CN115392567A - Power load prediction method, electronic equipment, device and readable storage medium - Google Patents
Power load prediction method, electronic equipment, device and readable storage medium Download PDFInfo
- Publication number
- CN115392567A CN115392567A CN202211026800.1A CN202211026800A CN115392567A CN 115392567 A CN115392567 A CN 115392567A CN 202211026800 A CN202211026800 A CN 202211026800A CN 115392567 A CN115392567 A CN 115392567A
- Authority
- CN
- China
- Prior art keywords
- load
- day
- factor set
- factor
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000003860 storage Methods 0.000 title claims abstract description 14
- 238000007621 cluster analysis Methods 0.000 claims description 10
- 238000000354 decomposition reaction Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 230000014509 gene expression Effects 0.000 description 14
- 238000013528 artificial neural network Methods 0.000 description 8
- 238000005070 sampling Methods 0.000 description 7
- 238000012549 training Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000001556 precipitation Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000010248 power generation Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000005855 radiation Effects 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 102000003712 Complement factor B Human genes 0.000 description 2
- 108090000056 Complement factor B Proteins 0.000 description 2
- 230000009471 action Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000002699 waste material Substances 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- 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)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a power load prediction method, electronic equipment, a device and a readable storage medium, wherein the power load prediction method comprises the steps of searching a target factor set belonging to the same category as a factor set of a day to be predicted in the factor sets of all historical days, and taking the historical day where the target factor set is located as a similar historical day, wherein the factor set of any day comprises data of all load influence factors on the day; taking the load data of the similar historical days and the load data before the prediction time point of the day to be predicted as an original load sequence, and decomposing the original load sequence to obtain a plurality of load subsequences; and predicting the power load after the prediction time point on the day to be predicted based on the load subsequence and the data of the load influence factors at the same time point as the load subsequence. The accuracy of the power load prediction can be improved.
Description
Technical Field
The invention relates to the technical field of power systems, in particular to a power load prediction method, electronic equipment, a device and a readable storage medium.
Background
In recent years, with the rapid increase of the economic level of China, the power industry is continuously developed. Power load prediction is increasingly important for safe and economic operation of power systems. The accuracy of the power load prediction is directly related to whether the power generation amount of the power system is accurate. If the predicted value of the power load is larger, the power production is excessive, and energy waste and production machine loss are caused; if the predicted value of the power load is smaller, the power production is insufficient, the life of people is affected, and the economic development is lost.
In general, the power load may be affected by a large number of factors, and besides known factors such as date type, weather conditions, climate, etc., there may be unknown factors, which makes the power load have obvious characteristics such as large data volume, nonlinearity, uncertainty, and poor robustness. In the current power load prediction method, researchers screen out historical data with similar power load characteristic changes by performing cluster analysis on historical data of a power system, so as to predict future power loads.
At present, when the historical data is subjected to cluster analysis, the influence of each influencing factor on the power load is considered in sequence to perform the cluster analysis on the historical data. For example, assume that there is an influencing factor A and an influencing factor B. In the existing method, firstly, the influence of the influence factor a on the power load is considered, historical data is divided into a plurality of large classes, and then, under each large class, the influence of the influence factor B on the power load is continuously considered, and classification is continuously performed. The number of the types of the historical data obtained by classification in the way is in an exponential relation with the number of the influencing factors. Under the condition of less historical data, the data volume of each type of historical data obtained by final classification is less, so that the training of a prediction model of the power load is not facilitated, the accuracy of the prediction model obtained by final training is not high, and the predicted power load is not accurate.
Disclosure of Invention
In view of this, embodiments of the present invention provide a power load prediction method, an electronic device, an electronic apparatus, and a computer-readable storage medium, which can improve the accuracy of power load prediction.
One aspect of the present invention provides a power load prediction method, including:
searching a target factor set belonging to the same category as the factor set of the day to be predicted in the factor sets of the historical days, and taking the historical day where the target factor set is located as a similar historical day, wherein the factor set of any day comprises data of each load influence factor on the day;
taking the load data of the similar historical days and the load data of the days to be predicted before the prediction time point as original load sequences, and decomposing the original load sequences to obtain a plurality of load subsequences; and
predicting the power load of the day to be predicted after the prediction time point based on the load subsequence and the data of the load influencing factors at the same time point as the load subsequence.
Another aspect of the present invention also provides an electronic device, including:
the similar historical date determining module is used for searching a target factor set which belongs to the same category as the factor set of the date to be predicted in the factor sets of all the historical dates, and taking the historical date where the target factor set is located as the similar historical date, wherein the factor set of any date comprises data of all the load influencing factors on the date;
the sequence decomposition module is used for taking the load data of the similar historical days and the load data of the days to be predicted before the prediction time point as original load sequences, and decomposing the original load sequences to obtain a plurality of load subsequences; and
and the prediction module is used for predicting the load data of the day to be predicted after the prediction time point based on the load subsequence and the data of the load influence factors at the same time point with the load subsequence.
In another aspect, the present invention also provides an electronic device, which includes a processor and a memory, where the memory is used to store a computer program, and the computer program is executed by the processor to implement the method as described above.
In another aspect, the present invention also provides a computer-readable storage medium for storing a computer program, which when executed by a processor implements the method as described above.
In some embodiments of the application, data of each load influence factor in one day is used as a factor set, a target factor set belonging to the same category as the factor set of a day to be predicted is searched in the factor set of a historical day to determine a similar historical day, and the classification method takes each load influence factor as a whole, considers the influence of all the load influence factors on the power load as a whole, does not need to consider the influence of each load influence factor on the power load in sequence, and can reduce the classification category. Under the condition of less historical data, the target factor set which belongs to the same category as the factor set of the day to be predicted can meet the data volume required by prediction, and the prediction accuracy is improved.
Drawings
The features and advantages of the present invention will be more clearly understood by reference to the accompanying drawings, which are illustrative and not to be construed as limiting the invention in any way, and in which:
FIG. 1 illustrates a flow diagram of a power load prediction method provided by an embodiment of the present application;
FIG. 2 illustrates a block diagram of an electronic device provided by an embodiment of the present application;
fig. 3 shows a schematic diagram of an electronic device provided by an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings of the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
Please refer to fig. 1, which is a flowchart illustrating a power load prediction method according to an embodiment of the present disclosure. The power load prediction method can be applied to electronic equipment for scheduling and controlling a power system. Electronic devices include, but are not limited to, computers, controllers, servers, and the like. The power load prediction method comprises the following steps:
and S11, searching a target factor set which belongs to the same category as the factor set of the day to be predicted in the factor sets of the historical days, and taking the historical day where the target factor set is located as a similar historical day, wherein the factor set of any day comprises data of each load influence factor on the day.
In this embodiment, the load influencing factors include temperature, humidity, relative air pressure, wind speed, precipitation, radiation intensity, holidays, and seasons. The change of the load influencing factor can affect the power load. For example, the increase or decrease of the temperature may increase the usage demand of the cooling device or the heating device, thereby increasing the power load; the change of the illumination intensity can affect the use demand of people on the lighting equipment, thereby affecting the power load; changes in precipitation, humidity, relative air pressure and wind speed can affect the frequency of people's activities and thus the electrical load.
It will be appreciated that the load influencing factors may be selected based on the actual circumstances, and the above listed load influencing factors do not constitute a limitation of the present application. The present application takes the load influencing factors including temperature, humidity, relative air pressure, wind speed, precipitation, radiation intensity, holidays, seasons as examples for explanation.
In some embodiments, among the above load influencing factors, the holiday and the season may be a certain value for a particular day, i.e. whether the day is a holiday and the season in which the day is located, but the temperature, humidity, relative air pressure, wind speed, precipitation, and radiation intensity may be constantly changing during the day. In view of this, for the load influencing factors of which data are continuously changed during the day, the data of the load influencing factors may be acquired once every preset time interval (for example, 15 minutes) every day, then the data of each day is averaged, and the obtained average value is used as the data of the load influencing factors on the day.
Here, it should be noted that, for each history day, the data of the load influencing factors may be known since these history days have passed. However, for the predicted day, the predicted day may not be completely passed, so that data of the load influencing factors in the future time period of the predicted day cannot be acquired. In view of this, the data for the load influencing factors in the future time period of the predicted day can be obtained by adopting a prediction method. For example, through weather forecast, the temperature, precipitation and the like at each time point in the future time period of the day can be predicted.
Based on the above description, in some embodiments, one set of factors may be represented as follows:
X a ＝{x 1 ,x 2 ,...,x n }
wherein X represents the set of factors at day a, X 1 ～x n Data representing 1 st to n th load influencing factors on day a, i.e. the shadow of these loadsAnd averaging the data of the response factors at all time points in the day a.
In some embodiments, before searching for a target factor set belonging to the same category as the factor set of the day to be predicted in the factor sets of the historical days, cluster analysis may be performed on the factor sets of the historical days to divide the factor sets of the historical days into a plurality of factor set classes, and then, in the plurality of factor set classes, a target factor set class to which the factor set of the day to be predicted belongs may be determined, and the factor set in the target factor set class is taken as the target factor set.
When the cluster analysis is performed on the factor sets of the historical days, all the load influence factors in each factor set can be subjected to the cluster analysis as a whole, that is, only the influence on the power load under the integral action of all the load influence factors can be considered, and the influence on the power load by a single load influence factor does not need to be considered. One factor set class may include a plurality (at least 1) of factor sets. The sets of factors may have the same or similar impact on the electrical load for different sets of factors in the same set of factors class.
Correspondingly, when the target factor set class to which the factor set of the day to be predicted belongs is searched, all load influence factors in the factor set of the day to be predicted can be taken as a whole, and the influence on the power load under the integral action of the load influence factors can be considered. Meanwhile, the factor sets of the days to be predicted are classified into the factor set classes with similar influence on the power load. For example, assuming that the set of factors for the day to be predicted may cause the power load to vary along the characteristic curve 1, the set of factors in the factor set class a may cause the power load to vary along the characteristic curve 1, and the set of factors in the factor set class B may cause the power load to vary along the characteristic curve 2. The factor set for the day to be predicted may be divided into a factor set class a. And the factor set in the factor set class A is a target factor set which belongs to the same class as the factor set of the day to be predicted.
In some embodiments, when the above-mentioned clustering analysis is performed on the factor sets of the historical days, the factor sets of the historical days may be divided into multiple categories based on a K-means clustering algorithm. Specifically, the cluster analysis may be performed on the factor sets of the historical days according to the number of preset factor set classes (i.e., the cluster number). According to the conventional clustering principle, the closer the data distance between the same classes is, the better the data distance is, and the farther the data distance between different classes is, the better the data distance is. Whether the number setting of the factor set classes is proper or not can directly influence the clustering effect. In the present application, a plurality of candidate numbers may be set for the factor set class, and then a clustering score corresponding to each candidate number is calculated through a calinski _ harabasz index (also referred to as a CH index). The higher the clustering score is, the more appropriate the corresponding candidate number is. The candidate number with the highest clustering score may be used as the number of the preset factor set classes. The conventional technique is used for calculating the clustering score through the calinski _ harabasz index, and details are not repeated here.
In some embodiments, the performing the cluster analysis on the factor sets of the historical days may include:
randomly selecting a plurality of factor sets from the factor sets of the historical days as clustering centers respectively;
and for the factor sets which are not taken as the clustering centers, carrying out clustering analysis on the factor sets of the historical days according to Euclidean distances from the factor sets to the clustering centers.
Here, it is assumed that the set of cluster centers is C = { C 1 ，C 2 ，...，C k }，C 1 ～C k K factor sets selected from the factor sets of the historical days are represented, and each cluster center serves as the center of one factor set class.
In some embodiments, for a factor set that is not a cluster center in the factor sets of the history day, the euclidean distance of each factor set to the respective cluster center may be calculated based on expression (1):
wherein，L ij Representing the Euclidean distance from the ith factor set to the jth cluster center; x it A tth load influencing factor representing an ith set of factors; c jt The tth load influencing factor representing the jth cluster center. Here, assume that the 1 st factor set is associated with the cluster center C 1 ～C k In Euclidean distance, the 1 st factor set and the cluster center C 2 Has the smallest Euclidean distance, the 1 st factor set is divided into the cluster center C 2 A class of factor sets that are central. In this way, the factor sets of the historical days may be divided into a plurality of factor set classes.
After the classification is completed, the mean value of the factor sets in each factor set class can be calculated according to the expression (2), and the calculated mean value is updated to the clustering center of the corresponding factor set class.
Wherein, C l Represents the ith cluster center, | C l L represents the number of factor sets in the l factor set class, X i Representing the ith factor set in the ith factor set class.
And (3) repeatedly executing the steps corresponding to the expression (1) and the expression (2) until the clustering center position of each factor set class is kept unchanged. In some embodiments, the cluster center position representing the factor set class remains unchanged when the squared error function E, which may be calculated based on expression (3), converges to a constant value (minimum value).
Based on the above expressions (1) to (3), the factor sets of the history day may be divided into a plurality of factor set classes.
Based on the divided factor set classes, in some embodiments, the determining a target factor set class to which the factor set of the day to be predicted belongs in the multiple factor set classes may include:
and respectively calculating the Euclidean distance between the factor set of the day to be predicted and each factor set class, and taking the factor set class with the minimum Euclidean distance from the factor set of the day to be predicted as a target factor set class to which the factor set of the day to be predicted belongs.
The calculation of the euclidean distance can be referred to the expression (1), which is not described herein again.
Based on the above process of dividing factor set classes, it can be understood that the factor set in each factor set class may be multiple factor sets on continuous days, multiple factor sets on discontinuous days, or one factor set class may also include only one factor set. For example, it is assumed that the factor set class a includes a factor set of No. 3/month 1/2020, no. 3/month 3/2020, no. 3/month 8/2020, and No. 3/month 12/2020; the factor set class B includes factor sets of 3/5/2020, 3/6/2020, and 3/7/2020. And assuming that the target factor set class to which the factor set of the day to be predicted belongs is the factor set class A, the target factor sets belonging to the same class as the factor set of the day to be predicted are the factor sets of No. 3/1/2020, no. 3/8/2020, and No. 3/12/2020. Meanwhile, year 2020 No. 3/month 1, year 2020 No. 3/month 3, year 2020 No. 3/month 8, and year 2020 No. 3/month 12 are similar history days of the day to be predicted.
And S12, taking the load data of the similar historical days and the load data before the prediction time point of the day to be predicted as an original load sequence, and decomposing the original load sequence to obtain a plurality of load subsequences.
In some embodiments, the prediction time point is the current time of the day to be predicted. The time before the predicted time point is the time that has passed, and the time after the predicted time point is the time that has not yet arrived, i.e., the time in the future. For example, the predicted time point is 5 pm of the day to be predicted, the time before 5 pm (for example, 4 pm) is the elapsed time, and the time after 5 pm (for example, 8 pm) is the time that has not yet arrived. In combination with the above description, it can be understood that, for the day to be predicted, the load data before the prediction time point and the data of the load influencing factors may be obtained, while the data of the load influencing factors after the prediction time point may be obtained by prediction or the like, but the load data after the prediction time point may not be obtained. The load data predicted in the present application is power load data after the prediction time point. By predicting future power load conditions, the operation of the power system can be controlled to avoid over-or under-production of power.
In some embodiments, the load data of the similar historical days and the load data of the days to be predicted before the prediction time point may be data obtained according to a preset sampling frequency.
In the present embodiment, the sampling frequency is 15 minutes/time, i.e., load data is acquired every 15 minutes, and 15 minutes is a sampling period. The load data acquired each time is the average value of the power load in the current sampling period.
Taking the present embodiment as an example, there are 96 load data for each similar history day. For the day to be predicted, the number of load data is related to the prediction time point. In the original payload sequence, these payload sequences may be arranged in chronological order.
In some embodiments, the original load sequence may be decomposed based on an EEMD (Ensemble Empirical Mode Decomposition) algorithm. Specifically, the original payload sequence that is discontinuous in the time domain may be converted into the original payload signal X (t) that is continuous in the time domain, and the original payload sequence may be decomposed by decomposing the original payload signal. This process may be as follows:
based on the set global average number of times N, adding the noise signal ω (t) with the standard normal distribution to the original load signal X (t) to obtain a new signal X' (t), as shown in expression (4):
X i ′(t)＝X(t)+ω i (t)，i＝1，2，...，N (4)
where i represents the i-th addition of white noise.
EMD decomposition is carried out on the new signal X' (t) to obtain IMF (Intrinsic Mode function) components of each order. As shown in expression (5):
wherein J represents the number of IMF components obtained by decomposition after the ith white noise is added, and r j (t) represents the residual component after decomposition, c j (t) represents an IMF component.
Repeatedly executing the steps corresponding to the expressions (4) and (5) for N times, adding white noise with the same intensity and different sequences every time, and finally obtaining X i ' (t) is as shown in expression (6).
Averaging the IMFs to obtain a final IMF component c by using the principle that the average value of a white noise frequency spectrum is zero j (t) as shown in expression (7)
For IMF component c j And (t) sampling to obtain the load subsequence obtained by decomposing the original load sequence.
In the application, the original load sequence is decomposed into a plurality of sub-load sequences, so that the complexity of the original load sequence can be reduced, and the accuracy of power load prediction can be improved.
In some embodiments, in view that the partial payload subsequence obtained by decomposition may have a small correlation with the original payload sequence, to reduce data processing amount, after decomposing the original payload sequence, the method of the present application may further include:
performing correlation calculation on each load subsequence and the original load sequence to obtain a correlation coefficient between each load subsequence and the original load sequence;
and eliminating the invalid load sub-sequences of which the correlation coefficients with the original load sequences are smaller than the threshold value of the correlation coefficients, and taking the load sub-sequences after the invalid load sub-sequences are eliminated as the load sub-sequences of the original load sequences.
In some embodiments, the above-mentioned correlating each load subsequence with the original load sequence separately may be to calculate the IMF component c corresponding to each load subsequence j (t) performing a correlation calculation with the original load signal X (t).
Specifically, the IMF component c may be calculated based on expression (8) j (t) correlation coefficient with the original load signal y (t):
wherein r is j The correlation coefficient of the jth IMF component and the original load signal y (t) is also the correlation coefficient of the jth load subsequence and the original load sequence;
n is the number of sample points of the signal.
In some embodiments, the correlation coefficient threshold may be obtained based on expression (9):
wherein, TH is a correlation coefficient threshold, and n represents the number of IMF components.
If the correlation coefficient r of the jth load subsequence with the original load sequence j If the correlation coefficient is greater than or equal to the correlation coefficient threshold TH, the jth load subsequence is reserved; if the correlation coefficient r of the jth load subsequence with the original load sequence j And if the correlation coefficient is smaller than the correlation coefficient threshold TH, rejecting the jth load subsequence.
After the invalid payload sub-sequence is eliminated, step S13 may be performed based on the remaining payload sub-sequences.
And step S13, predicting the power load after the prediction time point on the day to be predicted based on the load subsequence and the data of the load influence factors at the same time point with the load subsequence.
In some embodiments, when the load data is collected, the data of the load influencing factors may be collected at the same time, and the load data and the data of the load influencing factors at each sampling time point are obtained.
In some embodiments, predicting load data after the prediction time point for a day to be predicted comprises:
for any load subsequence, inputting the load subsequence and data of load influence factors at the same time point with the load subsequence into a prediction model corresponding to the load subsequence, and predicting to obtain a load component based on the load subsequence, wherein at least part of different load subsequences correspond to different prediction models;
and performing superposition reconstruction on the load components of each load subsequence to obtain the power load of the day to be predicted after the prediction time point.
The prediction model corresponding to each load subsequence can be obtained by training based on historical data, and the training process is similar to the using process of the prediction model and is not described herein any more.
In the present embodiment, each load subsequence corresponds to a prediction model. According to the data characteristics of each load subsequence, each prediction model can have different network structures, so that better prediction can be performed based on the load subsequences, and the prediction accuracy is improved.
In some embodiments, the predictive model may be a LSTNet model. The LSTNet model may consist of a nonlinear neural network and a linear neural network. The nonlinear neural network can comprise a CNN convolution layer, an RNN circulation layer and a jump layer; the linear neural network may include an autoregressive linear layer. The CNN convolutional layer can extract local dependency relationship between load data, and the RNN cyclic layer can capture long-term dependency relationship of load subsequences. The jump layer can capture an ultra-long-term dependence mode by utilizing the characteristic that the load subsequence has periodicity, and the optimization process is simplified.
In some embodiments, a traditional autoregressive linear model can be added on the basis of a nonlinear neural network, so that the nonlinear neural network has stronger robustness on the subload sequence with the scale violation change.
In some embodiments, the results of the nonlinear neural network and the linear neural network are reconstructed and superimposed, so as to obtain a load component corresponding to the load subsequence.
In some embodiments, there may be more data in a single load subsequence, which may present a problem of being computationally expensive when predicting through a prediction model. To solve the problem, for any load subsequence, when the load subsequence and the data of the load influencing factors at the same time point as the load subsequence are input into the prediction model corresponding to the load subsequence, the target load data of the load subsequence in a preset time period and the data of the load influencing factors at the same time point as the target load data can be input into the prediction model corresponding to the load subsequence.
Specifically, the load subsequence in a preset time period before the prediction time point and the data of the load influencing factors at the same time point as the load sequence can be input into the prediction model by taking the prediction time point as a reference. In the present embodiment, target load data 16 hours before the prediction time point in the load sub-sequence, and data of load influencing factors at the same time point as the target load data are input into the prediction model. Here, it should be noted that, since the load data included in one load sub-sequence may be the load data of discontinuous days, the preset time period may not be continuous in time. In this case, the preset period is constituted by a plurality of sub-periods, and the plurality of periods may not be completely continuous in time. For example, it is assumed that the to-be-detected day is 13/3/2020, the detection time point is 8 am of 13/3/2020, and the similar history days of the to-be-detected day are No. 3/1/2020, no. 3/2020, no. 3/8/2020, and No. 3/12/2020. Then the detection time point is 16 hours before 0 to 8 o ' clock (8 hours) on day 13 on month 3 of 2020, and 16 o ' clock to 24 o ' clock (8 hours) on day 12 on month 3 of 2020. According to the sampling frequency of 15 minutes/time, 64 pieces of target load data and data of load influence factors corresponding to the 64 pieces of target load data exist in the preset time period. The prediction model predicts the load components after the detection time point based on the 64 pieces of target load data and the data of the load influencing factors corresponding to the 64 pieces of target load data. Therefore, the data processing amount can be effectively reduced, the operation efficiency is improved, and the hardware requirement during the operation of the prediction model is reduced.
In some embodiments, the load component predicted by the predictive model is a load component at a plurality of time points within the prediction time period after the predicted time point. And building and superposing the load components output by each prediction model, so that the power loads of a plurality of time points in the prediction time period can be obtained.
In the present embodiment, the predicted power load is a power load at a plurality of time points within 4 hours after the predicted time point. Wherein the interval between adjacent time points may be 15 minutes. Thus, the power load conditions at 16 time points within 4 hours in the future can be predicted, the power generation amount of the power system can be controlled in advance, and the situations of excessive power generation or insufficient power generation can be prevented.
In some embodiments of the application, data of each load influence factor in one day is used as a factor set, a target factor set belonging to the same category as the factor set of a day to be predicted is searched in the factor set of a historical day to determine a similar historical day, and the classification method takes each load influence factor as a whole, considers the influence of all the load influence factors on the power load as a whole, does not need to consider the influence of each load influence factor on the power load in sequence, and can reduce the classification category. Under the condition of less historical data, the target factor set which belongs to the same category as the factor set of the day to be predicted can be ensured to meet the data volume required by prediction, and the prediction accuracy is improved. Meanwhile, in the training stage of the prediction model, historical data are divided into a plurality of classes according to the method, the number of the divided classes is small, the number of the data in each class is large, and when the prediction model is trained on the basis of the historical data in each class, the number of the training data is large, the precision of the trained prediction model is high, and therefore the precision of power load prediction is improved.
In addition, compared with the technology that the original load sequence is directly input into the prediction model for prediction, the load component is predicted through different prediction models for each load subsequence obtained through decomposition after the original load sequence is decomposed, so that the complexity of the original load sequence is reduced, meanwhile, the load component can be predicted through the prediction models with different network structures according to the data characteristics of each load subsequence, and the prediction accuracy is improved.
Please refer to fig. 2, which is a block diagram of an electronic device according to an embodiment of the present disclosure. The electronic device includes:
the similar historical date determining module is used for searching a target factor set which belongs to the same category as the factor set of the date to be predicted in the factor sets of the historical dates, and taking the historical date where the target factor set is located as the similar historical date, wherein the factor set of any date comprises data of each load influence factor on the date;
the sequence decomposition module is used for taking the load data of the similar historical days and the load data before the prediction time point of the day to be predicted as an original load sequence, and decomposing the original load sequence to obtain a plurality of load subsequences; and
and the prediction module is used for predicting the load data of the day to be predicted after the prediction time point based on the load subsequence and the data of the load influence factors at the same time point with the load subsequence.
Please refer to fig. 3, which is a schematic diagram of an electronic device according to an embodiment of the present application. The electronic device comprises a processor and a memory for storing a computer program which, when executed by the processor, implements the power load prediction method described above.
The processor may be a Central Processing Unit (CPU). The Processor may also be other general purpose processors, digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components, or a combination thereof.
The memory, which is a non-transitory computer-readable storage medium, may be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as program instructions/modules corresponding to the methods in the embodiments of the present invention. The processor executes various functional applications and data processing of the processor by executing non-transitory software programs, instructions and modules stored in the memory, that is, the method in the above method embodiment is realized.
The memory may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created by the processor, and the like. Further, the memory may include high speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, the memory optionally includes memory located remotely from the processor, and such remote memory may be coupled to the processor via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
An embodiment of the present application further provides a computer-readable storage medium for storing a computer program, which when executed by a processor, implements the power load prediction method described above.
Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations fall within the scope defined by the appended claims.
Claims (10)
1. A method of predicting a power load, the method comprising:
searching a target factor set which belongs to the same category as the factor set of the day to be predicted in the factor sets of the historical days, and taking the historical day where the target factor set is located as a similar historical day, wherein the factor set of any day comprises data of each load influence factor in the day;
taking the load data of the similar historical days and the load data of the days to be predicted before the prediction time point as original load sequences, and decomposing the original load sequences to obtain a plurality of load subsequences; and
predicting the power load of the day to be predicted after the prediction time point based on the load subsequence and the data of the load influencing factors at the same time point as the load subsequence.
2. The method of claim 1, wherein prior to finding a target set of factors that belong to the same category as the set of factors for the day to be predicted, the method further comprises:
performing cluster analysis on the factor sets of the historical days to divide the factor sets of the historical days into a plurality of factor set classes;
in the factor sets of each historical day, searching a target factor set belonging to the same category as the factor set of the day to be predicted comprises:
and determining a target factor set class to which the factor set of the day to be predicted belongs in the plurality of factor set classes, and taking the factor set in the target factor set class as the target factor set.
3. The method of claim 2, wherein the performing cluster analysis on the set of factors for the historical day comprises:
randomly selecting a plurality of factor sets from the factor sets of the historical days to be used as clustering centers respectively;
and for the factor sets which are not taken as the clustering centers, carrying out clustering analysis on the factor sets of the historical days according to Euclidean distances from the factor sets to the clustering centers.
4. The method as claimed in claim 2, wherein the determining the target factor set class to which the factor set of the day to be predicted belongs, among the plurality of factor set classes, comprises:
and respectively calculating the Euclidean distance between the factor set of the day to be predicted and each factor set class, and taking the factor set class with the minimum Euclidean distance from the factor set of the day to be predicted as a target factor set class to which the factor set of the day to be predicted belongs.
5. The method of claim 1, wherein after said decomposing said original payload sequence, said method further comprises:
performing correlation calculation on each load subsequence and the original load sequence to obtain a correlation coefficient between each load subsequence and the original load sequence;
and eliminating an invalid load subsequence of which the correlation coefficient with the original load sequence is smaller than a correlation coefficient threshold value, and taking the load subsequence after eliminating the invalid load subsequence as the load subsequence of the original load sequence.
6. The method of claim 1, wherein said predicting the electrical load of the day to be predicted after the predicted point in time comprises:
for any one of the load subsequences, inputting the load subsequences and the data of the load influence factors at the same time point as the load subsequences into a prediction model corresponding to the load subsequences, and predicting to obtain a load component based on the load subsequences, wherein at least part of different load subsequences correspond to different prediction models;
and performing superposition reconstruction on the load components of each load subsequence to obtain the power load of the day to be predicted after the prediction time point.
7. The method as claimed in claim 6, wherein for any of the load subsequences, inputting the load subsequences and the data of the load influencing factors at the same time point as the load subsequences into the prediction model corresponding to the load subsequences comprises:
and inputting the target load data of the load subsequence in a preset time interval and the data of the load influence factors at the same time point with the target load data into a prediction model corresponding to the load subsequence.
8. An electronic device, comprising:
the similar historical date determining module is used for searching a target factor set which belongs to the same category as the factor set of the date to be predicted in the factor sets of all the historical dates, and taking the historical date where the target factor set is located as the similar historical date, wherein the factor set of any date comprises data of all the load influencing factors on the date;
the sequence decomposition module is used for taking the load data of the similar historical days and the load data of the days to be predicted before the prediction time point as original load sequences, and decomposing the original load sequences to obtain a plurality of load subsequences; and
and the prediction module is used for predicting the load data of the day to be predicted after the prediction time point based on the load subsequence and the data of the load influence factors at the same time point with the load subsequence.
9. A computer-readable storage medium, characterized in that the computer-readable storage medium is used for storing a computer program which, when executed by a processor, implements the method of any one of claims 1 to 7.
10. An electronic device, characterized in that the electronic device comprises a processor and a memory for storing a computer program which, when executed by the processor, implements the method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211026800.1A CN115392567A (en) | 2022-08-25 | 2022-08-25 | Power load prediction method, electronic equipment, device and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211026800.1A CN115392567A (en) | 2022-08-25 | 2022-08-25 | Power load prediction method, electronic equipment, device and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115392567A true CN115392567A (en) | 2022-11-25 |
Family
ID=84122007
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211026800.1A Pending CN115392567A (en) | 2022-08-25 | 2022-08-25 | Power load prediction method, electronic equipment, device and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115392567A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116258280A (en) * | 2023-05-12 | 2023-06-13 | 国网湖北省电力有限公司经济技术研究院 | Short-term load prediction method based on time sequence clustering |
-
2022
- 2022-08-25 CN CN202211026800.1A patent/CN115392567A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116258280A (en) * | 2023-05-12 | 2023-06-13 | 国网湖北省电力有限公司经济技术研究院 | Short-term load prediction method based on time sequence clustering |
| CN116258280B (en) * | 2023-05-12 | 2023-08-11 | 国网湖北省电力有限公司经济技术研究院 | Short-term load prediction method based on time sequence clustering |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Sun et al. | Using Bayesian deep learning to capture uncertainty for residential net load forecasting | |
| CN110928993A (en) | User position prediction method and system based on deep cycle neural network | |
| CN112686464A (en) | Short-term wind power prediction method and device | |
| CN111382906B (en) | Power load prediction method, system, equipment and computer readable storage medium | |
| CN112100911B (en) | Solar radiation prediction method based on depth BILSTM | |
| CN112396234A (en) | User side load probability prediction method based on time domain convolutional neural network | |
| CN113240153A (en) | Photovoltaic power generation data prediction method and device, computing equipment and storage medium | |
| Yang et al. | Prediction method of PM2. 5 concentration based on decomposition and integration | |
| Liao et al. | Ultra-short-term interval prediction of wind power based on graph neural network and improved bootstrap technique | |
| CN116663746A (en) | Power load prediction method and device, computer equipment and storage medium | |
| CN115392567A (en) | Power load prediction method, electronic equipment, device and readable storage medium | |
| CN115659411A (en) | Method and device for data analysis | |
| CN111785093A (en) | Air traffic flow short-term prediction method based on fractal interpolation | |
| CN117613883A (en) | Method and device for predicting generated power, computer equipment and storage medium | |
| CN113128666A (en) | Mo-S-LSTMs model-based time series multi-step prediction method | |
| CN113380340A (en) | Training method and device of fly ash concentration prediction model and computer equipment | |
| Nawar et al. | Transfer learning in deep learning models for building load forecasting: Case of limited data | |
| CN117350336A (en) | Charging load prediction and model training method, device, equipment and storage medium | |
| CN117196105A (en) | People number prediction method, device, computer equipment and storage medium | |
| CN116341720A (en) | Multi-fan wind speed and direction prediction method based on dynamic graph convolution and transformation | |
| CN115907131A (en) | Method and system for building electric heating load prediction model in northern area | |
| CN116578858A (en) | Air compressor fault prediction and health degree evaluation method and system based on graphic neural network | |
| CN115965160A (en) | Data center energy consumption prediction method and device, storage medium and electronic equipment | |
| CN114139783A (en) | Wind power short-term power prediction method and device based on nonlinear weighted combination | |
| CN113723670A (en) | Photovoltaic power generation power short-term prediction method with variable time window |
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 |