associated daily load data is load data of different times of a preset day before the historical holidays. In this embodiment, the time 6 days before the current day is selected as the preset day. Therefore, the preset days before the historical holidays are the load data at different times 6 days before the historical holidays. The historical holiday load data is the load data of the historical holiday. The historical years may be selected to be the last three years, for example, the current day is the year 2021 year old day of the year new year, then the year history can obtain the metadata related load data of 2020, 2019 and 2018.
Further, step 202 is followed by:
and performing holiday alignment operation on the historical associated daily load data and the historical holiday load data in the historical load data set according to the holiday type of the current day.
And aligning the acquired historical holidays and historical associated days of the historical years according to the current holiday, taking the current day as a 2021 year New year day example, and aligning by taking the public 1 month and 1 day of the historical years as a reference, so that the 2021 year New year date is 12 months and 30 days, 12 months and 31 days, and 2021 month and 1 day of the 2021 year. Date alignment facilitates subsequent date-based joint analysis of load data.
The preset historical power transfer information table records the ID of the main transformer, the name of the main transformer, the power transfer starting time and the power transfer ending time. In the prior art, the situation that the load of the power supply is changed suddenly is not taken into deep consideration, but the load prediction result is directly influenced by sudden change or fluctuation of the load, so that the relevant load data of the historical years with the power supply switching operation is removed in the embodiment, and the influence of the relevant load data on the load prediction result of the current day is avoided. The process is a process of screening the historical load data set, and the accuracy of the prediction result is ensured by controlling the reliability of the index data.
The removing process comprises the following steps: acquiring a main transformer ID from a preset historical power transfer information table, searching power transfer data based on the main transformer ID, if a historical holiday period is contained in the power transfer data, indicating that power transfer operation occurs during the holiday of the year, and intensively removing the year data from historical load data.
And 204, if the screened historical load data set is a non-empty set, calculating annual load similarity according to the historical associated daily load data and the current holiday associated daily load data, and keeping the historical associated daily load data and the historical holiday load data of the year corresponding to the maximum annual load similarity.
If there is any historical load data in the screening historical load data set, the load similarity calculation for the associated day may be started, specifically, D may be defined as the screening historical load data set, and is expressed as D = { (D) 1 ,h 1 ),(d 2 ,h 2 ),(d 3 ,h 3 ) A few, wherein (d) i ,h i ) Load data of the i-th year, d i Associated daily load data, h, representing 6 days before the historical holiday i The data are load data of the current day of historical holidays, and the data are all composed of load data of a plurality of moments of each day. Definition (d) n ,h' n ) Load data for the current year, wherein d n Is the associated daily load data, h 'of the preset day before the current day' n Is the current holiday load data to be predicted. The similarity of the annual load calculated by calculating the historical associated daily load data and the current holiday associated daily load data can be expressed as follows:
where k is the load data length and abs (·) is the absolute value function.
Every year, a similarity value s can be calculated i Finding the maximum annual load similarity, and retaining the historical associated daily load data and the historical holiday load data of the corresponding year, namely (d) s ,h s )。
Further, step 204 further includes:
and if the screened historical load data set is an empty set, acquiring historical holiday load data of historical holidays of the recent historical year at each moment, which are different from the current day in type, and performing mean calculation to obtain the current holiday load data.
And if the screening historical load data set is an empty set, the selected holiday and festival related load data of all the years are related to power supply transferring operation, and all the holiday and festival related load data are removed from the data set in the screening stage. The occurrence probability of the condition of lacking the referable historical data is relatively small, and once the condition occurs, the average value can be directly calculated according to the historical holiday load data of the recent historical year at each moment of the historical holiday with different types from the current day, and the average value is used as the current holiday load data.
And step 205, calculating a historical time smooth value and a current time smooth value corresponding to different times according to the historical associated daily load data and the current holiday associated daily load data of the year corresponding to the maximum annual load similarity.
Defining the load of t moment of j day in N related days selected before the current holiday of the year as d njt The load at the t moment of the j day of the N associated days selected before the holiday of the most similar year is d sjt Where larger j indicates closer to holidays on the time axis. Taking the load value at the predicted time t as an example, t has 96 times, t = {1,2,3.. 96}, and the smoothed values a at the time t of the relevant day before the current year and the most similar year are calculated respectively 1t ，A 2t ：
Wherein, alpha is a smoothing coefficient of point-by-point load, generally takes the value as alpha belongs to [0.1,0.9], and the optimal parameter value is found to be 0.618 through experimental verification.
And step 206, predicting the current holiday load data through the historical time smooth value, the historical holiday load data and the current time smooth value.
The specific prediction process can be expressed as:
wherein h is st Is a smoothed value A 1t And load data of the corresponding historical holiday at the time t on the day. Predicting the load value of each moment of the current holiday according to the formulaA load prediction curve of the current holiday can be obtained.
And step 207, if the current day is judged to be a non-holiday according to the short-term date attribute, acquiring the historical associated time point temperature and the historical associated time point load data in a preset time period before the current day.
The preset time period may be set according to actual conditions, such as 10 days in the previous month of the current day of the current year, or data of the previous days. The load fluctuation of non-holidays is generally smaller, so that more load data can be referred to, mainly the selection of data indexes and the optimization of models. In addition to temperature data and load data, week attributes may be obtained as needed.
Further, step 207 is followed by:
and performing preset data processing on the temperature at the historical associated time point and the load data at the historical associated time point, wherein the preset data processing comprises data filling, data replacement and data association establishment.
The acquired data quality may have differences due to various reasons, and the data is subjected to preset data processing operation, so that the data quality can be improved, and a prediction result is guaranteed. The data recording is missing due to the failure of the metering device and other reasons, for example, the data at 96 time points selected in one day has multiple missing data, and the missing data needs to be filled; aiming at the holiday periods contained in the historical data, filling data with non-holidays with same-week attributes to ensure that all the data are non-holiday data and the integrity of the time sequence is ensured; the outlier data points are smoothed and replaced.
In addition, data association is needed, and the normalization data including line numbers, dates, times, loads, temperature values and other multivariate data are associated together according to the dates and the time points to form a data wide table, so that extraction and processing are facilitated.
And step 208, calculating the correlation between the temperature and the load according to the historical associated time point temperature and the historical associated time point load data.
Recording temperature vector X = [ X ] at historical associated time point 1 ,x 2 ,...]And load vector Y = [ Y ] 1 ,y 2 ,...]Correlation of (c) corr:
where n is the total number of data points acquired.
And 209, if the correlation is greater than the correlation threshold, constructing a preset time sequence feature vector according to the historical associated time point temperature and the historical associated time point load data.
In this embodiment, it is defined that, if abs (corr) > 0.5, it is determined that there is a strong correlation between the temperature and the load, otherwise, it is determined that there is not a large correlation between the temperature and the load. If the correlation between the temperature and the load is strong, a preset time sequence feature vector is constructed according to the historical correlation time point temperature and the historical correlation time point load data, and besides the two index data, data such as historical temperature average values, historical day attributes, current day temperature values, current day attribute values and current day average temperature values can be added to construct a multi-channel preset time sequence feature vector.
Further, step 209 further includes:
and if the correlation is smaller than the correlation threshold value, constructing a preset time sequence feature vector according to the historical associated time point load data.
If the correlation between the temperature and the load is not large, temperature data in the index data, including historical temperature data and current day temperature data, needs to be removed, a preset time sequence feature vector is constructed according to historical associated time point load data, and a multi-channel preset time sequence feature vector can be constructed by adding the historical day attribute and the current day attribute. In this embodiment, 96 time points are selected every day, and 96 preset time sequence feature vectors can be generated according to the time points.
And step 210, judging whether the load has a preset load fluctuation condition according to the load data of the historical associated time points, if so, performing load prediction according to a preset time sequence characteristic vector by adopting a preset short-step prediction model to obtain the load data of the current day.
By analyzing the change condition of the load data at the historical associated time point, whether the load has large fluctuation or not is judged, namely the preset load fluctuation condition is judged, an upper threshold value and a lower threshold value can be set, and the out-of-limit value is judged to have large fluctuation, which indicates that the load fluctuation is not stable and has sudden change. And at the moment, a preset short-step-length prediction model is adopted to process preset time sequence characteristic vectors, load prediction is carried out, and current daily load data are obtained. The step size selection for the preset short-step prediction model may be 2 days.
And step 211, if not, performing load prediction by using a preset long-step prediction model according to a preset time sequence characteristic vector to obtain current daily load data.
The preset step prediction model comprises a preset short step prediction model and a preset long step prediction model. The model can automatically sense the fluctuation situation of the load, and the prediction precision and robustness of the model are improved.
When the load sudden change amplitude is small, the load fluctuation is stable, the influence of the temperature accumulation effect and the periodicity on the load fluctuation is mainly considered, and accurate current daily load data can be predicted and obtained by adopting a preset long-step prediction model. The step size selection for the preset long-step prediction model may be 7 days.
The load fluctuation in the steady state has certain periodicity, but the periodicity in the non-steady state is broken, and the load fluctuation state cannot be sensed in time only by the periodicity law, so that prediction hysteresis of a certain degree is caused. Therefore, the preset step prediction model selected in the embodiment of the application can realize load prediction by automatically sensing the load fluctuation situation to alleviate the technical problem.
The above is an embodiment of a short-term load prediction method provided by the present application, and the following is an embodiment of a short-term load prediction apparatus provided by the present application.
For ease of understanding, referring to fig. 3, the present application further provides an embodiment of a short term load prediction apparatus, comprising:
a date acquisition module 301, configured to acquire a short-term date attribute, where the short-term date attribute includes a time attribute and a week attribute;
a holiday prediction module 302, configured to predict current holiday load data according to the historical associated day load data, the historical holiday load data, and the current holiday associated day load data based on a holiday preload similarity principle if the current day is judged to be a holiday according to the short-term date attribute;
and a non-holiday prediction module 303, configured to, if it is determined that the current day is a non-holiday according to the short-term date attribute, predict load data of the current day according to the history associated time point temperature and the history associated time point load data by using a preset step prediction model.
Further, the holiday prediction module 302 includes:
a first load obtaining submodule 3021, configured to, if it is determined that the current day is a holiday according to the short-term date attribute, obtain history associated day load data and history holiday load data corresponding to history holidays of the same type as the current day in a plurality of history years, to obtain a history load data set, where the history associated day load data is load data of different times of a preset day before the history holiday;
the load removing submodule 3022 is configured to remove historical associated daily load data and historical holiday load data corresponding to a historical year including a power transfer operation according to a preset historical power transfer information table, so as to obtain a filtered historical load data set;
a first similarity calculation submodule 3023, configured to calculate annual load similarity according to the historical associated daily load data and the current holiday associated daily load data if the screened historical load data set is a non-empty set, and retain historical associated daily load data and historical holiday load data of a year corresponding to the maximum annual load similarity;
a smoothing value operator module 3024, configured to calculate a smoothing value at a historical time and a smoothing value at a current time that correspond to different times according to the historical associated daily load data and the current holiday associated daily load data of the year corresponding to the maximum annual load similarity;
a holiday prediction sub-module 3025 for predicting the current holiday load data from the historical time smooth value, the historical holiday load data and the current time smooth value.
Further, the non-holiday prediction module 303 includes:
a second load obtaining submodule 3031, configured to obtain a history associated time point temperature and history associated time point load data in a preset time period before a current day if it is determined that the current day is a non-holiday according to the short-term date attribute;
the second similarity calculation submodule 3032 is configured to calculate the correlation between the temperature and the load according to the historical association time point temperature and the historical association time point load data;
the feature vector construction submodule 3033 is configured to construct a preset time sequence feature vector according to the historical associated time point temperature and the historical associated time point load data if the correlation is greater than the correlation threshold;
the first non-holiday prediction submodule 3034 is used for judging whether the load has a preset load fluctuation condition according to the load data of the historical association time point, if so, a preset short-step prediction model is adopted to carry out load prediction according to a preset time sequence eigenvector to obtain load data of the current day;
a second non-holiday prediction submodule 3035, configured to perform load prediction according to a preset time sequence feature vector by using a preset long-step prediction model if the current holiday load data is not the current holiday load data;
the preset step prediction model comprises a preset short step prediction model and a preset long step prediction model.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one type of logical functional division, and other divisions may be realized in practice, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, or portions or all or portions of the technical solutions that contribute to the prior art, may be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods described in the embodiments of the present application through a computer device (which may be a personal computer, a server, or a network device). And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above 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; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (6)
1. A method for short-term load prediction, comprising:
acquiring short-term date attributes, wherein the short-term date attributes comprise a time attribute and a week attribute;
the step of obtaining short-term date attributes includes:
acquiring date attributes of year, month, day, hour, minute, second and day of the current day, and acquiring a preset day time attribute and a preset week attribute before the current day;
if the current day is judged to be a holiday according to the short-term date attribute, predicting the current holiday load data according to the historical associated day load data, the historical holiday load data and the current holiday associated day load data based on a holiday before-load similarity principle;
if the current day is judged to be a holiday according to the short-term date attribute, predicting the current holiday load data according to the historical associated day load data, the historical holiday load data and the current holiday associated day load data based on a holiday association rule, wherein the step comprises the following steps:
if the current day is judged to be a holiday according to the short-term date attribute, acquiring historical associated daily load data and historical holiday load data, corresponding to historical holidays with the same type as the current day, of a plurality of historical years to obtain a historical load data set, wherein the historical associated daily load data are load data of different moments of a preset day before the historical holiday;
according to a preset historical power transfer information table, eliminating historical associated daily load data and historical holiday load data corresponding to historical years of power transfer operation to obtain a screened historical load data set;
if the screened historical load data set is a non-empty set, calculating annual load similarity according to the historical associated daily load data and the current holiday associated daily load data, and keeping the historical associated daily load data and the historical holiday load data of the year corresponding to the maximum annual load similarity;
calculating a historical time smooth value and a current time smooth value corresponding to different times according to the historical associated daily load data and the current holiday associated daily load data of the year corresponding to the maximum annual load similarity;
predicting current holiday load data through the historical time smooth value, the historical holiday load data and the current time smooth value;
if the current day is judged to be a non-holiday according to the short-term date attribute, a preset step prediction model is adopted to predict the load data of the current day according to the historical associated time point temperature and the historical associated time point load data;
if the current day is judged to be a non-holiday according to the short-term date attribute, a preset step length prediction model is adopted to predict the load data of the current day according to the historical associated time point temperature and the historical associated time point load data, and the step comprises the following steps:
if the current day is judged to be a non-holiday according to the short-term date attribute, acquiring historical association time point temperature and historical association time point load data in a preset time period before the current day;
calculating the correlation between the temperature and the load according to the historical associated time point temperature and the historical associated time point load data;
if the correlation is larger than a correlation threshold value, constructing a preset time sequence feature vector according to the historical correlation time point temperature and the historical correlation time point load data;
judging whether the load has a preset load fluctuation condition according to the historical associated time point load data, if so, adopting a preset short step prediction model to predict the load according to the preset time sequence characteristic vector to obtain the current day load data;
if not, a preset long-step prediction model is adopted to carry out load prediction according to the preset time sequence characteristic vector to obtain the current daily load data;
the preset step length prediction model comprises a preset short step length prediction model and a preset long step length prediction model.
2. The method of short term load prediction according to claim 1, wherein if the filtered historical load data set is a non-empty set, calculating annual load similarity from the historical associated daily load data and the current holiday associated daily load data, and retaining the historical associated daily load data and the historical holiday load data for a year corresponding to a maximum annual load similarity, further comprising:
and if the screened historical load data set is an empty set, acquiring historical holiday load data of the recent historical year at each moment of the historical holiday with the type different from that of the current day, and performing mean calculation to obtain the current holiday load data.
3. The short-term load prediction method according to claim 1, wherein the obtaining historical associated daily load data and historical holiday load data corresponding to historical holidays of the same type as the current day for a plurality of historical years to obtain a historical load data set further comprises:
and performing holiday alignment operation on the historical associated day load data and the historical holiday load data in the historical load data set according to the holiday type of the current day.
4. The method of claim 1, wherein if the correlation is greater than a correlation threshold, constructing a preset time series feature vector based on the historical associated time point temperature and the historical associated time point load data, further comprising:
and if the correlation is smaller than the correlation threshold value, constructing the preset time sequence feature vector according to the load data of the historical associated time point.
5. The method according to claim 1, wherein if the current day is judged to be a non-holiday according to the short-term date attribute, the method further comprises the following steps of obtaining historical associated time point temperature and historical associated time point load data in a preset time period before the current day:
and performing preset data processing on the temperature at the historical associated time point and the load data at the historical associated time point, wherein the preset data processing comprises data filling, data replacement and data association establishment.
6. A short-term load prediction apparatus, comprising:
the date acquisition module is used for acquiring short-term date attributes, wherein the short-term date attributes comprise a time attribute and a week attribute;
the step of obtaining short-term date attributes includes:
acquiring date attributes of year, month, day, hour, minute, second and day of the current day, and acquiring a preset day time attribute and a preset week attribute before the current day;
a holiday prediction module, configured to, if it is determined that the current day is a holiday according to the short-term date attribute, predict load data of the current holiday according to the historical associated day load data, the historical holiday load data, and the load data of the current holiday associated day based on a holiday preload similarity principle;
a non-holiday prediction module, configured to, if it is determined that the current day is a non-holiday according to the short-term date attribute, predict load data of the current day according to the history associated time point temperature and the history associated time point load data by using a preset step prediction model;
the holiday prediction module comprises:
a first load obtaining submodule, configured to obtain historical associated day load data and historical holiday load data corresponding to historical holidays of the same type as the current day in a plurality of historical years if it is determined that the current day is a holiday according to the short-term date attribute, to obtain a historical load data set, where the historical associated day load data are load data of different times of a preset day before the historical holiday;
the load removing submodule is used for removing the historical associated daily load data and the historical holiday load data corresponding to the historical years of the power transferring operation according to a preset historical power transferring information table to obtain a screened historical load data set;
a first similarity calculation submodule, configured to calculate annual load similarity according to the historical associated daily load data and the current holiday associated daily load data if the screened historical load data set is a non-empty set, and retain the historical associated daily load data and the historical holiday load data of a year corresponding to a maximum annual load similarity;
a smoothing value operator module used for calculating the smoothing values at the historical time and the current time corresponding to different times according to the historical associated daily load data and the current holiday associated daily load data of the year corresponding to the maximum annual load similarity;
the holiday prediction submodule is used for predicting current holiday load data through the historical time smooth value, the historical holiday load data and the current time smooth value;
the non-holiday prediction module comprises:
the second load obtaining submodule is used for obtaining the historical association time point temperature and the historical association time point load data in the preset time period before the current day if the current day is judged to be a non-holiday according to the short-term date attribute;
the second similarity calculation submodule is used for calculating the correlation between the temperature and the load according to the historical association time point temperature and the historical association time point load data;
the characteristic vector construction submodule is used for constructing a preset time sequence characteristic vector according to the historical association time point temperature and the historical association time point load data if the correlation is larger than a correlation threshold value;
the first non-holiday prediction submodule is used for judging whether a preset load fluctuation condition exists in the load according to the historical associated time point load data, if yes, a preset short step length prediction model is adopted to carry out load prediction according to the preset time sequence characteristic vector, and current day load data are obtained;
the second non-holiday prediction submodule is used for performing load prediction according to the preset time sequence characteristic vector by adopting a preset long-step prediction model if the current holiday load data is not the current holiday load data;
the preset step length prediction model comprises a preset short step length prediction model and a preset long step length prediction model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110604166.4A CN113326985B (en) | 2021-05-31 | 2021-05-31 | Short-term load prediction method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110604166.4A CN113326985B (en) | 2021-05-31 | 2021-05-31 | Short-term load prediction method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113326985A CN113326985A (en) | 2021-08-31 |
| CN113326985B true CN113326985B (en) | 2023-01-20 |
Family
ID=77422862
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110604166.4A Active CN113326985B (en) | 2021-05-31 | 2021-05-31 | Short-term load prediction method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113326985B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113705929B (en) * | 2021-09-15 | 2024-05-07 | 中国南方电网有限责任公司 | Spring festival holiday load prediction method based on load characteristic curve and typical characteristic value fusion |
| CN114221353B (en) * | 2021-12-06 | 2023-09-26 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | Power grid regulation method, device, equipment, storage medium and computer program product |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8065098B2 (en) * | 2008-12-12 | 2011-11-22 | Schneider Electric USA, Inc. | Progressive humidity filter for load data forecasting |
| JP6118975B2 (en) * | 2015-02-27 | 2017-05-10 | 株式会社Ｅ．Ｉ．エンジニアリング | Energy prediction system, method of operating energy prediction system, computer program for executing the same, computer-readable recording medium storing this program, and driving support system |
| CN106600463B (en) * | 2016-12-21 | 2020-08-14 | 广东电网有限责任公司电力调度控制中心 | Local shape similarity ultra-short term load prediction method and device |
| CN109190856A (en) * | 2018-10-30 | 2019-01-11 | 深圳供电局有限公司 | Urban power grid short-term load prediction method and system for distinguishing days to be predicted |
| CN111553529B (en) * | 2020-04-27 | 2023-02-17 | 新奥数能科技有限公司 | Load prediction method and device, computer readable storage medium and electronic equipment |
| CN112232600A (en) * | 2020-11-17 | 2021-01-15 | 烟台海颐软件股份有限公司 | Short-term power load prediction method and system based on variational modal decomposition |
-
2021
- 2021-05-31 CN CN202110604166.4A patent/CN113326985B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113326985A (en) | 2021-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113326985B (en) | Short-term load prediction method and device | |
| US7421460B2 (en) | Method for determining execution of backup on a database | |
| CN111064614B (en) | Fault root cause positioning method, device, equipment and storage medium | |
| US10580095B2 (en) | Method and system for water production and distribution control | |
| CN109682436B (en) | Method for additionally reading meter reading data of gas meter of Internet of things | |
| CN111831704B (en) | Determination method and device of abnormal data, storage medium and electronic equipment | |
| US10740336B2 (en) | Computerized methods and systems for grouping data using data streams | |
| JP6245030B2 (en) | Power consumption prediction method, power consumption prediction program, and power consumption prediction apparatus | |
| CN108897821B (en) | Method and device for automatically generating data conclusion | |
| CN111414410B (en) | Data processing method, device, equipment and storage medium | |
| CN112307002A (en) | Data compensation method, system and computer readable storage medium | |
| JP6012860B2 (en) | Work time estimation device | |
| CN112165412B (en) | Flow prediction method based on time series model | |
| CN116450280A (en) | Container mixed expansion and contraction method and system based on machine learning | |
| CN112926801B (en) | Load curve combined prediction method and device based on quantile regression | |
| CN112732544B (en) | Computer hardware adaptation intelligent analysis system | |
| CN112767056B (en) | Service data prediction method, device, computer equipment and storage medium | |
| CN115034906A (en) | Financial product configuration scheme memorizing and recommending method and system | |
| CN113703974A (en) | Method and device for predicting server capacity | |
| CN115424826B (en) | Method, device, equipment and storage medium for determining cooling performance of converter transformer | |
| JP7515322B2 (en) | Setting management device, setting management method, and setting management program | |
| CN115459299B (en) | Low-voltage distribution reactive power regulation method, device, computer equipment and storage medium | |
| JP2023183921A (en) | Determination apparatus, determination system, determination method, and program | |
| Cailloux | Dealing with numerical imprecision in mathematical programs for Electre Tri models disaggregation | |
| CN118228069A (en) | Method, apparatus, device, medium and program product for predicting electric load |
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 |