CN116432510A - Flexible load prediction method based on ensemble learning-longhorn beetle whisker optimization algorithm - Google Patents
Flexible load prediction method based on ensemble learning-longhorn beetle whisker optimization algorithm Download PDFInfo
- Publication number
- CN116432510A CN116432510A CN202211364298.5A CN202211364298A CN116432510A CN 116432510 A CN116432510 A CN 116432510A CN 202211364298 A CN202211364298 A CN 202211364298A CN 116432510 A CN116432510 A CN 116432510A
- Authority
- CN
- China
- Prior art keywords
- day
- data
- load
- prediction
- historical
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 92
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000005457 optimization Methods 0.000 title claims abstract description 49
- 241000254173 Coleoptera Species 0.000 title claims abstract description 40
- 238000005070 sampling Methods 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 22
- 238000010801 machine learning Methods 0.000 claims abstract description 15
- 230000004044 response Effects 0.000 claims abstract description 12
- 230000002159 abnormal effect Effects 0.000 claims description 18
- 238000004458 analytical method Methods 0.000 claims description 17
- 230000008859 change Effects 0.000 claims description 17
- 238000010606 normalization Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 12
- 230000005611 electricity Effects 0.000 claims description 11
- 230000006870 function Effects 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000011161 development Methods 0.000 claims description 6
- 230000008901 benefit Effects 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 5
- 230000007613 environmental effect Effects 0.000 claims description 5
- 238000012360 testing method Methods 0.000 claims description 4
- 101100268668 Caenorhabditis elegans acc-2 gene Proteins 0.000 claims description 3
- 101100268670 Caenorhabditis elegans acc-3 gene Proteins 0.000 claims description 3
- 101100268671 Caenorhabditis elegans acc-4 gene Proteins 0.000 claims description 3
- 101100373202 Rattus norvegicus Cx3cl1 gene Proteins 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 238000002790 cross-validation Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 238000010586 diagram Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000012098 association analyses Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000007621 cluster analysis Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000005251 gamma ray Effects 0.000 description 1
- 238000005338 heat storage Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 230000000087 stabilizing effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/04—Power grid distribution networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/06—Power analysis or power optimisation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a flexible load prediction method based on an ensemble learning-longhorn beetle whisker optimization algorithm, which comprises the steps of obtaining historical daily load data and quantized similar daily influence factor data; normalizing the processed data; selecting a history day as a history prediction day for training and sampling, and outputting the sampling data of the next step of history similar day; n pieces of historical similar day data output by a similar day algorithm are used as input of an integrated learning algorithm, and a sub-learner is trained according to the existing historical data; adopting a longhorn beetle whisker optimization algorithm to intelligently adjust and optimize parameters of a plurality of machine learning algorithms; training each sub learner and obtaining prediction accuracy under the optimal parameters; and assisting power system staff to complete the prediction work of flexible load demand response. According to the invention, through the developed flexible load prediction algorithm for the power demand response, the problem of orderly power utilization of the power grid in the power utilization peak period is solved, the load regulation capability is accurately known in real time, and the demand response is converted into non-real-time response.
Description
Technical Field
The invention relates to the technical field of multisource coordination control and flexible load prediction, in particular to a flexible load prediction method based on an integrated learning-longhorn beetle whisker optimization algorithm.
Background
With the continuous development of the power distribution network, the flexible load scheduling technology represented by the demand response can enable the original rigid load demand to become a part of schedulable resources, and meanwhile, the intelligent power utilization regulation technology of the power distribution and utilization system is researched and developed, so that the problem of peak power utilization demand is increasingly important. The flexible load can be understood as a load with the electricity consumption capable of being changed in a designated interval or being transferred in different time intervals, the part of load can play a role in stabilizing new energy fluctuation and peak clipping and valley filling by adjusting the electricity consumption of the load, peak shaving pressure and standby cost of a power supply side are shared, and win-win of the power supply side and the electricity consumption side is realized. The building environment to which the temperature control load represented by the air conditioner belongs has heat storage capability, can convert electric energy into heat energy for storage in a specific time, has no obvious influence on a human body in a certain temperature range, and therefore has a better application prospect compared with other traditional loads. Through reasonable regulation and control, not only can load be reduced fast, gentle load peak pressure can provide multiple auxiliary services moreover, guarantees the safe and stable operation of electric wire netting, improves system operation efficiency, compares with the construction power plant, and investment cost is low, has good social and economic benefits. It is expected that unified regulation of such distributed resources will become an important means for guaranteeing the real-time supply and demand balance of the future power grid.
At present, two main research methods of distributed flexible load prediction exist, wherein the methods based on experience knowledge comprise an expert system method, a growth coefficient method and the like, and the methods based on data driving comprise a neural network method, cluster analysis, association analysis and the like. However, both of these approaches currently have certain drawbacks, in that the empirical knowledge based approach is only suitable for certain scenarios, and the data driven based approach requires a large data set. In recent years, with the rapid development of machine learning theory, novel algorithms such as random forest algorithm, gradient elevator GBM algorithm, catboost algorithm and the like are emerging to provide theoretical basis for prediction and identification. The learning ability of the network learning is very strong, good prediction results can be obtained, however, different machine learning algorithms are applicable to different scenes and different requirements on the characteristics of the data set, so that the built model is only applicable to partial data, and the universality of the traditional prediction algorithm model prevents the application of the diversified algorithm in actual engineering.
The invention solves the problem of orderly power utilization of the power grid in the power utilization peak period through the researched and developed power demand response flexible load prediction algorithm based on the power distribution and utilization global sensing technology. The load regulation capability can be accurately known in real time through the global perception technology, so that the demand response is converted into the real-time response from non-real-time.
Disclosure of Invention
This section is intended to outline some aspects of embodiments of the invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section as well as in the description summary and in the title of the application, to avoid obscuring the purpose of this section, the description summary and the title of the invention, which should not be used to limit the scope of the invention.
The present invention has been made in view of the above and/or problems associated with existing flexible load prediction methods based on ensemble learning-longhorn beetle whisker optimization algorithms.
Therefore, the problem to be solved by the invention is how to provide a flexible load prediction method based on an ensemble learning-longhorn beetle whisker optimization algorithm.
In order to solve the technical problems, the invention provides the following technical scheme: a flexible load prediction method based on an ensemble learning-longhorn beetle whisker optimization algorithm comprises the steps of acquiring historical daily load data and quantized similar daily impact factor data from a power system load database, preprocessing the data, and selecting the similar daily impact factor data with highest correlation with flexible load change influence;
traversing the daily load data and carrying out normalization processing;
selecting a history day as a history prediction day for training and sampling, and outputting the sampling data of the next step of history similar day;
n pieces of historical similar day data output by a similar day algorithm are used as input of an integrated learning algorithm, and after key indexes are extracted, a sub-learner is trained according to the existing historical data;
adopting a longhorn beetle whisker optimization algorithm to intelligently adjust and optimize parameters of a plurality of machine learning algorithms;
training each sub-learner is completed, and the prediction accuracy of each sub-learner under the optimal parameters is obtained;
and calling the integrated learning prediction model which is completed with training to output a load prediction data result and a prediction curve, and assisting power system staff to complete the prediction work of flexible load demand response.
As a preferable scheme of the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm, the invention comprises the following steps: the similar day influence factor data at least comprises one or more of a historical day difference factor, a day type factor and a meteorological condition factor, and the similar day influence factor data respectively correspond to the periodicity, the continuity and the characteristics which are easily influenced by environmental factors of the power load;
preprocessing data, analyzing by using a correlation index, and selecting the data of the similar day influence factors with highest influence correlation with flexible load change, wherein the calculation formula of the correlation coefficient is as follows:
as a preferable scheme of the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm, the invention comprises the following steps: the step of traversing the history daily load data and normalizing the data comprises the following steps:
traversing the history daily load data, searching the missing data and supplementing the missing data set by using a piecewise linear interpolation method;
judging abnormal data and deleting the abnormal data, and adding a missing data set by piecewise linear interpolation, wherein the abnormal data reference formula is as follows:
wherein P is avg (t) is the average load of the sample set for 30 days at time t; epsilon is a constant, the abnormal tolerance range is calibrated, and abnormal data with different degrees can be removed according to the value of epsilon;
the data normalization processing is carried out, and the method comprises the following steps:
processing continuous data by using a maximum and minimum normalization method, and converting the data into continuous variables between 0 and 1 on the premise of not changing data distribution, thereby accelerating convergence of machine learning;
wherein x is min ,x max Representing the minimum and maximum values of the original data, respectively.
As a preferable scheme of the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm, the invention comprises the following steps: the step of selecting one history day as a history prediction day, training and sampling, and outputting the sampling data of the next step of history similar days comprises the following steps:
selecting a historical day as a historical prediction day, wherein all the historical data during sample training select data before a time axis of the historical prediction day, and the data and the day to be predicted are used as inputs of a similar day algorithm;
calculating the comprehensive similarity coefficient of each historical day sample;
and selecting N historical days with the greatest comprehensive similarity as similar days of the days to be predicted, sampling, and outputting the sampling data of the next historical similar days.
As a preferable scheme of the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm, the invention comprises the following steps: the history daily load data is traversed, normalization processing is carried out, the comprehensive similarity coefficient is calculated, N history days with the largest comprehensive similarity are selected as similar days of the days to be predicted, sampling is carried out, and a training set foundation is provided for integrated learning.
As a preferable scheme of the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm, the invention comprises the following steps: and integrating the sub learners by adopting the integrated learning algorithm, and playing the advantages of each sub learner in different scenes.
As a preferable scheme of the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm, the invention comprises the following steps: the intelligent optimization of the parameters of a plurality of machine learning algorithms by adopting the longhorn beetle whisker optimization algorithm comprises the optimization of the parameters of the algorithm by taking the prediction accuracy of each learner as an fitness function, so that the automatic adjustment of the parameters is realized, the model is easy to migrate to other data sets for application, and the universality of the model and the algorithm is enhanced.
As one of the flexible load prediction methods based on the ensemble learning-longhorn beetle whisker optimization algorithm
A preferred embodiment, wherein: the date difference factor date difference index aims at evaluating whether the characteristics of periodicity, near-large and far-small can be embodied between the historical date data and the target prediction date, quantifying the date difference factor and introducing a time similarity coefficient delta i (t) as a factor for evaluating the effect of the date gap:
wherein i is a historical day number, and t is the i-th day distance target forecast day number; beta 1 、β 2 And beta 3 For the decay factor, the values of 0.9 to 0.98 and β1 are generally given for similar reductions for each increase in the distance between the historical day and the predicted day for one day, one week and one year, respectively>β2>Beta 3; mod () is a remainder function, and int () is a rounding function; n1, N2 and N3 are all constants, where N1 and N2 are days of the week 7 and N3 is days of the year 365;
the day type factor includes: the load of the power system is periodically reflected in the regularity of load change in one week, and the power load during working dates mainly consists of commercial building electricity and industrial load, and the load change during working days caused by the stable operation of the power system has certain similarity; during the holiday, the electric load consists of resident electricity and service loads; dividing the day types into seven types according to the difference of average loads of seven days of the week;
let the day average load from monday to sunday be P (i=1, 2, …, 7), wherein the highest day type average load Pm occurs on the m-th day, and its map value is defined as 1; the lowest day type average load P occurs on the nth day, its mapping value is defined as 0.1, and the day types are mapped one by one according to the following formula:
the method comprises the steps of introducing a day type factor similarity coefficient to represent the similarity degree of a history day and a predicted day on the day type, wherein the larger the day type similarity degree is, the larger the day type factor similarity coefficient is, and calculating the day type factor similarity coefficient yi of the ith history day and the predicted day by adopting the following formula:
γ i ＝1-|f(P i )-f(P 0 )|
in the above, f (P i ) And f (P) 0 ) The day type map values of the i-th historical day and the predicted day are respectively shown.
As a preferable scheme of the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm, the invention comprises the following steps: the meteorological condition factors include:
after each index of the meteorological factors is quantitatively processed and corrected, introducing a meteorological factor similarity coefficient, and representing the similarity degree of the prediction day and the history day on the weather; because most of daily weather factor data are sequences which change along with time, evaluating the weather factor similarity of the predicted day and the historical day is to compare the similarity of the predicted day sequence and the historical day sequence of each index; according to the characteristic of the meteorological factors, a gray correlation analysis method is adopted to conduct similarity analysis, and the method can measure the correlation degree between the factors according to the same or different development trend degrees (namely gray correlation degrees) between the factors in the system so as to reflect the similarity degree; the correlation coefficient xi of all meteorological factors is synthesized i (k) The correlation degree of the forecast day and the ith historical day on the meteorological factors is as follows:
when the similarity of the forecast day and the history day on the meteorological factors is larger, the association coefficient is larger, the association degree is also larger, so that the association degree r is taken i As a meteorological factor similarity coefficient.
As a preferable scheme of the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm, the invention comprises the following steps: the intelligent optimization of the parameters of a plurality of machine learning algorithms by adopting the longhorn beetle whisker optimization algorithm comprises the following steps of: 4, carrying out 5 times of cross validation set and test set total data set prediction accuracy, wherein each prediction accuracy is respectively recorded as Acc1, acc2, acc3 and Acc4, the adaptive function of the algorithm is the weighted sum of the prediction accuracies of each sub-learner, and the 4 prediction accuracies respectively occupy 0.25 weight for balancing each evaluation index.
The flexible load prediction algorithm based on integrated learning and longhorn beetle whisker optimization intelligent parameter adjustment has the beneficial effects that (1) the flexible load prediction algorithm based on integrated learning and longhorn beetle whisker optimization intelligent parameter adjustment is provided, the calculation speed is high, the flexible load prediction algorithm is suitable for different data set scenes under the action of an integrated learning network for completing training, and a basis and a foundation are provided for on-line prediction of flexible load prediction. (2) Compared with other traditional algorithms, the method has the advantages of short iteration time, high execution efficiency, no need of a large amount of storage space, breaking of the previous limitation, avoiding of pathological convergence by the integrated optimization idea of weighted voting, and solving of the defects of a single intelligent algorithm. Big data and heuristic intelligent optimization algorithm are fully utilized, and a complicated engineering modeling process is avoided; (3) The method for integrated learning and parameter automatic optimization is applied to flexible load prediction for the first time, and a new research thought is provided for flexible load prediction.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. Wherein:
FIG. 1 is a schematic diagram of an implementation flow chart of the present invention.
FIG. 2 is a flowchart of an algorithm based on ensemble learning according to the present invention.
Fig. 3 is a program flow chart of the longhorn beetle whisker optimization algorithm of the present invention.
Fig. 4 is a plot of the resident load prediction relative error scatter of the present invention.
Fig. 5 is a charge amount relative error scatter diagram of the electric vehicle according to the present invention.
Fig. 6 is a charge amount curve of the electric vehicle of the present invention.
Detailed Description
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the invention. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Example 1
Referring to fig. 1 to 3, a first embodiment of the present invention provides a flexible load prediction method based on an ensemble learning-longhorn beetle whisker optimization algorithm, the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm including
S1, acquiring historical daily load data and quantized similar daily impact factor data from a power system load database, preprocessing the data, and selecting the similar daily impact factor data with the highest influence correlation with flexible load change.
Further, the similar day influence factor data at least comprises one or more of a historical day difference factor, a day type factor and a meteorological condition factor, and the similar day influence factor data respectively correspond to the characteristics of periodicity, continuity and susceptibility to environmental factors of the power load;
preprocessing data, analyzing by using a correlation index, and selecting the data of the similar day influence factors with highest influence correlation with flexible load change.
Specifically, the input body of the similar day algorithm is historical day load data adjacent to the predicted day in a set time range, and the accurate algorithm result must be established on the basis of comprehensive and accurate historical load data of the power system. Through the SCADA system of the data acquisition and monitoring control system, in the process of summarizing the electric power data by the acquisition end through layer-by-layer transmission, each stage is likely to have deviation, if the electric power system fails to overhaul or accident fails to work, the acquired electric power data records abnormal load changes and has no analysis significance. Therefore, the elimination of the deviation data and the abnormal load data is a necessary measure for ensuring that the result of the similar day algorithm meets the expectations. And applies the correlation index analysis to select the data of the similar day influence factors with the highest influence correlation with the flexible load change,
the calculation formula of the correlation coefficient is as follows:
s2, traversing the daily load data and carrying out normalization processing.
Furthermore, traversing the history daily load data, searching the missing data and supplementing the missing data set by using a piecewise linear interpolation method;
specifically, the load curve is greatly affected by daily characteristic-related factors and generally has continuity over a period of time, so that if the date difference is too large, the load may be significantly different from day to day. Therefore, in the data preprocessing operation, if an extreme value appears at the time t, historical data of the first 30 days are taken as a comparison sample set. For the moment t, the load P (t).
Judging abnormal data and deleting the abnormal data, and adding a missing data set by piecewise linear interpolation, wherein the abnormal data reference formula is as follows:
wherein P is avg (t) is the average load of the sample set for 30 days at time t; epsilon is a constant, the abnormal tolerance range is calibrated, and abnormal data with different degrees can be removed according to the value of epsilon;
after the abnormal data is removed, the power data acquired by the power system is needed to be completed, and the situation that the power data is not disassembled or recorded by the system exists for various reasons. This missing data is complemented by piecewise linear interpolation, the following equation.
Wherein P (ti) is the power data at the ith and the tth time.
And carrying out data normalization processing. Data normalization is a means by which data is often processed before many arithmetic operations to cancel differences between data that differ in dimension or magnitude. The method can reduce the possibility of occurrence of prediction error increase. And processing continuous data by using a maximum and minimum normalization method, and converting the data into continuous variables between 0 and 1 on the premise of not changing data distribution, thereby accelerating convergence of machine learning.
The method comprises the following steps:
processing continuous data by using a maximum and minimum normalization method, and converting the data into continuous variables between 0 and 1 on the premise of not changing data distribution, thereby accelerating convergence of machine learning;
wherein x is min ,x max Representing the minimum and maximum values of the original data, respectively.
S3, selecting a historical day as a historical prediction day to train and sample, and outputting the sampling data of the next historical similar day.
Further, selecting a history day as a history prediction day, training and sampling, and outputting the next step of history similar day sampling data, wherein the step of outputting the next step of history similar day sampling data comprises the following steps:
selecting a historical day as a historical prediction day, wherein all the historical data during sample training select data before a time axis of the historical prediction day, and the data and the day to be predicted are used as inputs of a similar day algorithm;
calculating the comprehensive similarity coefficient of each historical day sample;
and selecting N historical days with the greatest comprehensive similarity as similar days of the days to be predicted, sampling, and outputting the sampling data of the next historical similar days.
Further, the similar day effect factors include a date gap factor, a day type factor, and a weather condition factor. Because the power load has periodicity and continuity in a period, the basis for selecting the data sample set is provided for the similar day algorithm. In addition, the electric load is also characterized by being influenced by environmental factors, such as different electric loads in summer and winter due to different temperatures, different electric loads in working days and resting days, and the like. In summary, three factors, namely date gap, day type and meteorological condition, are selected as the influence factors of similar days, and correspond to the characteristics of periodicity, continuity and susceptibility to environmental factors of the power load.
Specifically, a date gap factor is calculated. The periodicity and the continuity of the load data of the power system are reflected in a shorter time range, the power load has periodicity, namely, the date difference factor is influenced by the periodic cycle, and the continuity is reflected as the smaller the date difference is, the closer the load characteristic is to the target predicted daily load characteristic. Therefore, the date gap index aims at evaluating whether the periodicity and the near-far-small characteristics can be represented between the historical day data and the target prediction day. Quantizing the date difference factor, and introducing a time similarity coefficient delta i (t) as a factor for evaluating the effect of the date gap:
wherein i is a historical day number, and t is the i-th day distance target forecast day number; beta 1 、β 2 And beta 3 For attenuation coefficients, respectively representing calendarsShi Ri and the like reduction ratio of each day, week and year of the increase of the predicted daily period distance generally take values of 0.9 to 0.98, and β1>β2>Beta 3; mod () is a remainder function, and int () is a rounding function; n1, N2 and N3 are all constants, where N1 and N2 are days of the week 7 and N3 is days of the year 365.
Specifically, a day type factor is measured. The load periodicity of the power system is represented by the regularity of load change in one week. The power load during working days mainly comprises commercial building power consumption and industrial load, and the stable operation of the power load causes the working day load change to have certain similarity; during the holiday, the electrical load consists of residential electricity and service loads. As a result of investigation, the day types are classified into seven categories, since the average loads are different from each other on seven days of the week. Let the day average load from monday to sunday be P (i=1, 2, …, 7), wherein the highest day type average load Pm occurs on the m-th day, and its map value is defined as 1; the lowest day type average load P occurs on day n, and its map value is defined as 0.1. Mapping the day types one by one according to the following steps:
and introducing a date type factor similarity coefficient to represent the similarity degree of the historical date and the predicted date on the date type. The greater the day type similarity, the greater the day type factor similarity coefficient. Calculating a day type factor similarity coefficient yi of the ith historical day and the predicted day by adopting the following formula:
γ i ＝1-|f(P i )-f(P 0 )|
in the above, f (P i ) And f (P) 0 ) The day type map values of the i-th historical day and the predicted day are respectively shown.
Specifically, the weather factor is measured. After each index of the meteorological factors is quantized and corrected, introducing a meteorological factor similarity coefficient, and representing and predictingDay and history day are weather-like. As most of daily weather factor data are sequences which change along with time, evaluating the weather factor similarity of the predicted day and the historical day is to compare the similarity of the predicted day sequence and the historical day sequence of each index. According to the characteristic of the meteorological factors, a gray correlation analysis method is adopted to conduct similarity analysis, and the method can measure the correlation degree between the factors according to the same or different development trend degrees (namely gray correlation degrees) between the factors in the system, so that the similarity degree is reflected. The correlation coefficient xi of all meteorological factors is synthesized i (k) The correlation degree of the forecast day and the ith historical day on the meteorological factors is as follows:
when the similarity of the forecast day and the history day on the meteorological factors is larger, the association coefficient is larger, the association degree is also larger, so that the association degree r is taken i As a meteorological factor similarity coefficient. So far, the similarity coefficient of each influence factor is obtained.
Specifically, the comprehensive similarity is measured and calculated. The similarity coefficient of each factor quantifies the similarity of the predicted day and the historical day in the date difference and day type. In practice, the load change will be affected by a combination of these factors, and thus a combination of all factors needs to be considered. The comprehensive similarity coefficient is introduced, and when the total similarity of the historical day and the prediction day is quantized, a mode of multiplying the similarity coefficients of all factors is selected when the total similarity calculation method is constructed, the dominant factors can be simply and automatically identified, and the weight setting problem of all the factors can be solved. Thus, the comprehensive similarity coefficient theta between the predicted day and the ith history day is calculated according to the following formula i ：
θ i ＝δ i γ i r i
In the above, delta i A time factor similarity coefficient representing a predicted day and an i-th historical day; gamma ray i And the date type factor similarity coefficient of the forecast date and the ith historical date is represented.
The comprehensive similarity coefficient can quantify the comprehensive similarity of the date difference between the predicted date and the historical date and the date type. Therefore, when the similar days are selected, N historical days close to the predicted daily load characteristic are selected according to the comprehensive similarity coefficient from large to small. And selecting N historical days with the greatest comprehensive similarity as similar days of the days to be predicted, sampling, and outputting the sampling data of the next historical similar days.
S4, utilizing N pieces of historical similar day data output by the similar day algorithm as input of the integrated learning algorithm, and training the sub-learner according to the existing historical data after extracting key indexes.
Furthermore, N historical similar day data output by the similar day algorithm is used as input of the integrated learning algorithm, and after key indexes are extracted, the sub-learner is trained according to the existing historical data.
S5, intelligently adjusting parameters of a plurality of machine learning algorithms by adopting a longhorn beetle whisker optimization algorithm.
Furthermore, the parameter tuning of a plurality of sub learners involves a large number of manual operations, in order to realize the self-adaptive tuning of the parameters and save the labor cost, the invention proposes that the parameters of a plurality of machine learning algorithms are intelligently tuned by adopting a longhorn beetle whisker optimization algorithm, each sub learner is in a training set, a 5-time cross validation set (divided according to the proportion of 1:4), a test set, the prediction accuracy of a total data set, each prediction accuracy is respectively recorded as Acc1, acc2, acc3 and Acc4, the fitness function of the algorithm is the weighted sum of the prediction accuracy of each sub learner, and the invention balances each evaluation index, and 4 prediction accuracy respectively accounts for 0.25 weight.
And S6, training each sub-learner is completed, and the prediction accuracy of each sub-learner under the optimal parameters is obtained.
Furthermore, after each sub-learner finishes training, the prediction accuracy of each sub-learner under the optimal parameters can be obtained. Under the condition of ensuring the accuracy of the model algorithm, the applicability of the model and the algorithm is improved, the flexible load prediction is carried out by adopting the thought of a weighted voting method, and the category with the highest final score is the final category.
And S7, calling an integrated learning prediction model which is trained to output a load prediction data result and a prediction curve, and assisting power system staff to complete the prediction work of flexible load demand response.
Furthermore, on the test set applying the flexible load, an integrated learning prediction model which is used for completing training is called to output a load prediction data result and a prediction curve, so that power system staff is assisted to complete the prediction work of flexible load demand response.
Example 2
Referring to fig. 1 to 6, a second embodiment of the present invention is different from the first embodiment in that the present embodiment takes load prediction on certain dates in summer in 2017 in two regions as an example, and it is verified whether the load potential analysis method is reliable. On one hand, the main research content of the embodiment is load potential analysis, and on the other hand, the rapid development of the electric automobile and the bright prospect thereof lead the load potential analysis work of the electric automobile to be gradually scheduled, and the precision of the load prediction of the electric automobile also needs to be considered.
The present embodiment performs an example analysis from two angles of resident electricity load potential analysis and electric vehicle load potential analysis, respectively. And predicting domestic and civil electric conditions by using the data of a certain tropical monsoon climate type region, and predicting the load of the electric automobile by using the data of a certain subtropical monsoon climate type region.
When the load prediction of the electric vehicle is performed, the present embodiment analyzes the charged electric quantity thereof. Because the charging condition of the electric automobile can reversely indicate the electricity consumption condition to a certain extent, the prediction of the charging condition can reflect the electricity consumption load prediction condition.
First, residential and civil electrical load prediction analysis
The method predicts the load condition of residents in a certain district in a certain month in summer in a certain tropical monsoon climate type region. The cell is provided with an independent power supply circuit, so that the load interference of other circuits can be eliminated to a certain extent, the relative error condition of 96 nodes is obtained through calculation, and 0 is listed in the temporary list: 00 to 11:45 the relative error value at these 48 points in time. The relative prediction error of the residential and civil electrical load is shown in table 1-
Table 1 relative error in household electrical load prediction
| Time point | Relative error/% | Time point | Relative error/% | Time point | Relative error/% |
| 0：00 | -8.977 | 4：00 | 7.896 | 8：00 | 0.000 |
| 0：15 | -7.896 | 4：15 | -8.977 | 8：15 | -8.491 |
| 0：30 | -8.977 | 4：30 | -4:110 | 8：30 | 7.767 |
| 0：45 | -4.110 | 4：45 | -3.530 | 8：45 | 6.186 |
| 1：00 | -3.530 | 5：00 | 0.000 | 9：00 | 6.384 |
| 1：15 | -3.798 | 5：15 | -7.394 | 9：15 | 0.000 |
| 1：30 | -7.589 | 5：30 | 0.000 | 9：30 | -2.830 |
| 1：45 | -3.798 | 5：45 | -6.394 | 9：45 | 6.001 |
| 2：00 | -8.589 | 6：00 | -7.394 | 10：00 | -5.661 |
| 2：15 | 8.220 | 6：15 | 7.896 | 10：15 | 9.001 |
| 2：30 | 4.286 | 6：30 | -8.977 | 10：30 | -5.661 |
| 2：45 | 3.798 | 6：45 | -7.318 | 10：45 | 8.739 |
| 3：00 | -7.896 | 7：00 | 7.596 | 11：00 | 0.000 |
| 3：15 | -3.798 | 7：15 | -3.948 | 11:15 | 2.913 |
| 3：30 | 0.000 | 7：30 | 7.896 | 11:30 | 8.001 |
| 3:45 | -3.798 | 7：45 | 9.636 | 11:45 | 0.000 |
And obtaining the maximum similar day which is the most consistent with the similar day factor of the predicted day and is the nearest similar day to the predicted day based on the algorithm flow. And obtaining daily power load data from the corrected power load database, comparing the daily power load data with the actual power load data of the predicted day as a prediction result, and performing error analysis by using a relative error formula. To more intuitively see the distribution of the relative error values, a plot of the relative error scatter of the resident load predictions at 96 time points is drawn, see fig. 4.
As can be seen from table 1 and fig. 4, the relative error is within the range of 10.000%, most of the errors are distributed within 8.000%, the error distribution is relatively uniform, the relative error is 19.677% at maximum, 0.000% at minimum, and there are 14 time points at which the error is 0.000%. The relative error is within the allowable range. The prediction result is reliable through quantitative calculation of relative errors and drawing of a scatter diagram. In conclusion, the method has certain effect when the residential electric load is predicted.
Next, a single day predictive analysis of the charge amount of the electric vehicle is performed
The method selects the day in a month in summer in a subtropical monsoon climate type region as the prediction day. And obtaining data such as time period and electric quantity of each charging of the electric vehicle charging pile in the region for several months, and obtaining the charging quantity database of the electric vehicle in the region after calculation. And predicting the charge quantity of the electric automobile so as to reflect the load and the load prediction condition of the electric automobile in the region.
Similarly, the maximum likelihood that the factors of the likelihood days are the most consistent with the predicted day and the nearest likelihood day to the predicted day is obtained by the load prediction flow. The charging amounts of the electric vehicles at 96 times in the day are obtained from the corrected electric load database, and are used as prediction results of each time, and are compared with actual charging amounts at each time in the prediction day, and error analysis is performed by using a relative error formula. Calculating to obtain the relative error value of the charge quantity prediction at each moment, and temporarily listing 0:00 to 11:45 the relative error of these 48 time points.
The relative error of the electric vehicle charge quantity prediction is shown in table 2, and in order to more intuitively observe the relative error numerical distribution, an electric vehicle charge quantity relative error scatter diagram of 96 time points is drawn, and is shown in fig. 5.
Table 2 relative error in electric vehicle charge amount prediction
| Time point | Relative error/% | Time point | Relative error/% | Time point | Relative error/% |
| 0：00 | 1.504 | 4：00 | 4.305 | 8：00 | -10.337 |
| 0：15 | 1.504 | 4：15 | 4.305 | 8：15 | -8.280 |
| 0：30 | 1.504 | 4：30 | 4.305 | 8：30 | -9.114 |
| 0：45 | 1.504 | 4：45 | 4.305 | 8：45 | -6.518 |
| 1：00 | 1.504 | 5：00 | 4.305 | 9：00 | 0.008 |
| 1：15 | 1.504 | 5：15 | 4.305 | 9：15 | 8.240 |
| 1：30 | 1.504 | 5：30 | 4.305 | 9：30 | 8.842 |
| 1：45 | 1.504 | 5：45 | 4.305 | 9：45 | 9.884 |
| 2：00 | 1.504 | 6：00 | 1.846 | 10：00 | 9.884 |
| 2：15 | 1.504 | 6：15 | 1.846 | 10：15 | 9.695 |
| 2：30 | 1.504 | 6：30 | 1.846 | 10：30 | 8.695 |
| 2：45 | 1.504 | 6：45 | 1.846 | 10：45 | 1.251 |
| 3：00 | 1.504 | 7：00 | 1.846 | 11：00 | 4.766 |
| 3：15 | 1.504 | 7：15 | 1.725 | 11:15 | -8.642 |
| 3：30 | 1.504 | 7：30 | -1.093 | 11:30 | -8.642 |
| 3:45 | 4.305 | 7：45 | -1.093 | 11:45 | 10.993 |
By quantitative calculation of relative errors and drawing of a scatter diagram, it is known that the charge amount prediction of the electric vehicle in one day is more reliable to a certain extent.
Again, one week predictive analysis of electric vehicle charge
To further verify the reliability of the method, the study continued to predict the electric vehicle charge daily for consecutive weeks in month 2017 in the area. The same method is adopted to obtain the maximum similar day of each day, and then the actual value and the predicted value of the predicted daily charge quantity are obtained. And drawing an electric vehicle charge quantity curve, wherein the daily charge quantity of the electric vehicle is kept at a higher level from Monday to Friday, and the charge quantity on Saturday and Sunday is obviously reduced greatly compared with that on weekdays, as shown in fig. 6. The charge amount on sunday is the least in one week, and friday is a small charge peak. The trend of the two curves in the graph is basically consistent, and the predicted value and the actual value of each day are relatively close.
Table 3 predicts the relative error for one week of charge as follows:
table 3 charge one week prediction relative error
| Number of weeks | Relative error/% |
| Monday | 6.045 |
| Zhoudi (Zhoudi) | -1.625 |
| Wednesday | -8.880 |
| Zhou four | 2.030 |
| Friday (friday) | 5.922 |
| Saturday (Saturday) | -3.296 |
| (Sunday) | 6.121 |
The prediction error of the 7-day charge was calculated, and the one-week charge prediction relative error is shown in table 3.
As can be seen from Table 3, the relative error predicted daily remains between + -9.000% for this week, mostly within + -7.000%. Wherein the maximum relative error is-8.880%, and the minimum relative error is-1.625%. The relative error is within the allowable range.
As can be seen from fig. 6 and table 3, this method has a certain reliability in daily charge amount prediction. The charge quantity prediction condition can reflect the load prediction condition to a certain extent, and the method is capable of integrating the single-day charge quantity prediction condition and the one-week charge quantity prediction condition of the electric vehicle, so that the method has certain reliability in the aspect of short-term load prediction of the electric vehicle. In conclusion, the method has certain reliability in terms of short-term load potential analysis.
It should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered in the scope of the claims of the present invention.
Claims (10)
1. A flexible load prediction method based on an ensemble learning-longhorn beetle whisker optimization algorithm is characterized by comprising the following steps of: comprising the steps of (a) a step of,
acquiring historical daily load data and quantized similar daily impact factor data from a power system load database, preprocessing the data, and selecting the similar daily impact factor data with the highest correlation with the flexible load change influence;
traversing the daily load data and carrying out normalization processing;
selecting a history day as a history prediction day for training and sampling, and outputting the sampling data of the next step of history similar day;
n pieces of historical similar day data output by a similar day algorithm are used as input of an integrated learning algorithm, and after key indexes are extracted, a sub-learner is trained according to the existing historical data;
adopting a longhorn beetle whisker optimization algorithm to intelligently adjust and optimize parameters of a plurality of machine learning algorithms;
training each sub-learner is completed, and the prediction accuracy of each sub-learner under the optimal parameters is obtained;
and calling the integrated learning prediction model which is completed with training to output a load prediction data result and a prediction curve, and assisting power system staff to complete the prediction work of flexible load demand response.
2. The flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm as set forth in claim 1, wherein: the similar day influence factor data at least comprises one or more of a historical day difference factor, a day type factor and a meteorological condition factor, and the similar day influence factor data respectively correspond to the periodicity, the continuity and the characteristics which are easily influenced by environmental factors of the power load;
preprocessing data, analyzing by using a correlation index, and selecting the data of the similar day influence factors with highest influence correlation with flexible load change, wherein the calculation formula of the correlation coefficient is as follows:
3. the flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm according to claim 1 or 2, wherein: the step of traversing the history daily load data and normalizing the data comprises the following steps:
traversing the history daily load data, searching the missing data and supplementing the missing data set by using a piecewise linear interpolation method;
judging abnormal data and deleting the abnormal data, and adding a missing data set by piecewise linear interpolation, wherein the abnormal data reference formula is as follows:
wherein P is avg (t) is the average load of the sample set for 30 days at time t; epsilon is a constant, the abnormal tolerance range is calibrated, and abnormal data with different degrees can be removed according to the value of epsilon;
the data normalization processing is carried out, and the method comprises the following steps:
processing continuous data by using a maximum and minimum normalization method, and converting the data into continuous variables between 0 and 1 on the premise of not changing data distribution, thereby accelerating convergence of machine learning;
wherein x is min ,x max Representing the minimum and maximum values of the original data, respectively.
4. The flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm according to claim 3, wherein: the step of selecting one history day as a history prediction day, training and sampling, and outputting the sampling data of the next step of history similar days comprises the following steps:
selecting a historical day as a historical prediction day, wherein all the historical data during sample training select data before a time axis of the historical prediction day, and the data and the day to be predicted are used as inputs of a similar day algorithm;
calculating the comprehensive similarity coefficient of each historical day sample;
and selecting N historical days with the greatest comprehensive similarity as similar days of the days to be predicted, sampling, and outputting the sampling data of the next historical similar days.
5. The flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm according to claim 1 or 4, wherein: the history daily load data is traversed, normalization processing is carried out, the comprehensive similarity coefficient is calculated, N history days with the largest comprehensive similarity are selected as similar days of the days to be predicted, sampling is carried out, and a training set foundation is provided for integrated learning.
6. The flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm according to claim 5, wherein: and integrating the sub learners by adopting the integrated learning algorithm, and playing the advantages of each sub learner in different scenes.
7. The flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm of claim 6, wherein: the intelligent optimization of the parameters of a plurality of machine learning algorithms by adopting the longhorn beetle whisker optimization algorithm comprises the optimization of the parameters of the algorithm by taking the prediction accuracy of each learner as an fitness function, so that the automatic adjustment of the parameters is realized, the model is easy to migrate to other data sets for application, and the universality of the model and the algorithm is enhanced.
8. The flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm of claim 7, wherein: the date difference factor date difference index aims at evaluating whether the characteristics of periodicity, near-large and far-small can be embodied between the historical date data and the target prediction date, quantifying the date difference factor and introducing a time similarity coefficient delta i (t) as a factor for evaluating the effect of the date gap:
wherein i is a historical day number, and t is the i-th day distance target forecast day number; beta 1 、β 2 And beta 3 For the decay factor, the values of 0.9 to 0.98 and β1 are generally given for similar reductions for each increase in the distance between the historical day and the predicted day for one day, one week and one year, respectively>β2>Beta 3; mod () is a remainder function, and int () is a rounding function; n1, N2 and N3 are all constants, where N1 and N2 are days of the week 7 and N3 is days of the year 365;
the day type factor includes: the load of the power system is periodically reflected in the regularity of load change in one week, and the power load during working dates mainly consists of commercial building electricity and industrial load, and the load change during working days caused by the stable operation of the power system has certain similarity; during the holiday, the electric load consists of resident electricity and service loads; dividing the day types into seven types according to the difference of average loads of seven days of the week;
let the day average load from monday to sunday be P (i=1, 2, …, 7), wherein the highest day type average load Pm occurs on the m-th day, and its map value is defined as 1; the lowest day type average load P occurs on the nth day, its mapping value is defined as 0.1, and the day types are mapped one by one according to the following formula:
the method comprises the steps of introducing a day type factor similarity coefficient to represent the similarity degree of a history day and a predicted day on the day type, wherein the larger the day type similarity degree is, the larger the day type factor similarity coefficient is, and calculating the day type factor similarity coefficient yi of the ith history day and the predicted day by adopting the following formula:
γ i ＝1-|f(P i )-f(P 0 )|
in the above, f (P i ) And f (P) 0 ) The day type map values of the i-th historical day and the predicted day are respectively shown.
9. The flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm of claim 8, wherein: the meteorological condition factors include:
after each index of the meteorological factors is quantitatively processed and corrected, introducing a meteorological factor similarity coefficient, and representing the similarity degree of the prediction day and the history day on the weather; because most of daily weather factor data are sequences which change along with time, evaluating the weather factor similarity of the predicted day and the historical day is to compare the similarity of the predicted day sequence and the historical day sequence of each index; according to the characteristic of the meteorological factors, a gray correlation analysis method is adopted to conduct similarity analysis, and the method can measure the correlation degree between the factors according to the same or different development trend degrees (namely gray correlation degrees) between the factors in the system so as to reflect the similarity degree; the correlation coefficient xi of all meteorological factors is synthesized i (k) The correlation degree of the forecast day and the ith historical day on the meteorological factors is as follows:
when the similarity of the forecast day and the history day on the meteorological factors is larger, the association coefficient is larger, the association degree is also larger, so that the association degree r is taken i As a meteorological factor similarity coefficient.
10. The flexible load prediction method based on the ensemble learning-longhorn beetle whisker optimization algorithm of claim 9, wherein: the intelligent optimization of the parameters of a plurality of machine learning algorithms by adopting the longhorn beetle whisker optimization algorithm comprises the following steps of: 4, carrying out 5 times of cross validation set and test set total data set prediction accuracy, wherein each prediction accuracy is respectively recorded as Acc1, acc2, acc3 and Acc4, the adaptive function of the algorithm is the weighted sum of the prediction accuracies of each sub-learner, and the 4 prediction accuracies respectively occupy 0.25 weight for balancing each evaluation index.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211364298.5A CN116432510A (en) | 2022-11-02 | 2022-11-02 | Flexible load prediction method based on ensemble learning-longhorn beetle whisker optimization algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211364298.5A CN116432510A (en) | 2022-11-02 | 2022-11-02 | Flexible load prediction method based on ensemble learning-longhorn beetle whisker optimization algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116432510A true CN116432510A (en) | 2023-07-14 |
Family
ID=87089574
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211364298.5A Pending CN116432510A (en) | 2022-11-02 | 2022-11-02 | Flexible load prediction method based on ensemble learning-longhorn beetle whisker optimization algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116432510A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116610911A (en) * | 2023-07-19 | 2023-08-18 | 南昌工程学院 | Power consumption data restoration method and system based on Bayesian Gaussian tensor decomposition model |
-
2022
- 2022-11-02 CN CN202211364298.5A patent/CN116432510A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116610911A (en) * | 2023-07-19 | 2023-08-18 | 南昌工程学院 | Power consumption data restoration method and system based on Bayesian Gaussian tensor decomposition model |
| CN116610911B (en) * | 2023-07-19 | 2023-09-19 | 南昌工程学院 | Power consumption data restoration method and system based on Bayesian Gaussian tensor decomposition model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2020104000A4 (en) | Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model | |
| CN111860979B (en) | Short-term load prediction method based on TCN and IPSO-LSSVM combined model | |
| US11581740B2 (en) | Method, system and storage medium for load dispatch optimization for residential microgrid | |
| CN109558975B (en) | Integration method for multiple prediction results of power load probability density | |
| CN103295075B (en) | A kind of ultra-short term load forecast and method for early warning | |
| CN110474339B (en) | Power grid reactive power control method based on deep power generation load prediction | |
| CN111915092B (en) | Ultra-short-term wind power prediction method based on long-short-term memory neural network | |
| CN110889527A (en) | Electric vehicle charging load prediction method based on LSTM neural network | |
| CN106650979A (en) | Feeder baseline load prediction method | |
| CN110909958A (en) | Short-term load prediction method considering photovoltaic grid-connected power | |
| CN116432510A (en) | Flexible load prediction method based on ensemble learning-longhorn beetle whisker optimization algorithm | |
| CN112348287A (en) | Electric power system short-term load probability density prediction method based on LSTM quantile regression | |
| CN116596139A (en) | Short-term load prediction method and system based on Elman neural network | |
| CN110991747A (en) | Short-term load prediction method considering wind power plant power | |
| CN114329347A (en) | Method and device for predicting metering error of electric energy meter and storage medium | |
| CN104252647B (en) | Electro-load forecast method based on anti-distance weighting interpolation method | |
| CN106022542A (en) | Enterprise gateway load prediction method based on operation load characteristics | |
| CN111697560B (en) | Method and system for predicting load of power system based on LSTM | |
| CN115965134A (en) | Regional power grid wind power generation power prediction optimization method | |
| CN115860797A (en) | Electric quantity demand prediction method suitable for new electricity price reform situation | |
| CN112052430B (en) | Photovoltaic station correlation comprehensive evaluation system based on improved TOPSIS | |
| CN114759579A (en) | Power grid active power optimization control system, method and medium based on data driving | |
| CN113313312A (en) | Spring festival load rolling prediction method based on short, medium and long time scales | |
| CN113487131A (en) | Station network interaction method based on charging demand and energy storage adjustment | |
| CN117595231B (en) | Intelligent power grid distribution management system and method thereof |
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 |