CN112949207A - Short-term load prediction method based on improved least square support vector machine - Google Patents
Short-term load prediction method based on improved least square support vector machine Download PDFInfo
- Publication number
- CN112949207A CN112949207A CN202110321179.0A CN202110321179A CN112949207A CN 112949207 A CN112949207 A CN 112949207A CN 202110321179 A CN202110321179 A CN 202110321179A CN 112949207 A CN112949207 A CN 112949207A
- Authority
- CN
- China
- Prior art keywords
- support vector
- vector machine
- prediction
- improved
- square support
- 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
- 238000012843 least square support vector machine Methods 0.000 title claims abstract description 58
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 58
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 241000256844 Apis mellifera Species 0.000 claims description 68
- 241000257303 Hymenoptera Species 0.000 claims description 25
- 238000012706 support-vector machine Methods 0.000 claims description 18
- 235000012907 honey Nutrition 0.000 claims description 13
- 235000013305 food Nutrition 0.000 claims description 11
- 238000011156 evaluation Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 6
- 238000010187 selection method Methods 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000013499 data model Methods 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 2
- 230000035772 mutation Effects 0.000 claims 1
- 238000005070 sampling Methods 0.000 claims 1
- 238000009826 distribution Methods 0.000 abstract description 12
- 230000008901 benefit Effects 0.000 abstract description 8
- 230000006870 function Effects 0.000 description 14
- 239000000523 sample Substances 0.000 description 10
- 230000007547 defect Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 201000004569 Blindness Diseases 0.000 description 1
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 1
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 239000013068 control sample Substances 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000013213 extrapolation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 210000003734 kidney Anatomy 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000009394 selective breeding Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/08—Probabilistic or stochastic CAD
Abstract
The invention relates to a short-term load prediction method based on an improved least square support vector machine, which comprises the following steps: step 1: collecting historical load data, and selecting the historical load data by using a similar daily method to obtain a training sample; step 2: establishing a prediction model of a least square support vector machine to carry out load prediction on a training sample, and when the accuracy of a prediction result is judged to be not up to standard, improving the prediction model of the least square support vector machine by utilizing an improved artificial bee colony algorithm to obtain the improved prediction model of the least square support vector machine; and step 3: and inputting the training sample into a prediction model of the improved least square support vector machine for load prediction and error analysis, and outputting a prediction result. Compared with the prior art, the method has the advantages that the numerical value of the load is accurately predicted before the electric quantity scheduling of the power distribution network, the network loss is favorably reduced, the redundant electric quantity and the economic loss are reduced, the safety of the line is enhanced, and the like.
Description
Technical Field
The invention relates to the technical field of distribution network load prediction, in particular to a short-term load prediction method based on an improved least square support vector machine.
Background
Distribution network load prediction techniques fall into three broad categories. The first type is a traditional load prediction method, such as a load derivation method, a similar daily method, a kalman filtering method, an exponential smoothing method, and the like. The second type is a classic load prediction method, which mainly comprises a time series method, a regression analysis method, a trend extrapolation method and the like. The third kind of methods are intelligent methods, and with the continuous application of a series of intelligent algorithms such as artificial neural networks, support vector machines and the like in recent years, more and more intelligent algorithms are developing the way in load prediction. The load prediction of the intelligent algorithm mainly adopts the methods of an artificial neural network method, a support vector machine method, a fuzzy prediction method and the like.
The first type of traditional load prediction method has the defects of large and tedious calculated amount, low prediction precision, incomplete consideration and the like in the traditional technology. And other coupling factors cannot be considered, or the method has the limitation that larger deviation exists in the predicted result. The second type is a classic load prediction method, and a time series method is mainly used. The time sequence method has the advantages that the time sequence relevance of the load data is considered, and the defect is that the fitting capacity of the nonlinear relation of the load data is limited. The third type is an intelligent algorithm, and the intelligent algorithm applied to the load prediction of the power distribution network mainly comprises a support vector machine method, a neural network method and the like. The neural network method has the defects of difficult problems of model structure building, learning speed optimization, local minimum points and the like. The support vector machine prediction method has the outstanding advantages of strong generalization capability, global optimization, high calculation speed and the like. However, there are large artifacts in the selection of the optional parameters and the kernel function, which have a great influence on the prediction accuracy. The support vector machine needs to select parameters through other intelligent algorithms.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a short-term load prediction method based on an improved least square support vector machine, and in order to promote the construction of a smart grid, the prediction of loads expected to occur in the future in a power system is also important. The available generating capacity in the system should meet the load requirements in the system at any time under normal operating conditions. If the load prediction is low, the actual power grid cannot meet the power supply requirement, even the power grid may lack power, and necessary measures should be taken to increase the power generation capacity; if the load forecast is too high, it will result in installing too many power generation equipment that can not be fully utilized, thereby causing waste of investment. Therefore, the load prediction of the power distribution network is a premise and a basis for effectively making a power distribution plan, reasonably arranging the start, stop and maintenance of a generator set and effectively ensuring the stable operation of a power system, and is also indispensable work for safe production, reliable operation and planning and scheduling of the power distribution network. The accurate load prediction can avoid major accidents, guarantee the safety of production and life, save economy and improve social benefits.
The purpose of the invention can be realized by the following technical scheme:
a short-term load prediction method based on an improved least square support vector machine comprises the following steps:
step 1: collecting historical load data, and selecting the historical load data by using a similar daily method to obtain a training sample;
step 2: establishing a prediction model of a least square support vector machine to carry out load prediction on a training sample, and when the accuracy of a prediction result is judged to be not up to standard, improving the prediction model of the least square support vector machine by utilizing an improved artificial bee colony algorithm to obtain the improved prediction model of the least square support vector machine;
and step 3: and inputting the training sample into a prediction model of the improved least square support vector machine for load prediction and error analysis, and outputting a prediction result.
Further, the step 1 comprises the following sub-steps:
step 101: collecting historical load data comprising weather types, date types and date differences;
step 102: preprocessing abnormal or missing load data in the acquired data;
step 103: and selecting a proper evaluation function for evaluating the similarity between the historical days and the prediction days according to the preprocessed data, and selecting the data of the historical days with higher similarity as training samples.
Further, in step 103, the evaluation function selected by the degree of similarity between the evaluation historical day and the prediction day has a corresponding mathematical formula:
in the formula, rijFor the degree of similarity between historical and predicted days, XikFor feature quantities quantized for historical day i, XjkThe characteristic vector of the day j to be measured, m is the characteristic vector, and k is a natural number.
Further, the step 2 comprises the following sub-steps:
step 201: establishing a prediction model of a least square support vector machine to carry out load prediction on a training sample, and judging whether the precision of a prediction result reaches the standard or not;
step 202: when the precision is not up to the standard, improving the artificial bee colony algorithm, and optimizing the parameters of the least square support vector machine by using the improved artificial bee colony algorithm;
step 203: and establishing a corresponding model based on the optimized parameters of the least square support vector machine.
Further, in the artificial bee colony algorithm improved in step 201, the random search equation of the collected bees is improved by using the variation operation in the differential evolution algorithm, and the improved search equation corresponds to:
vij＝xbest，j+φij(xr1，j-xr2，j)
in the formula, vijAs a new source of honey, xbest，jThe j weft value phi of the optimal honey sourceijIs [ -1, 1 [ ]]Random number between, xr1，jAnd xr2，jIs two different from iAnd (4) carrying out random individuals.
Further, in step 201, the method of determining whether the accuracy of the prediction result meets the standard is determined by using a relative error.
Further, in the prediction model of the least square support vector machine in step 201, the kernel function of the prediction model adopts a radial basis function, and the corresponding mathematical description formula is as follows:
wherein ε is a nuclear parameter.
Further, in the prediction model of the least square support vector machine in step 201, the regression constraint condition is:
wherein | | w | | non-conducting phosphor2To control the complexity of the model, eiError of sample data prediction for the trained data model, C is regularization parameter, b is constant, yiIn order to output the data, the data is output,
for mapping from a low-dimensional space to a high-dimensional space, W is a weight vector.
Further, in the artificial bee colony algorithm modified in step 201, the bee collecting unit transmits the collected bee source information to the observing bees, and the observing bees use the wheel selection method to have a probability piSelecting food sources, searching for new solutions around the food sources, calculating fitness values, and greedy selecting, wherein the probability piThe mathematical description formula of (a) is:
in the formula, fitiTo solve XiThe fitness value of (1, 2., SN), SN is the number of solutions in the population.
Further, in the artificial bee colony algorithm improved in step 201, if the optimal value of the honey bee is not found after the honey bee is searched for many times, the honey bee is changed into a scout bee, and a new solution is randomly generated to replace an old solution according to a corresponding formula, wherein the corresponding formula is as follows:
xij＝ximin+rand(0，1)(ximax-ximin)
in the formula, ximaxAnd ximinRepresenting the maximum and minimum around the old solution.
Compared with the prior art, the invention has the following advantages:
(1) for the problem of parameter selection of the least square support vector machine, the method overcomes the error existing in artificial blind selection, and optimizes the parameters by adopting an improved artificial bee colony algorithm. The accuracy of the parameter selection is related to the accuracy of the load prediction.
(2) For the fact that the historical load data are numerous and the influence factors on load prediction are numerous, the load data are selected by a similar day selection method, the load data with the date with high similarity are selected as training samples of the load prediction, and the accuracy of the load prediction is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a schematic diagram of the main steps of load prediction in an embodiment of the present invention;
FIG. 2 is a flowchart of load prediction for a least squares support vector machine according to an embodiment of the present invention;
FIG. 3 is a flow chart of the load prediction of the improved least squares support vector machine in an embodiment of the present invention;
FIG. 4 is a comparison graph of load predictions for different algorithms in an embodiment of the present invention;
FIG. 5 is a graph comparing error values for different algorithms in an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, shall fall within the scope of protection of the present invention.
Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings or the orientations or positional relationships that the products of the present invention are conventionally placed in use, and are only used for convenience in describing the present invention and simplifying the description, but do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus, should not be construed as limiting the present invention. Furthermore, the terms "first," "second," "third," and the like are used solely to distinguish one from another and are not to be construed as indicating or implying relative importance.
Furthermore, the terms "horizontal", "vertical" and the like do not imply that the components are required to be absolutely horizontal or pendant, but rather may be slightly inclined. For example, "horizontal" merely means that the direction is more horizontal than "vertical" and does not mean that the structure must be perfectly horizontal, but may be slightly inclined.
In the description of the present invention, it should also be noted that, unless otherwise explicitly specified or limited, the terms "disposed," "mounted," "connected," and "connected" are to be construed broadly and may, for example, be fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
The invention predicts the short-term load of the distribution network. A method for optimizing a least squares support vector machine (IABC-LS-SVM) by using an improved artificial bee colony algorithm is adopted. Because the artificial bee colony algorithm has the defects of low convergence speed, easy trapping in local optimum and poor exploitation capability, the search mode of the artificial bee colony algorithm is improved, then the improved artificial bee colony algorithm is used for optimizing the parameters of the least square support vector machine, and finally a load prediction model is established based on the optimized least square support vector machine to predict the short-term load of the power distribution network in a certain place. Due to a plurality of factors influencing load prediction, collected data are preprocessed, historical load data are selected by using a similar daily method, training samples are formed, and load prediction is carried out.
The specific embodiment is as follows:
the specific main steps of the load prediction of the least square support vector machine introduced with the improved artificial bee colony algorithm are shown in fig. 1:
(1) historical load data, weather type, date type and date difference are collected.
(2) Abnormal or missing load data is analyzed and filled in, and the weather type is normalized as shown in the following table 1.
TABLE 1 weather type normalization
| Weather type | All-weather | Cloudy | Yin (kidney) | Rain water | Other bad weather |
| Normalizing numerical values | 0.9 | 0.7 | 0.5 | 0.3 | 0.1 |
The specification processing for the date type is shown in table 2 below:
TABLE 2 date type values
| Monday | Tuesday | Wednesday | Thursday | Friday of week | Saturday | Sunday |
| 0.7 | 0.8 | 0.8 | 0.8 | 0.8 | 0.4 | 0.3 |
The date is poor, and the principle of' big and small is met. The closer the history day is to the prediction day, the greater the influence on the prediction day. The similarity formula for date difference is as follows:
in the formula: m isiSimilarity of day differences on day i; beta is an attenuation coefficient, and the attenuation coefficient is reduced when the date difference is increased by one day, and the value is between 0.90 and 0.98. k is the number of days of difference between the historical day and the predicted day, and alpha is the lowest similarity.
(3) The similar day method is that factors influencing the load form a day feature vector, then a proper evaluation function is selected to evaluate the similarity between the historical day and the predicted day, and the data of the historical day with higher similarity is selected as a training sample. The similarity is calculated as shown below.
In the formula: xi＝[Xi1，Xi2，…，Xim]And (4) quantifying the characteristic quantity of the historical day i. Xj＝[Xj1，Xj2，…，Xjm]The feature vector of the day to be measured. m is the number of features. The daily feature vector is: weather type, date difference, etc. The larger the similarity value is, the higher the similarity is, and the larger the adoption rate is.
(4) A model of a least squares support vector machine was established, the prediction flow chart of which is shown in fig. 2.
The least square support vector machine (LS-SVM) is an extension and improvement of a standard Support Vector Machine (SVM), the standard support vector machine solves a quadratic programming problem with inequality constraint conditions, the least square support vector machine is further improved on the basis of the standard support vector machine, inequality constraint in quadratic programming processing in the standard support vector machine is converted into equality constraint, a plurality of uncertain influence factors are reduced, and the solution of the quadratic programming problem is converted into the solution of a linear equation set. Compared with a standard support vector machine, the least square support vector machine has fewer parameters for selection and lower computational complexity.
The basic principle of LS-SVM is: the load data of the training sample is (x)i，yi)，i＝1，2，…，N。xiTo input data, yiFor output data, N is the total number of samples. Mapping the training samples to a high-dimensional feature space through a nonlinear mapping phi (x), and then performing linear regression in the high-dimensional feature space, wherein the regression function is as follows:
y＝f(x)＝w·φ(x)+b
in the formula: w is the weight vector, b is a constant, and φ (x) is a mapping from a low-dimensional space to a high-dimensional space.
According to the risk minimization principle, the regression constraint conditions of the LS-SVM are as follows:
in the formula: | w | non-woven phosphor2To control the complexity of the model; e.g. of the typeiError of sample data prediction for the trained data model; and C is a regularization parameter and represents the punishment degree of the control sample exceeding the error.
In order to solve the constraint condition conveniently, a Lagrangian function is introduced, and the constraint problem is changed into an unconstrained problem:
from KKT condition on w, b, ei、aiThe partial derivative is zero, and the linear equation system is obtained as follows:
elimination of w and eiThen the linear equation is as follows:
wherein a ═ a1，a2，…，aN]T；E＝[1，1，…，1]T；Y＝[y1，y2，…，yn]T。
Where K is a kernel function satisfying the Mercer condition, i.e.
Representing non-linearities from an input space to a high-dimensional feature spaceAnd (6) mapping.
The kernel function has many options, and is more commonly: radial Basis Functions (RBF), fourier kernel functions, etc. The method selects the radial basis function as the kernel function, and has the advantages of wide convergence domain, strong adaptability to different samples and the like. The expression is as follows:
where ε is the kernel width coefficient. The final prediction model of the LS-SVM is:
the regularization parameter C and the kernel parameter epsilon in the least square support vector method are 2 important parameters related to the accuracy of the load prediction model, and too large or too small data can affect the prediction accuracy. In the past, parameters are valued by an artificial experience selection method, but the artificial selection can generate larger errors and influence the accuracy of an LS-SVM prediction model, so that the parameters are selected by an intelligent algorithm with better performance. Because the artificial bee colony algorithm has the advantages of less setting parameters, simple algorithm, strong global search capability, high convergence speed and the like, the artificial bee colony algorithm is adopted to improve the parameters of the prediction model. However, the artificial bee colony algorithm has disadvantages, so the artificial bee colony algorithm is improved firstly.
(5) And optimizing parameters of the least square support vector machine by using an improved artificial bee colony algorithm.
An artificial bee colony Algorithm (ABC) is an intelligent optimization algorithm for simulating the bee honey collection behavior. The algorithm has the advantages of few control parameters, easiness in implementation and simplicity in calculation, and is applied to multiple fields.
In the artificial bee colony algorithm, an artificial bee colony comprises 3 components: bee picking, bee observation and bee reconnaissance. The main tasks of the bees are to explore and develop food sources, go out to search honey sources, and share the information of the food sources with other bees. The main task of observing bees is to wait for the return of bees in the bee nest, not only share the information of food sources from the bees, but also select the food sources with a certain selection probability by utilizing a strategy and mine the surrounding. If some bees can not search better bee sources, the bees are changed into scout bees, new better food sources are searched near the bee nest to replace the original worse food sources, and the number of the scout bees is generally 5% -10% of the number of bee colonies.
The artificial bee colony algorithm mainly comprises the following steps:
1) and (3) collecting the searching behavior of the bees: the bee is adopted to search in the original honey source field according to the following formula to generate a new solution vijAnd calculating its fitness value to solve x for the initial solutionijAnd new solution vijA greedy selection is performed.
vij＝xij+φij(xij-xkj)
In the formula: x is the number ofijThe j dimension value represents the ith honey source, i ═ 1,2, …, SN, and j ═ 1,2, 3, …, D. k represents a source of honey different from source i. Phi is aijIs [ -1, 1 [ ]]Is responsible for xijGeneration of a source location within a neighborhood. Comparing the position of the original honey source with the random honey source in the neighborhood to obtain a new honey source vijThe position of (a).
2) Roulette selection method for observing bees: the collected bee source information is transmitted to the observation bees by the bee collecting bees, and the observation bees use a roulette wheel selection method to obtain the probability piFood sources are selected, a new solution is searched around the food sources, fitness values are calculated, and greedy selection is performed.
In the formula: fitiTo solve XiThe fitness value of (1,2, …, SN), SN is the number of solutions in the population.
3) Generation of random solutions for scout bees: and if the optimal value of the honey bee is not found after the honey bee is searched for many times, the honey bee is changed into a scout bee, and a new solution is randomly generated to replace the old solution according to the following formula.
xij＝ximin+rand(0，1)(ximax-ximin)
In the formula: x is the number ofimaxAnd ximinRepresenting the maximum and minimum around the old solution.
The artificial bee colony Algorithm (ABC) has good global search capability, but the later local search capability is weak and the convergence speed is slow. Inspired by the idea of differential evolution, the random search equation of the bee is improved by using the variation operation in the differential evolution algorithm, and the improved search equation is as follows:
vij＝xbest，j+φij(xr1，j-xr2，j)
in the formula: r1 and r2 are two random individuals different from i, phiijIs [ -1, 1 [ ]]A random number in between.
In the improved search equation, a new candidate solution is obtained by extracting a new search area near the optimal solution obtained by the last iteration, the information of the optimal solution is used for guiding the search of the solution to improve the extraction capacity of the solution, and two individuals are randomly selected to make a difference to guide the bees to develop and explore the optimal area as much as possible so as to improve the development capacity and the search speed of the bees.
(6) A final load prediction model is determined.
(7) And performing load prediction and error analysis, and outputting prediction data.
The load prediction flow chart of the improved least squares support vector machine is shown in fig. 3.
Example analysis
1. Error index of load prediction
Through error calculation between the actual value and the predicted value, the accuracy of the load prediction model can be visually seen. The analysis of the error is of great significance to prove the effectiveness of the load prediction algorithm. The error is divided into an absolute error and a relative error, and the accuracy of load prediction is judged mainly through the relative error.
(1) Absolute error
In the formula: AEiIs the absolute error at time i, XiIs the actual value at time i and,
is the predicted value at time i.
(2) Relative error
In the formula: REiRelative error at time i, AEiIs the absolute error at time i, XiThe actual value at time i.
(3) Mean relative error (MAPE): the absolute value of the prediction error between the predicted and measured values is divided by the average of the measured value sums.
2. Simulation of experiment
In order to prove the effectiveness and the accuracy of the algorithm, 720 historical load data of No. 8/month 1 to No. 8/month 30 of 2020 in summer are selected, a load numerical value of a date with the similarity value higher than 0.6 is selected as a training sample by using a similarity method, and the load prediction is carried out on No. 8/month 31 of 2020. And (3) predicting the load No. 8/month & 31 in 2020 by using a data training sample as an input factor and using three models of an improved artificial bee colony-least square support vector machine (IABC-LS-SVM), a particle swarm-least square support vector machine (PSO-LS-SVM) and a least square support vector machine (LS-SVM), respectively, and comparing the predicted load with the actual load in the current day to calculate an error value.
The parameters of the improved artificial bee colony algorithm are set as follows: the bee colony size NP is 40, the number of the bee collecting bees and the number of the observation bees are equal to 20, the maximum number N of each search of the bee collecting bees and the observation bees is 50, and the total number of iterations of the algorithm is Max is 150. And a regularization parameter C and a kernel parameter epsilon in the least square support vector machine respectively represent parameters of a honey source with the optimal fitness in the improved artificial bee colony algorithm. The parameters of the least square support vector machine are selected by using a modified artificial bee colony algorithm, and the search range of C and epsilon is [0.01,100 ].
The load prediction results of the different algorithms are shown in fig. 4 below:
TABLE 3 comparison of errors in load prediction for different algorithms
In FIG. 4, there are four curves, which are the actual value curve, LS-SVM, PSO-LS-SVM and IABC-LS-SVM, respectively, and the values of the curves are close but have a certain error. Table 3 and fig. 5 may be more convenient to compare the predicted load values and the errors of different algorithms. It can be seen that the average relative error of the load prediction of the standard least square support vector machine is about 2.42%, the average relative error of the load prediction of the particle swarm improved least square support vector machine is about 1.56%, and the average relative error of the load prediction of the least square support vector machine of the improved artificial bee colony algorithm is about 0.85%. Compared with the least square support vector machine using the improved artificial bee colony algorithm, the prediction accuracy is higher, and the prediction error value is smaller.
The invention uses the LS-SVM method to predict the short-term load of the power distribution network, introduces an improved artificial bee colony algorithm to optimize the parameters of the least square support vector machine for the parameter selection problem of the LS-SVM, and overcomes the blindness of manual selection. The artificial bee colony algorithm with the improved differential evolution idea enables the artificial bee colony algorithm to have better searching performance and overcomes the defects of weak local searching capability and low convergence speed. And historical load data is selected by adopting a similar daily method, so that the accuracy of load prediction is higher. By example analysis, the load prediction method of the least square support vector machine for improving the artificial bee colony algorithm is proved to be capable of effectively improving the accuracy of load prediction. And the method is rarely used in the field of distribution network load prediction. Before the electric quantity scheduling is carried out on the power distribution network, the numerical value of the load is accurately predicted, so that the network loss is favorably reduced, the redundant electric quantity and the economic loss are reduced, and the safety of a circuit is enhanced.
While the invention has been described with reference to specific embodiments, the invention is not limited thereto, and various equivalent modifications and substitutions can be easily made by those skilled in the art within the technical scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A short-term load prediction method based on an improved least square support vector machine is characterized by comprising the following steps:
step 1: collecting historical load data, and selecting the historical load data by using a similar daily method to obtain a training sample;
step 2: establishing a prediction model of a least square support vector machine to carry out load prediction on a training sample, and when the accuracy of a prediction result is judged to be not up to standard, improving the prediction model of the least square support vector machine by utilizing an improved artificial bee colony algorithm to obtain the improved prediction model of the least square support vector machine;
and step 3: and inputting the training sample into a prediction model of the improved least square support vector machine for load prediction and error analysis, and outputting a prediction result.
2. The method for predicting the short-term load based on the improved least square support vector machine as claimed in claim 1, wherein the step 1 comprises the following substeps:
step 101: collecting historical load data comprising weather types, date types and date differences;
step 102: preprocessing abnormal or missing load data in the acquired data;
step 103: and selecting a proper evaluation function for evaluating the similarity between the historical days and the prediction days according to the preprocessed data, and selecting the data of the historical days with higher similarity as training samples.
3. The method of claim 2, wherein the evaluation function selected in step 103 for evaluating the similarity between the historical days and the predicted days is represented by the following mathematical formula:
in the formula, rijFor the degree of similarity between historical and predicted days, XikFor feature quantities quantized for historical day i, XjkThe characteristic vector of the day j to be measured, m is the characteristic vector, and k is a natural number.
4. The method for predicting short-term load based on the improved least square support vector machine as claimed in claim 1, wherein the step 2 comprises the following sub-steps:
step 201: establishing a prediction model of a least square support vector machine to carry out load prediction on a training sample, and judging whether the precision of a prediction result reaches the standard or not;
step 202: when the precision is not up to the standard, improving the artificial bee colony algorithm, and optimizing the parameters of the least square support vector machine by using the improved artificial bee colony algorithm;
step 203: and establishing a corresponding model based on the optimized parameters of the least square support vector machine.
5. The short-term load prediction method based on the improved least square support vector machine as claimed in claim 4, wherein the artificial bee colony algorithm improved in step 201 improves the random search equation of the bee sampling by using the mutation operation in the differential evolution algorithm, and the improved search equation corresponds to:
vij＝xbest，j+φij(xr1，j-xr2，j)
in the formula, vijAs a new source of honey, xbest，jThe j weft value phi of the optimal honey sourceijIs [ -1, 1 [ ]]Random number between, xr1，jAnd xr2，jTwo random individuals different from i.
6. The method of claim 4, wherein in step 201, the relative error is used to determine whether the accuracy of the prediction result meets the standard.
7. The method of claim 4, wherein the kernel function of the prediction model of the least squares support vector machine in step 201 is a radial basis function, and the mathematical description formula is as follows:
wherein ε is a nuclear parameter.
8. The method of claim 4, wherein the regression constraint of the prediction model of the least squares support vector machine in the step 201 is:
in the formula, | w |)2To control the complexity of the model, eiError of sample data prediction for the trained data model, C is regularization parameter, b is constant, yiIn order to output the data, the data is output,
for mapping from a low-dimensional space to a high-dimensional space, W is a weight vector.
9. The method of claim 4, wherein the artificial bee colony algorithm modified in step 201 is implemented by using bees to transmit the collected bee source information to observers, and the observers use a wheel selection method to predict the short-term load with a probability piSelecting food sources, searching for new solutions around the food sources, calculating fitness values, and greedy selecting, wherein the probability piThe mathematical description formula of (a) is:
in the formula, fitiTo solve XiThe fitness value of (1, 2., SN), SN is the number of solutions in the population.
10. The method of claim 4, wherein in the artificial bee colony algorithm improved in the step 201, if the optimal value of the honey bee is not found after a plurality of searches, the honey bee is changed into a scout bee, and a new solution is randomly generated to replace the old solution according to a corresponding formula, wherein the corresponding formula is as follows:
xij＝ximin+rand(0，1)(ximax-ximin)
in the formula, ximaxAnd ximinRepresenting the maximum and minimum around the old solution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110321179.0A CN112949207A (en) | 2021-03-25 | 2021-03-25 | Short-term load prediction method based on improved least square support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110321179.0A CN112949207A (en) | 2021-03-25 | 2021-03-25 | Short-term load prediction method based on improved least square support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112949207A true CN112949207A (en) | 2021-06-11 |
Family
ID=76226729
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110321179.0A Pending CN112949207A (en) | 2021-03-25 | 2021-03-25 | Short-term load prediction method based on improved least square support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112949207A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113890094A (en) * | 2021-10-11 | 2022-01-04 | 燕山大学 | Multi-target distributed power supply optimal configuration method considering power supply setting conditions |
| CN114180023A (en) * | 2021-12-08 | 2022-03-15 | 中国船舶重工集团公司第七一六研究所 | Multi-energy ship control management method and device based on load prediction algorithm |
| CN114397289A (en) * | 2022-01-14 | 2022-04-26 | 长春工业大学 | Quantitative analysis method for heavy metal elements in soil and related equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106022465A (en) * | 2016-05-19 | 2016-10-12 | 江南大学 | Extreme learning machine method for improving artificial bee colony optimization |
| CN106855957A (en) * | 2015-12-09 | 2017-06-16 | 四川大学 | Factory's bus load prediction based on similar day and least square method supporting vector machine |
| CN108416366A (en) * | 2018-02-06 | 2018-08-17 | 武汉大学 | A kind of power-system short-term load forecasting method of the weighting LS-SVM based on Meteorological Index |
| CN108829957A (en) * | 2018-06-01 | 2018-11-16 | 韩山师范学院 | A kind of welded grider design method based on mixing difference artificial bee colony algorithm |
| CN111931990A (en) * | 2020-07-13 | 2020-11-13 | 江西理工大学 | Power load prediction method based on optimized least square support vector machine |
-
2021
- 2021-03-25 CN CN202110321179.0A patent/CN112949207A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106855957A (en) * | 2015-12-09 | 2017-06-16 | 四川大学 | Factory's bus load prediction based on similar day and least square method supporting vector machine |
| CN106022465A (en) * | 2016-05-19 | 2016-10-12 | 江南大学 | Extreme learning machine method for improving artificial bee colony optimization |
| CN108416366A (en) * | 2018-02-06 | 2018-08-17 | 武汉大学 | A kind of power-system short-term load forecasting method of the weighting LS-SVM based on Meteorological Index |
| CN108829957A (en) * | 2018-06-01 | 2018-11-16 | 韩山师范学院 | A kind of welded grider design method based on mixing difference artificial bee colony algorithm |
| CN111931990A (en) * | 2020-07-13 | 2020-11-13 | 江西理工大学 | Power load prediction method based on optimized least square support vector machine |
Non-Patent Citations (3)
| Title |
|---|
| WEIFENG GAO ET AL.: "a global best artificial bee colony algorithm for global optimization", 《JOURNAL OF COMPUTATIONAL AND APPLIED MATHEMATICS》 * |
| 王雪微等: "基于相似日的支持向量机短期负荷预测", 《计算机测量与控制》 * |
| 陈阳: "基于人工蜂群优化LS-SVM短期负荷预测研究", 《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113890094A (en) * | 2021-10-11 | 2022-01-04 | 燕山大学 | Multi-target distributed power supply optimal configuration method considering power supply setting conditions |
| CN113890094B (en) * | 2021-10-11 | 2022-06-03 | 燕山大学 | Multi-target distributed power supply optimal configuration method considering power supply setting conditions |
| CN114180023A (en) * | 2021-12-08 | 2022-03-15 | 中国船舶重工集团公司第七一六研究所 | Multi-energy ship control management method and device based on load prediction algorithm |
| CN114180023B (en) * | 2021-12-08 | 2023-09-01 | 中国船舶集团有限公司第七一六研究所 | Multi-energy ship control management method and device based on load prediction algorithm |
| CN114397289A (en) * | 2022-01-14 | 2022-04-26 | 长春工业大学 | Quantitative analysis method for heavy metal elements in soil and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113962364B (en) | Multi-factor power load prediction method based on deep learning | |
| CN108491969B (en) | Big data-based space load prediction model construction method | |
| CN111783953B (en) | 24-point power load value 7-day prediction method based on optimized LSTM network | |
| CN112949207A (en) | Short-term load prediction method based on improved least square support vector machine | |
| CN108921339B (en) | Quantile regression-based photovoltaic power interval prediction method for genetic support vector machine | |
| CN112734128A (en) | 7-day power load peak value prediction method based on optimized RBF | |
| CN111160626B (en) | Power load time sequence control method based on decomposition fusion | |
| CN116316599A (en) | Intelligent electricity load prediction method | |
| CN112329990A (en) | User power load prediction method based on LSTM-BP neural network | |
| CN110796307A (en) | Distributed load prediction method and system for comprehensive energy system | |
| CN116227637A (en) | Active power distribution network oriented refined load prediction method and system | |
| CN114021848A (en) | Generating capacity demand prediction method based on LSTM deep learning | |
| CN115860177A (en) | Photovoltaic power generation power prediction method based on combined machine learning model and application thereof | |
| CN113011680A (en) | Power load prediction method and system | |
| CN115545333A (en) | Method for predicting load curve of multi-load daily-type power distribution network | |
| Alharbi et al. | Short-term wind speed and temperature forecasting model based on gated recurrent unit neural networks | |
| CN113887809A (en) | Power distribution network supply and demand balance method, system, medium and computing equipment under double-carbon target | |
| CN113033898A (en) | Electrical load prediction method and system based on K-means clustering and BI-LSTM neural network | |
| CN115905319B (en) | Automatic identification method and system for abnormal electricity fees of massive users | |
| CN108345996B (en) | System and method for reducing wind power assessment electric quantity | |
| CN113723670B (en) | Photovoltaic power generation power short-term prediction method with variable time window | |
| CN115759395A (en) | Training of photovoltaic detection model, detection method of photovoltaic power generation and related device | |
| CN114819337A (en) | Multi-task learning-based comprehensive energy system multi-load prediction method | |
| CN113537607A (en) | Power failure prediction method | |
| Shen et al. | An interval analysis scheme based on empirical error and mcmc to quantify uncertainty of wind speed |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210611 |