CN110264006B - Wind power probabilistic prediction method based on chaotic firefly algorithm and Bayesian network - Google Patents

Wind power probabilistic prediction method based on chaotic firefly algorithm and Bayesian network Download PDF

Info

Publication number
CN110264006B
CN110264006B CN201910541439.8A CN201910541439A CN110264006B CN 110264006 B CN110264006 B CN 110264006B CN 201910541439 A CN201910541439 A CN 201910541439A CN 110264006 B CN110264006 B CN 110264006B
Authority
CN
China
Prior art keywords
firefly
time point
wind power
value
sequence
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910541439.8A
Other languages
Chinese (zh)
Other versions
CN110264006A (en
Inventor
何耀耀
祝贺功
施诺
赵秋宇
李路遥
范慧玲
张婉莹
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hefei University of Technology
Original Assignee
Hefei University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hefei University of Technology filed Critical Hefei University of Technology
Priority to CN201910541439.8A priority Critical patent/CN110264006B/en
Publication of CN110264006A publication Critical patent/CN110264006A/en
Application granted granted Critical
Publication of CN110264006B publication Critical patent/CN110264006B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G06F18/24155—Bayesian classification
    • 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Tourism & Hospitality (AREA)
  • Artificial Intelligence (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • General Health & Medical Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Computing Systems (AREA)
  • Primary Health Care (AREA)
  • Public Health (AREA)
  • Operations Research (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Game Theory and Decision Science (AREA)
  • Molecular Biology (AREA)
  • Water Supply & Treatment (AREA)
  • Development Economics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a wind power probabilistic prediction method based on a chaotic firefly algorithm and a Bayesian network, which comprises the following steps: 1, acquiring wind speed, wind direction, air temperature and wind power actual power data, preprocessing the data, and selecting training set and test set data; 2, performing empirical mode decomposition on the wind power original data to make the wind power time sequence more stable; 3, constructing a Bayesian network model to obtain an initial prediction interval; 4, calculating the range of interval change amplitude, and obtaining the optimal interval change amplitude by using a chaotic firefly algorithm; and 5, performing chaotic search near the optimal interval change amplitude to obtain a final prediction interval. The method can measure the uncertainty of the wind power through the construction prediction region, thereby providing effective reference for power scheduling decision.

Description

Wind power probabilistic prediction method based on chaotic firefly algorithm and Bayesian network
Technical Field
The invention relates to the technical field of wind power generation, in particular to a wind power prediction model method based on EMD decomposition, a Bayesian network and a chaotic firefly algorithm.
Background
In recent years, with the development of socio-economic, the demand for energy is increasing, fossil energy is gradually facing the risk of exhaustion, and the use of a large amount of fossil fuel brings about a serious pollution problem. In order to cope with the problems of fuel energy and environmental pollution, renewable energy is widely used. Among them, wind energy has been rapidly developed and applied over the past decades as a clean, renewable, most potentially exploited source of energy. With the gradual expansion of the scale of wind power generation, wind power occupies a greater and greater proportion in a power system. The weak controllability of wind power caused by the intermittency and uncertainty of wind power brings great challenges to the safe and stable operation of the power system. In order to reduce the risk of wind power network access, reduce the construction cost of system reserve capacity and help the power department to make an accurate scheduling scheme, reliable, timely and accurate wind power prediction becomes a problem which needs to be solved urgently at present.
Wind power generation is affected by factors such as weather factors, unit faults and data misloading, so that the collected data often has abnormal values and missing values, and certain trouble is brought to wind power prediction. Meanwhile, the wind power data has large volatility, and if the data is not subjected to certain technical processing, the prediction result obtained by directly predicting the data is often poor. However, the current research generally aims at processing abnormal values and missing values existing in a data set, no targeted countermeasures are provided for strong fluctuation of wind power data, and the accuracy of a prediction result often does not reach an ideal level.
At present, the methods for wind power prediction at home and abroad mainly include a physical method based on numerical weather forecast, a statistical method based on time series historical data and a prediction method of the mixture of the two methods. Generally speaking, a prediction mode which can simultaneously consider numerical weather physical data and time series historical data for wind power prediction can remarkably improve the prediction accuracy, but how to effectively combine data of two components and predict the data is a problem which is difficult to solve in the past. Researchers have achieved excellent research results in the field of point prediction of wind power prediction for decades, and prediction accuracy has already approached an ideal level. However, due to the uncertainty of wind energy and the inherent defects of the model, the traditional deterministic point prediction method has the problems that point prediction errors cannot be eliminated, the uncertainty of the result cannot be measured, the fluctuation range of the wind power cannot be given, and the like.
In the current method for carrying out probabilistic prediction on wind power, most models are optimized by using an evolutionary algorithm, but the traditional evolutionary algorithm has strong global searching capability and can effectively improve the convergence speed of the models, but the traditional evolutionary algorithm generally has the defects of low later-stage convergence speed, early maturity of the models and easy falling into local optimal solution. The operation time of the algorithm is long in wind power prediction, and the time availability requirement cannot be met; the prediction precision is not high, and the requirement of accuracy usability cannot be met.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides a wind power probabilistic prediction method based on a chaotic firefly algorithm and a Bayesian network, and can provide effective reference for power scheduling decision by measuring the uncertainty of the wind power through a construction interval.
The invention adopts the following technical scheme for solving the technical problems:
the invention relates to a wind power probabilistic prediction method based on a chaotic firefly algorithm and a Bayesian network, which is characterized by comprising the following steps of:
step 1, acquiring wind speed, wind direction, air temperature and wind power actual power data and carrying out data preprocessing:
step 1.1, collecting historical data of wind speed to form an original wind speed sequence, and filling missing values and abnormal values and normalizing the original wind speed sequence to obtain a preprocessed wind speed sequence which is recorded asV＝[v1,v2,...vi,...,vN]，viI is more than or equal to 1 and less than or equal to N, and N is the total number of samples, wherein the wind speed data of the ith time point in the preprocessed wind speed sequence V is the wind speed data of the ith time point;
acquiring historical data of wind directions to form an original wind direction sequence, and performing missing value and abnormal value filling and normalization processing on the original wind direction sequence to obtain a preprocessed wind direction sequence which is recorded as F ═ F1,f2,...fi...fN],fiWind direction data of the ith time point in the preprocessed wind direction sequence F are obtained;
collecting historical data of air temperature to form an original air temperature sequence, and performing missing value and abnormal value filling and normalization processing on the original air temperature sequence to obtain a preprocessed air temperature sequence which is marked as T ═ w1,w2,...wi,...wN]，wiThe temperature data of the ith time point in the preprocessed temperature sequence T is obtained;
acquiring actual power historical data of wind power to form an original wind power sequence, and filling missing values and abnormal values and normalizing the original wind power sequence to obtain a preprocessed wind power sequence which is recorded as P ═ P1,p2,...,pi,...,pN]，piWind power data of the ith time point in the wind power sequence P after the pretreatment is obtained;
dividing a data set consisting of the preprocessed wind speed sequence, wind direction sequence, air temperature sequence and wind power sequence into training set data and test set data;
step 1.2, carrying out empirical mode decomposition on the preprocessed wind power sequence P to obtain a data set consisting of k IMF components and a margin; the k IMF components are denoted as
Figure GDA0002764153530000021
Figure GDA0002764153530000022
Is the jth IMF scoreAmount of, and
Figure GDA0002764153530000023
Figure GDA0002764153530000024
is the jth IMF component
Figure GDA0002764153530000025
The wind power decomposition value of the ith time point;
step 2, constructing a Bayesian network model by using training set data:
step 2.1, the jth IMF component
Figure GDA0002764153530000026
Wind power decomposition value of the ith time point
Figure GDA0002764153530000027
And the wind speed data v of the ith time pointiWind direction data fiAnd gas temperature data wiAs an influence factor in the bayesian network model, the jth IMF component is added
Figure GDA0002764153530000031
Wind power decomposition value of the (i + 1) th time point
Figure GDA0002764153530000032
As an output node, thereby constructing a Bayesian network model;
step 2.2, respectively calculating the conditional probability of the output node corresponding to each influence factor when the influence factors take different values according to historical data, thereby obtaining a conditional probability table;
step 2.3, obtaining the jth IMF component according to the conditional probability table and the Bayesian network model
Figure GDA0002764153530000033
Power resolution value of the (i + 1) th time point
Figure GDA0002764153530000034
And taking a power interval corresponding to the highest point on the probability distribution curve as the jth IMF component
Figure GDA0002764153530000035
Wind power decomposition value of the (i + 1) th time point
Figure GDA0002764153530000036
The initial prediction interval of (1);
step 2.4, repeating the steps 2.1 to 2.3, so as to obtain an initial prediction interval of the power decomposition value of the (i + 1) th time point in the k IMF components, taking the sum of the initial prediction intervals of the power decomposition value of the (i + 1) th time point in the k IMF components as the wind power prediction interval of the (i + 1) th time point, and recording the sum as the wind power prediction interval of the (i + 1) th time point
Figure GDA0002764153530000037
Step 3, obtaining an optimal interval amplitude variation range by using a chaotic firefly algorithm:
step 3.1, predicting the interval according to the wind power of the (i + 1) th time point
Figure GDA0002764153530000038
Determining the median of the prediction interval of the (i + 1) th time point
Figure GDA0002764153530000039
Wherein,
Figure GDA00027641535300000310
the predicted value is used as the wind power predicted value of the (i + 1) th time point;
predicting the wind power of the (i + 1) th time point
Figure GDA00027641535300000311
Lower limit of (2)
Figure GDA00027641535300000312
Wind power predicted value divided by (i + 1) th time point
Figure GDA00027641535300000313
The lower ratio of the (i + 1) th time point is recorded as
Figure GDA00027641535300000314
Thereby obtaining the lower limit ratios of the N time points, and selecting the maximum value and the minimum value in the lower limit ratios of the N time points as the lower limit ratio betalowThe variation range of (a);
predicting the wind power of the (i + 1) th time point
Figure GDA00027641535300000315
Upper limit of (2)
Figure GDA00027641535300000316
Wind power predicted value divided by (i + 1) th time point
Figure GDA00027641535300000317
The upper ratio of the time points at i +1 is recorded as
Figure GDA00027641535300000318
Thereby obtaining the upper limit ratios of the N time points, and selecting the maximum value and the minimum value in the upper limit ratios of the N time points as the upper limit ratio betahighThe variation range of (a);
lower limit ratio betalowAnd an upper limit ratio betahighThe method is used as a parameter to be optimized in the chaotic firefly algorithm, and two corresponding variation ranges are used as the variation ranges of the population;
step 3.2, initializing the population:
setting a compression operator as g, a population scale as M, a current evolutionary algebra as T, and a maximum evolutionary algebra as TmaxThe current chaotic search frequency is D, and the maximum chaotic search frequency is DmaxThe absorption factor of light intensity is gamma, gamma belongs to [0,1 ]]Maximum attraction ω0Step size factor eta, eta is equal to [0,1 ]]；
When the initialization t is 0, M random numbers are generated in the variation range of the population and are used as M firefsInitial position of firefly, wherein the initial position of the nth firefly at the t-th evolutionary iteration is recorded as
Figure GDA0002764153530000041
The initial position of the mth firefly is
Figure GDA0002764153530000042
N is more than or equal to 1 and less than or equal to M, M is more than or equal to 1 and less than or equal to M, wherein,
Figure GDA0002764153530000043
the nth firefly is in a lower limit ratio beta in the t evolution iterationlowThe random number in the variation range of (a),
Figure GDA0002764153530000044
the nth firefly at the t evolution iteration is the upper limit ratio betahighThe random number in the variation range of (a),
Figure GDA0002764153530000045
the mth firefly is in the lower limit ratio beta in the t evolution iterationlowThe random number in the variation range of (a),
Figure GDA0002764153530000046
the mth firefly at the tth evolution iteration is the upper limit ratio betahighRandom number in the variation range of (2), the initial position of the firefly population is recorded as
Figure GDA0002764153530000047
Step 3.3, calculating the fluorescence brightness of the firefly:
the luminance of the nth firefly in the t evolution iteration is constructed by using the formula (1)
Figure GDA0002764153530000048
And the luminance of the mth firefly in the tth evolutionary iteration is obtained according to the formula (1)
Figure GDA0002764153530000049
Figure GDA00027641535300000410
In the formula (1), u is a confidence level,
Figure GDA00027641535300000411
the average bandwidth of the prediction interval of the nth firefly in the t evolution iteration is obtained by the formula (2),
Figure GDA00027641535300000412
the predicted interval coverage rate of the nth firefly in the t evolution iteration is obtained by the formula (3),
Figure GDA00027641535300000413
to predict the interval coverage
Figure GDA00027641535300000414
Is obtained by the formula (4), and lambda is the predicted interval coverage
Figure GDA00027641535300000415
Penalty factor when confidence level mu is not reached;
Figure GDA00027641535300000416
Figure GDA00027641535300000417
Figure GDA00027641535300000418
in the formula (2), R is a wind power true value sequence P ═ P1,p2,...,pi,...,pN]The difference between the maximum value and the minimum value in (b),
Figure GDA0002764153530000051
the lower bound of the prediction interval of the (i + 1) th time point of the nth firefly in the t evolution iteration is obtained by the formula (5),
Figure GDA0002764153530000052
the upper bound of the prediction interval of the (i + 1) th time point of the nth firefly in the t evolution iteration is obtained by the formula (6);
in the formula (3), bn,i+1Is the Boolean constant of the (i + 1) th time point of the nth firefly, if the real value p of the wind power of the (i + 1) th time pointi+1The prediction interval range of the (i + 1) th time point of the nth firefly in the t evolution iteration
Figure GDA0002764153530000053
In the interior, then order b n,i+11 is ═ 1; otherwise, let bn,i+1＝0；
Figure GDA0002764153530000054
Figure GDA0002764153530000055
In the formulae (5) and (6),
Figure GDA0002764153530000056
predicting the median of the interval for the (i + 1) th time point of the nth firefly;
according to the initial position and the fitness function of the fireflies, the fitness value of each firefly is used as the initial fluorescence brightness of the firefly
Figure GDA0002764153530000057
The fluorescence brightness of the t-th evolution iteration
Figure GDA0002764153530000058
Minimum value ofIs marked as
Figure GDA0002764153530000059
Wherein,
Figure GDA00027641535300000510
for the t-th iteration of evolution, the firefly population
Figure GDA00027641535300000511
Is located at the optimum position in the (c),
Figure GDA00027641535300000512
to an optimum position
Figure GDA00027641535300000513
Is determined by the first component of (a),
Figure GDA00027641535300000514
to an optimum position
Figure GDA00027641535300000515
A second component of (a); let Igmin(Xgbest) Is the global optimum fitness value after t evolutionary iterations, and
Figure GDA00027641535300000516
make global optimum position
Figure GDA00027641535300000517
Step 3.4, the firefly position moves:
according to the light intensity absorption factor gamma and the maximum attraction omega0Obtaining the mutual attraction degree omega of the mth firefly and the nth firefly; if it is
Figure GDA00027641535300000518
The mth firefly moves toward the nth firefly according to the mutual attraction degree omega; and obtaining the updated position of the mth firefly and using the updated position as the position of the mth firefly in the (t + 1) th evolution iteration and recording the position as the updated position
Figure GDA00027641535300000519
Otherwise, the nth firefly moves towards the mth firefly according to the mutual attraction omega; and obtaining the updated position of the nth firefly and using the updated position as the position of the nth firefly in the (t + 1) th evolution iteration, and recording the position as the position of the nth firefly
Figure GDA00027641535300000520
Thereby obtaining the updated positions of all the fireflies and forming a firefly population during the (t + 1) th evolutionary iteration;
step 3.5, obtaining the brightness of each individual in the firefly population during the t +1 th evolution iteration by using the formula (1), and comparing the brightness among the firefly individuals, so as to find out the fitness value of the firefly with the minimum brightness in the firefly population during the t +1 th evolution iteration, and recording the fitness value as the fitness value
Figure GDA0002764153530000061
Wherein,
Figure GDA0002764153530000062
the optimal position in the firefly population after the t +1 th evolutionary iteration;
if the global optimum fitness value
Figure GDA0002764153530000063
Then will be
Figure GDA0002764153530000064
Is assigned to XgbestThereby updating the global optimum position Xgbest(ii) a Otherwise, the global optimal position XgbestKeeping the same;
step 3.6, the optimal position in the firefly population during the t +1 th evolution iteration
Figure GDA0002764153530000065
Local search is performed nearby:
initialization d ═ 0, at [ -1,1 [ ]]Randomly generating q-dimension chaotic variable in d-th chaotic search in interval
Figure GDA0002764153530000066
q represents a dimension, and q is 1, 2;
the optimal position of the t +1 th evolution iteration
Figure GDA0002764153530000067
Assigning to the optimal coordinate y in the d-th chaotic search(d)And q is 1,
Figure GDA0002764153530000068
when q is equal to 2, the reaction is carried out,
Figure GDA0002764153530000069
Figure GDA00027641535300000610
for the optimal coordinate y in the d-th chaotic search(d)The q-th component of (a);
step 3.7, chaotic search is carried out by using the logistic mapping shown in the formula (7) to obtain chaotic variables during the d +1 th chaotic search
Figure GDA00027641535300000611
Figure GDA00027641535300000612
In the formula (7), s is a constant and is more than 0 and less than or equal to 2;
step 3.8, obtaining the optimal coordinate y in the d +1 th chaotic search by using the formula (8)(d+1)Q component of
Figure GDA00027641535300000613
Figure GDA00027641535300000614
In formula (8), when q is 1, xq,maxAs the first component of the initial position of the firefly population
Figure GDA00027641535300000615
Maximum value of (1), xq,minAs the first component of the initial position of the firefly population
Figure GDA00027641535300000616
Minimum value of (1); when q is 2, xq,maxSecond component of initial position of firefly population
Figure GDA00027641535300000617
Maximum value of (1), xq,minSecond component of initial position of firefly population
Figure GDA00027641535300000618
Minimum value of (1); g(t)For the compression operator after the t-th evolutionary iteration, the method is obtained by using the formula (9):
Figure GDA00027641535300000619
in formula (9), L is a constant greater than 0;
step 3.9, if D is less than DmaxIf so, assigning d +1 to d, and returning to the step 3.7; otherwise, obtaining the optimal coordinate sequence through chaotic search
Figure GDA0002764153530000071
And proceeds to step 3.9.1;
step 3.9.1, calculating the fitness value of the optimal coordinate sequence using equation (1)
Figure GDA0002764153530000072
Finding the minimum fitness value in the fitness values of the optimal coordinate sequence and recording the minimum fitness value as Imin(ybest) Wherein, ybestFor the optimal coordinate sequence
Figure GDA0002764153530000073
The optimal position of (1);
if it is not
Figure GDA0002764153530000074
Then pass through
Figure GDA0002764153530000075
Updating
Figure GDA0002764153530000076
Otherwise, it orders
Figure GDA0002764153530000077
Keeping the same;
step 3.9.2 if the global optimum fitness value
Figure GDA0002764153530000078
Then pass through
Figure GDA0002764153530000079
Updating the global optimal position; otherwise, let the global optimum position XgbestKeeping the same;
step 3.10, if T is less than TmaxAssigning t +1 to t and returning to the step 3.3; otherwise, the global optimal position X in the firefly population is obtainedgbestObtaining the optimal solution of the fitness function
Figure GDA00027641535300000710
And
Figure GDA00027641535300000711
step 4, according to the optimal solution
Figure GDA00027641535300000712
And
Figure GDA00027641535300000713
respectively calculating the optimal lower bound of the prediction interval of the (i + 1) th time point
Figure GDA00027641535300000714
The (i + 1) th time pointIs optimally upper bound of the prediction interval
Figure GDA00027641535300000715
And by an optimal lower bound of the prediction interval
Figure GDA00027641535300000716
And an optimal upper bound
Figure GDA00027641535300000717
And forming a final prediction interval, and substituting the test set data into the Bayesian network model so as to complete the probabilistic prediction of the wind power.
Compared with the prior art, the invention has the beneficial effects that:
1. the empirical mode decomposition adopted by the invention is a self-adaptive data processing and mining method, and can be used for carrying out stabilization processing on data, so that the wind power is more stable, and the fluctuation of the wind power is reduced.
2. The method utilizes the Bayesian network prediction model to carry out probabilistic interval prediction to overcome the defects of low reliability, complex calculation and the like of the existing prediction method, and improves the prediction quality of the wind power.
3. According to the method, the cleaned wind power data is combined with the chaotic firefly algorithm and the Bayesian network method, a prediction model is constructed, the optimal interval amplitude variation range is obtained, and a certain reference is provided for resource scheduling personnel.
4. Aiming at the problem of wind power probabilistic prediction, the invention provides a novel chaotic firefly algorithm, which carries out chaotic search near the optimal solution solved by the firefly algorithm, improves the global optimization capability of the chaotic firefly algorithm, and solves the problem that the chaotic search cannot be carried out in a negative value interval by carrying out the chaotic search by using logistic mapping, thereby improving the prediction quality of the probabilistic interval and facilitating a decision maker to carry out scientific and reasonable decision.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
fig. 2 is a block diagram of a bayesian network of the present invention.
Detailed Description
In this embodiment, a wind power probabilistic prediction method based on a chaotic firefly algorithm and a bayesian network, as shown in fig. 1, includes: acquiring wind speed, wind direction, air temperature and wind power actual power data, and preprocessing the data; EMD decomposition is carried out on the actual wind power, and the fluctuation of the wind power is reduced; establishing a Bayesian network model to obtain an initial prediction interval; and calculating the range of the interval change amplitude, obtaining the optimal interval change range when the fitness function is optimal by using the chaotic firefly algorithm, thereby obtaining a final prediction interval, and analyzing and evaluating a prediction result. Specifically, the method comprises the following steps:
step 1, acquiring wind speed, wind direction, air temperature and wind power actual power data and carrying out data preprocessing:
step 1.1, collecting historical data of wind speed to form an original wind speed sequence, and performing missing value and abnormal value filling and normalization processing on the original wind speed sequence to obtain a preprocessed wind speed sequence which is recorded as V ═ V1,v2,...vi,...,vN]，viThe wind speed data of the ith time point in the wind speed sequence V after pretreatment, i is more than or equal to 1 and less than or equal to N, and N is the total number of samples;
acquiring historical data of wind directions to form an original wind direction sequence, and performing missing value and abnormal value filling and normalization processing on the original wind direction sequence to obtain a preprocessed wind direction sequence, wherein the preprocessed wind direction sequence is marked as F ═ F1,f2,...fi...fN],fiWind direction data of the ith time point in the preprocessed wind direction sequence F;
collecting historical data of the air temperature to form an original air temperature sequence, and performing missing value and abnormal value filling and normalization processing on the original air temperature sequence to obtain a preprocessed air temperature sequence which is marked as T ═ w1,w2,...wi,...wN]，wiThe temperature data of the ith time point in the preprocessed temperature sequence T are obtained;
gather reality of wind-powered electricity generationThe power historical data form an original wind power sequence, and the original wind power sequence is subjected to missing value and abnormal value filling and normalization processing, so that a preprocessed wind power sequence is obtained and is marked as P ═ P1,p2,...,pi,...,pN]，piWind power data of the ith time point in the wind power sequence P after pretreatment;
dividing a data set consisting of the preprocessed wind speed sequence, wind direction sequence, air temperature sequence and wind power sequence into training set data and test set data;
step 1.2, carrying out empirical mode decomposition on the preprocessed wind power sequence P to obtain a data set consisting of k IMF components and a margin; wherein k IMF components are denoted as
Figure GDA0002764153530000081
Figure GDA0002764153530000082
Is the jth IMF component, and
Figure GDA0002764153530000083
Figure GDA0002764153530000084
is the jth IMF component
Figure GDA0002764153530000085
The wind power decomposition value of the ith time point;
step 2, constructing a Bayesian network model by using training set data:
step 2.1, the jth IMF component
Figure GDA0002764153530000091
Wind power decomposition value of the ith time point
Figure GDA0002764153530000092
And the wind speed data v of the ith time pointiWind direction data fiAnd gas temperature data wiAs an influence factor in the bayesian network model, the jth IMF component is added
Figure GDA0002764153530000093
Wind power decomposition value of the (i + 1) th time point
Figure GDA0002764153530000094
As output nodes, thereby constructing a Bayesian network model, as shown in FIG. 2;
step 2.2, respectively calculating the conditional probability of the output node corresponding to each influence factor when the influence factors take different values according to historical data, thereby obtaining a conditional probability table;
step 2.3, obtaining the jth IMF component according to the conditional probability table and the Bayesian network model
Figure GDA0002764153530000095
Power resolution value of the (i + 1) th time point
Figure GDA0002764153530000096
And taking a power interval corresponding to the highest point on the probability distribution curve as the jth IMF component
Figure GDA0002764153530000097
Wind power decomposition value of the (i + 1) th time point
Figure GDA0002764153530000098
The initial prediction interval of (1);
step 2.4, repeating the steps 2.1 to 2.3, so as to obtain an initial prediction interval of the power decomposition value of the (i + 1) th time point in the k IMF components, taking the sum of the initial prediction intervals of the power decomposition value of the (i + 1) th time point in the k IMF components as the wind power prediction interval of the (i + 1) th time point, and recording the sum as the wind power prediction interval of the (i + 1) th time point
Figure GDA0002764153530000099
Step 3, obtaining an optimal interval amplitude variation range by using a chaotic firefly algorithm:
step 3.1, predicting the interval according to the wind power of the (i + 1) th time point
Figure GDA00027641535300000910
Determining the median of the prediction interval of the (i + 1) th time point
Figure GDA00027641535300000911
Wherein,
Figure GDA00027641535300000912
the predicted value is used as the wind power predicted value of the (i + 1) th time point;
predicting the wind power of the (i + 1) th time point
Figure GDA00027641535300000913
Lower limit of (2)
Figure GDA00027641535300000914
Wind power predicted value divided by (i + 1) th time point
Figure GDA00027641535300000915
The lower ratio of the (i + 1) th time point is recorded as
Figure GDA00027641535300000916
Thereby obtaining the lower limit ratios of the N time points, and selecting the maximum value and the minimum value in the lower limit ratios of the N time points as the lower limit ratio betalowThe variation range of (A) is [0.4,1.2 ]]；
Predicting the wind power of the (i + 1) th time point
Figure GDA00027641535300000917
Upper limit of (2)
Figure GDA00027641535300000918
Wind power predicted value divided by (i + 1) th time point
Figure GDA00027641535300000919
As a firstThe upper ratio of the i +1 time points is recorded as
Figure GDA00027641535300000920
Thereby obtaining the upper limit ratios of the N time points, and selecting the maximum value and the minimum value in the upper limit ratios of the N time points as the upper limit ratio betahighThe variation range of (A) is [0.8,1.8 ]]；
Lower limit ratio betalowAnd an upper limit ratio betahighThe method is used as a parameter to be optimized in the chaotic firefly algorithm, and two corresponding variation ranges are used as the variation ranges of the population;
step 3.2, initializing the population:
setting a compression operator as g, a population scale as M, a current evolutionary algebra as T, and a maximum evolutionary algebra as TmaxThe current chaotic search frequency is D, and the maximum chaotic search frequency is DmaxThe absorption factor of light intensity is gamma, gamma belongs to [0,1 ]]Maximum attraction ω0Step size factor eta, eta is equal to [0,1 ]]；
When the initialization t is 0, generating M random numbers in the variation range of the population and taking the random numbers as the initial positions of M fireflies, wherein the initial position of the nth fireflies in the t evolution iteration
Figure GDA0002764153530000101
The initial position of the mth firefly is
Figure GDA0002764153530000102
N is more than or equal to 1 and less than or equal to M, M is more than or equal to 1 and less than or equal to M, wherein,
Figure GDA0002764153530000103
the nth firefly is in a lower limit ratio beta in the t evolution iterationlowThe random number in the variation range of (a),
Figure GDA0002764153530000104
the nth firefly at the t evolution iteration is the upper limit ratio betahighThe random number in the variation range of (a),
Figure GDA0002764153530000105
the mth firefly is in the lower limit ratio beta in the t evolution iterationlowThe random number in the variation range of (a),
Figure GDA0002764153530000106
the mth firefly at the tth evolution iteration is the upper limit ratio betahighRandom number in the variation range of (2), the initial position of the firefly population is recorded as
Figure GDA0002764153530000107
Step 3.3, calculating the fluorescence brightness of the firefly:
the luminance of the nth firefly in the t evolution iteration is constructed by using the formula (1)
Figure GDA0002764153530000108
And the luminance of the mth firefly in the tth evolutionary iteration is obtained according to the formula (1)
Figure GDA0002764153530000109
Figure GDA00027641535300001010
In the formula (1), u is a confidence level,
Figure GDA00027641535300001011
the average bandwidth of the prediction interval of the nth firefly in the t evolution iteration is obtained by the formula (2),
Figure GDA00027641535300001012
the predicted interval coverage rate of the nth firefly in the t evolution iteration is obtained by the formula (3),
Figure GDA00027641535300001013
to predict the interval coverage
Figure GDA00027641535300001014
Is obtained by the formula (4), and lambda is the predicted interval coverage
Figure GDA00027641535300001015
Penalty factor when confidence level mu is not reached;
Figure GDA00027641535300001016
Figure GDA00027641535300001017
Figure GDA0002764153530000111
in the formula (2), R is a wind power true value sequence P ═ P1,p2,...,pi,...,pN]The difference between the maximum value and the minimum value in (b),
Figure GDA0002764153530000112
the lower bound of the prediction interval of the (i + 1) th time point of the nth firefly in the t evolution iteration is obtained by the formula (5),
Figure GDA0002764153530000113
the upper bound of the prediction interval of the (i + 1) th time point of the nth firefly in the t evolution iteration is obtained by the formula (6);
in the formula (3), bn,i+1Is the Boolean constant of the (i + 1) th time point of the nth firefly, if the real value p of the wind power of the (i + 1) th time pointi+1The prediction interval range of the (i + 1) th time point of the nth firefly in the t evolution iteration
Figure GDA0002764153530000114
In the interior, then order b n,i+11 is ═ 1; otherwise, let bn,i+1＝0；
Figure GDA0002764153530000115
Figure GDA0002764153530000116
In the formulae (5) and (6),
Figure GDA0002764153530000117
predicting the median of the interval for the (i + 1) th time point of the nth firefly;
according to the initial position and the fitness function of the fireflies, the fitness value of each firefly is used as the initial fluorescence brightness of the firefly
Figure GDA0002764153530000118
The fluorescence brightness of the t-th evolution iteration
Figure GDA0002764153530000119
Minimum value of (1) is noted
Figure GDA00027641535300001110
Wherein,
Figure GDA00027641535300001111
for the t-th iteration of evolution, the firefly population
Figure GDA00027641535300001112
Is located at the optimum position in the (c),
Figure GDA00027641535300001113
to an optimum position
Figure GDA00027641535300001114
Is determined by the first component of (a),
Figure GDA00027641535300001115
to an optimum position
Figure GDA00027641535300001116
A second component of (a); let Igmin(Xgbest) Is the global optimum fitness value after t evolutionary iterations, and
Figure GDA00027641535300001117
make global optimum position
Figure GDA00027641535300001118
Step 3.4, the firefly position moves:
according to the light intensity absorption factor gamma and the maximum attraction omega0Obtaining the mutual attraction degree omega of the mth firefly and the nth firefly according to the formula (7);
Figure GDA00027641535300001119
in the formula (7), the reaction mixture is,
Figure GDA00027641535300001120
is the distance between firefly m and firefly n in the population at the t iteration, and has:
Figure GDA0002764153530000121
if it is
Figure GDA0002764153530000122
The mth firefly moves toward the nth firefly according to the formula (9);
Figure GDA0002764153530000123
in the formula (9), the reaction mixture is,
Figure GDA0002764153530000124
the position of the mth firefly in the (t + 1) th evolution iteration;
if it is
Figure GDA0002764153530000125
The nth firefly moves toward the mth firefly according to the formula (10);
Figure GDA0002764153530000126
in the formula (10), the compound represented by the formula (10),
Figure GDA0002764153530000127
the position of the nth firefly in the t +1 evolution iteration;
step 3.5, obtaining the brightness of each individual in the updated firefly population by using the formula (1), and comparing the brightness of the firefly individuals, so as to search the fitness value of the firefly with the minimum brightness in the t +1 th evolution iteration in the updated population
Figure GDA0002764153530000128
Figure GDA0002764153530000129
The optimal position in the firefly population after the t +1 th evolutionary iteration;
if the global optimum fitness value
Figure GDA00027641535300001210
Then will be
Figure GDA00027641535300001211
Is assigned to XgbestThereby updating the global optimum position Xgbest(ii) a Otherwise, the global optimal position XgbestKeeping the same;
step 3.6, the optimal position in the firefly population during the t +1 th evolution iteration
Figure GDA00027641535300001212
Local search is performed nearby:
initialization d ═ 0, at [ -1,1 [ ]]Randomly generating within the intervalQ-dimension chaotic variable in d chaotic searches
Figure GDA00027641535300001213
q represents a dimension, and q is 1, 2;
the optimal position obtained by the firefly algorithm in the t +1 th evolution iteration
Figure GDA00027641535300001214
Assigning to the optimal coordinate y in the d-th chaotic search(d)And q is 1,
Figure GDA00027641535300001215
when q is equal to 2, the reaction is carried out,
Figure GDA00027641535300001216
Figure GDA00027641535300001217
for the optimal coordinate y in the d-th chaotic search(d)The q-th component of (a);
step 3.7, chaotic search is carried out by using the logistic mapping shown in the formula (11) to obtain chaotic variables during the d +1 th chaotic search
Figure GDA00027641535300001218
Figure GDA00027641535300001219
In the formula (11), s is a constant and is more than 0 and less than or equal to 2;
step 3.8, obtaining the optimal coordinate y in the d +1 th chaotic search by using the formula (12)(d+1)Q component of
Figure GDA00027641535300001220
Figure GDA0002764153530000131
In the formula (12), the reaction mixture is,when q is 1, xq,maxAs the first component of the initial position of the firefly population
Figure GDA0002764153530000132
Maximum value of (1), xq,minAs the first component of the initial position of the firefly population
Figure GDA0002764153530000133
Minimum value of (1); when q is 2, xq,maxSecond component of initial position of firefly population
Figure GDA0002764153530000134
Maximum value of (1), xq,minSecond component of initial position of firefly population
Figure GDA0002764153530000135
Minimum value of (1); g(t)For the compression operator after the t-th evolutionary iteration, the method is obtained by using the formula (13):
Figure GDA0002764153530000136
in formula (13), L is a constant greater than 0;
step 3.9, if D is less than DmaxIf so, assigning d +1 to d, and returning to the step 3.7; otherwise, obtaining the optimal coordinate sequence through chaotic search
Figure GDA0002764153530000137
And proceeds to step 3.9.1;
step 3.9.1, calculating the fitness value of the optimal coordinate sequence using equation (1)
Figure GDA0002764153530000138
Finding the minimum fitness value in the fitness values of the optimal coordinate sequence is recorded as Imin(ybest) Wherein, ybestFor the optimal coordinate sequence
Figure GDA0002764153530000139
The optimal position of (1);
if it is not
Figure GDA00027641535300001310
Then pass through
Figure GDA00027641535300001311
Updating
Figure GDA00027641535300001312
Otherwise, it orders
Figure GDA00027641535300001313
Keeping the same;
step 3.9.2 if the global optimum fitness value
Figure GDA00027641535300001314
Then pass through
Figure GDA00027641535300001315
Updating the global optimal position; otherwise, let the global optimum position XgbestKeeping the same;
step 3.10, if T is less than TmaxAssigning t +1 to t and returning to the step 3.3; otherwise, the global optimal position X in the firefly population is obtainedgbestObtaining the optimal solution of the fitness function
Figure GDA00027641535300001316
And
Figure GDA00027641535300001317
step 4, substituting the test set data into the Bayesian network model to obtain the wind power prediction interval of the (i + 1) th time point
Figure GDA00027641535300001318
Optimal solution obtained according to step 3.10
Figure GDA00027641535300001319
And
Figure GDA00027641535300001320
respectively calculating the optimal lower bound of the prediction interval of the (i + 1) th time point
Figure GDA00027641535300001321
Optimal upper bound of prediction interval of i +1 th time point
Figure GDA00027641535300001322
And by an optimal lower bound of the prediction interval
Figure GDA00027641535300001323
And an optimal upper bound
Figure GDA00027641535300001324
And forming a final prediction interval, and substituting the test set data into the Bayesian network model so as to complete the probabilistic prediction of the wind power.
According to the invention, the EMD decomposition, the chaotic firefly algorithm and the Bayesian network are combined, the EMD decomposition enables the wind power to be more stable, the establishment of a model is facilitated, environmental factors influencing wind power generation can be considered through the Bayesian network model, the wind power prediction is closer to the reality, the chaotic firefly algorithm is used, the prediction interval is more accurate, and the model can provide a certain reference in the aspect of resource scheduling.

Claims (1)

1. A wind power probabilistic prediction method based on a chaotic firefly algorithm and a Bayesian network is characterized by comprising the following steps:
step 1, acquiring wind speed, wind direction, air temperature and wind power actual power data and carrying out data preprocessing:
step 1.1, collecting historical data of wind speed to form an original wind speed sequence, and performing missing value and abnormal value filling and normalization processing on the original wind speed sequence to obtain a preprocessed wind speed sequence which is recorded as V ═ V1,v2,...vi,...,vN]，viI is more than or equal to 1 and less than or equal to N, and N is the total number of samples, wherein the wind speed data of the ith time point in the preprocessed wind speed sequence V is the wind speed data of the ith time point;
acquiring historical data of wind directions to form an original wind direction sequence, and performing missing value and abnormal value filling and normalization processing on the original wind direction sequence to obtain a preprocessed wind direction sequence which is recorded as F ═ F1,f2,...fi...fN],fiWind direction data of the ith time point in the preprocessed wind direction sequence F are obtained;
collecting historical data of air temperature to form an original air temperature sequence, and performing missing value and abnormal value filling and normalization processing on the original air temperature sequence to obtain a preprocessed air temperature sequence which is marked as T ═ w1,w2,...wi,...wN]，wiThe temperature data of the ith time point in the preprocessed temperature sequence T is obtained;
acquiring actual power historical data of wind power to form an original wind power sequence, and filling missing values and abnormal values and normalizing the original wind power sequence to obtain a preprocessed wind power sequence which is recorded as P ═ P1,p2,...,pi,...,pN]，piWind power data of the ith time point in the wind power sequence P after the pretreatment is obtained;
dividing a data set consisting of the preprocessed wind speed sequence, wind direction sequence, air temperature sequence and wind power sequence into training set data and test set data;
step 1.2, carrying out empirical mode decomposition on the preprocessed wind power sequence P to obtain a data set consisting of k IMF components and a margin; the k IMF components are denoted as
Figure FDA0002787310800000011
Figure FDA0002787310800000012
Is the jth IMF component, and
Figure FDA0002787310800000013
Figure FDA0002787310800000014
is the jth IMF component
Figure FDA0002787310800000015
The wind power decomposition value of the ith time point;
step 2, constructing a Bayesian network model by using training set data:
step 2.1, the jth IMF component
Figure FDA0002787310800000016
Wind power decomposition value of the ith time point
Figure FDA0002787310800000017
And the wind speed data v of the ith time pointiWind direction data fiAnd gas temperature data wiAs an influence factor in the bayesian network model, the jth IMF component is added
Figure FDA0002787310800000018
Wind power decomposition value of the (i + 1) th time point
Figure FDA0002787310800000019
As an output node, thereby constructing a Bayesian network model;
step 2.2, respectively calculating the conditional probability of the output node corresponding to each influence factor when the influence factors take different values according to historical data, thereby obtaining a conditional probability table;
step 2.3, obtaining the jth IMF component according to the conditional probability table and the Bayesian network model
Figure FDA0002787310800000021
Power resolution value of the (i + 1) th time point
Figure FDA0002787310800000022
And taking a power interval corresponding to the highest point on the probability distribution curve as the jth IMF component
Figure FDA0002787310800000023
Wind power decomposition value of the (i + 1) th time point
Figure FDA0002787310800000024
The initial prediction interval of (1);
step 2.4, repeating the steps 2.1 to 2.3, so as to obtain an initial prediction interval of the power decomposition value of the (i + 1) th time point in the k IMF components, taking the sum of the initial prediction intervals of the power decomposition value of the (i + 1) th time point in the k IMF components as the wind power prediction interval of the (i + 1) th time point, and recording the sum as the wind power prediction interval of the (i + 1) th time point
Figure FDA0002787310800000025
Step 3, obtaining an optimal interval amplitude variation range by using a chaotic firefly algorithm:
step 3.1, predicting the interval according to the wind power of the (i + 1) th time point
Figure FDA0002787310800000026
Determining the median of the prediction interval of the (i + 1) th time point
Figure FDA0002787310800000027
Wherein,
Figure FDA0002787310800000028
the predicted value is used as the wind power predicted value of the (i + 1) th time point;
predicting the wind power of the (i + 1) th time point
Figure FDA0002787310800000029
Lower limit of (2)
Figure FDA00027873108000000210
Wind power predicted value divided by (i + 1) th time point
Figure FDA00027873108000000211
The lower ratio of the (i + 1) th time point is recorded as
Figure FDA00027873108000000212
Thereby obtaining the lower limit ratios of the N time points, and selecting the maximum value and the minimum value in the lower limit ratios of the N time points as the lower limit ratio betalowThe variation range of (a);
predicting the wind power of the (i + 1) th time point
Figure FDA00027873108000000213
Upper limit of (2)
Figure FDA00027873108000000214
Wind power predicted value divided by (i + 1) th time point
Figure FDA00027873108000000215
The upper ratio of the time points at i +1 is recorded as
Figure FDA00027873108000000216
Thereby obtaining the upper limit ratios of the N time points, and selecting the maximum value and the minimum value in the upper limit ratios of the N time points as the upper limit ratio betahighThe variation range of (a);
lower limit ratio betalowAnd an upper limit ratio betahighThe method is used as a parameter to be optimized in the chaotic firefly algorithm, and two corresponding variation ranges are used as the variation ranges of the population;
step 3.2, initializing the population:
setting a compression operator as g, a population scale as M, a current evolutionary algebra as T, and a maximum evolutionary algebra as TmaxThe current chaotic search frequency is D, and the maximum chaotic search frequency is DmaxLight intensity ofAbsorption factor gamma, gamma is in the range of 0,1]Maximum attraction ω0Step size factor eta, eta is equal to [0,1 ]]；
When the initialization t is 0, generating M random numbers in the variation range of the population and taking the random numbers as the initial positions of M fireflies, wherein the initial position of the nth fireflies in the t evolution iteration
Figure FDA00027873108000000217
The initial position of the mth firefly is
Figure FDA00027873108000000218
N is more than or equal to 1 and less than or equal to M, M is more than or equal to 1 and less than or equal to M, wherein,
Figure FDA00027873108000000219
the nth firefly is in a lower limit ratio beta in the t evolution iterationlowThe random number in the variation range of (a),
Figure FDA0002787310800000031
the nth firefly at the t evolution iteration is the upper limit ratio betahighThe random number in the variation range of (a),
Figure FDA0002787310800000032
the mth firefly is in the lower limit ratio beta in the t evolution iterationlowThe random number in the variation range of (a),
Figure FDA0002787310800000033
the mth firefly at the tth evolution iteration is the upper limit ratio betahighRandom number in the variation range of (2), the initial position of the firefly population is recorded as
Figure FDA0002787310800000034
Step 3.3, calculating the fluorescence brightness of the firefly:
the luminance of the nth firefly in the t evolution iteration is constructed by using the formula (1)
Figure FDA0002787310800000035
And the luminance of the mth firefly in the tth evolutionary iteration is obtained according to the formula (1)
Figure FDA0002787310800000036
Figure FDA0002787310800000037
In the formula (1), u is a confidence level,
Figure FDA0002787310800000038
the average bandwidth of the prediction interval of the nth firefly in the t evolution iteration is obtained by the formula (2),
Figure FDA0002787310800000039
the predicted interval coverage rate of the nth firefly in the t evolution iteration is obtained by the formula (3),
Figure FDA00027873108000000310
to predict the interval coverage
Figure FDA00027873108000000311
Is obtained by the formula (4), and lambda is the predicted interval coverage
Figure FDA00027873108000000312
Penalty factor when confidence level mu is not reached;
Figure FDA00027873108000000313
Figure FDA00027873108000000314
Figure FDA00027873108000000315
in the formula (2), R is a wind power true value sequence P ═ P1,p2,...,pi,...,pN]The difference between the maximum value and the minimum value in (b),
Figure FDA00027873108000000316
the lower bound of the prediction interval of the (i + 1) th time point of the nth firefly in the t evolution iteration is obtained by the formula (5),
Figure FDA00027873108000000317
the upper bound of the prediction interval of the (i + 1) th time point of the nth firefly in the t evolution iteration is obtained by the formula (6);
in the formula (3), bn,i+1Is the Boolean constant of the (i + 1) th time point of the nth firefly, if the real value p of the wind power of the (i + 1) th time pointi+1The prediction interval range of the (i + 1) th time point of the nth firefly in the t evolution iteration
Figure FDA00027873108000000318
In the interior, then order bn,i+11 is ═ 1; otherwise, let bn,i+1＝0；
Figure FDA0002787310800000041
Figure FDA0002787310800000042
In the formulae (5) and (6),
Figure FDA0002787310800000043
predicting the median of the interval for the (i + 1) th time point of the nth firefly;
according to the initial position and the fitness function of the fireflies, the fitness value of each firefly is used as the initial fluorescence brightness of the firefly
Figure FDA0002787310800000044
The fluorescence brightness of the t-th evolution iteration
Figure FDA0002787310800000045
Minimum value of (1) is noted
Figure FDA0002787310800000046
Wherein,
Figure FDA0002787310800000047
for the t-th iteration of evolution, the firefly population
Figure FDA0002787310800000048
Is located at the optimum position in the (c),
Figure FDA0002787310800000049
to an optimum position
Figure FDA00027873108000000410
Is determined by the first component of (a),
Figure FDA00027873108000000411
to an optimum position
Figure FDA00027873108000000412
A second component of (a); let Igmin(Xgbest) Is the global optimum fitness value after t evolutionary iterations, and
Figure FDA00027873108000000413
make global optimum position
Figure FDA00027873108000000414
Step 3.4, the firefly position moves:
according to the light intensity absorption factor gamma and the maximum attraction omega0Obtaining the mutual attraction degree omega of the mth firefly and the nth firefly; if it is
Figure FDA00027873108000000415
The mth firefly moves toward the nth firefly according to the mutual attraction degree omega; and obtaining the updated position of the mth firefly and using the updated position as the position of the mth firefly in the (t + 1) th evolution iteration and recording the position as the updated position
Figure FDA00027873108000000416
Otherwise, the nth firefly moves towards the mth firefly according to the mutual attraction omega; and obtaining the updated position of the nth firefly and using the updated position as the position of the nth firefly in the (t + 1) th evolution iteration, and recording the position as the position of the nth firefly
Figure FDA00027873108000000417
Thereby obtaining the updated positions of all the fireflies and forming a firefly population during the (t + 1) th evolutionary iteration;
step 3.5, obtaining the brightness of each individual in the firefly population during the t +1 th evolution iteration by using the formula (1), and comparing the brightness among the firefly individuals, so as to find out the fitness value of the firefly with the minimum brightness in the firefly population during the t +1 th evolution iteration, and recording the fitness value as the fitness value
Figure FDA00027873108000000418
Wherein,
Figure FDA00027873108000000419
the optimal position in the firefly population after the t +1 th evolutionary iteration;
if the global optimum fitness value
Figure FDA00027873108000000420
Then will be
Figure FDA00027873108000000421
Is assigned to XgbestThereby updating the global optimum position Xgbest(ii) a Otherwise, the global optimal position XgbestKeeping the same;
step 3.6, the optimal position in the firefly population during the t +1 th evolution iteration
Figure FDA00027873108000000422
Local search is performed nearby:
initialization d ═ 0, at [ -1,1 [ ]]Randomly generating q-dimension chaotic variable in d-th chaotic search in interval
Figure FDA00027873108000000423
q represents a dimension, and q is 1, 2;
the optimal position of the t +1 th evolution iteration
Figure FDA0002787310800000051
Assigning to the optimal coordinate y in the d-th chaotic search(d)And q is 1,
Figure FDA0002787310800000052
when q is equal to 2, the reaction is carried out,
Figure FDA0002787310800000053
Figure FDA0002787310800000054
for the optimal coordinate y in the d-th chaotic search(d)The q-th component of (a);
step 3.7, chaotic search is carried out by using the logistic mapping shown in the formula (7) to obtain chaotic variables during the d +1 th chaotic search
Figure FDA0002787310800000055
Figure FDA0002787310800000056
In the formula (7), s is a constant and is more than 0 and less than or equal to 2;
step 3.8, obtaining the optimal coordinate y in the d +1 th chaotic search by using the formula (8)(d+1)Q component of
Figure FDA0002787310800000057
Figure FDA0002787310800000058
In formula (8), when q is 1, xq,maxAs the first component of the initial position of the firefly population
Figure FDA0002787310800000059
Maximum value of (1), xq,minAs the first component of the initial position of the firefly population
Figure FDA00027873108000000510
Minimum value of (1); when q is 2, xq,maxSecond component of initial position of firefly population
Figure FDA00027873108000000511
Maximum value of (1), xq,minSecond component of initial position of firefly population
Figure FDA00027873108000000512
Minimum value of (1); g(t)Compressing the operator after the t-th evolution iteration and obtaining the result by using an equation (9):
Figure FDA00027873108000000513
in formula (9), L is a constant greater than 0;
step 3.9, if D is less than DmaxIf so, assigning d +1 to d, and returning to the step 3.7; otherwise, obtaining the optimal coordinate sequence through chaotic search
Figure FDA00027873108000000514
And proceeds to step 3.9.1;
step 3.9.1, calculating the fitness value of the optimal coordinate sequence using equation (1)
Figure FDA00027873108000000515
Finding the minimum fitness value in the fitness values of the optimal coordinate sequence and recording the minimum fitness value as Imin(ybest) Wherein, ybestFor the optimal coordinate sequence
Figure FDA00027873108000000516
The optimal position of (1);
if it is not
Figure FDA00027873108000000517
Then pass through
Figure FDA00027873108000000518
Updating
Figure FDA00027873108000000519
Otherwise, it orders
Figure FDA00027873108000000520
Keeping the same;
step 3.9.2 if the global optimum fitness value
Figure FDA00027873108000000521
Then pass through
Figure FDA00027873108000000522
Updating the global optimal position; otherwise, let the global optimum position XgbestKeeping the same;
step 3.10, if T is less than TmaxAssigning t +1 to t and returning to the step 3.3; otherwise, the global optimal position X in the firefly population is obtainedgbestObtaining the optimal solution of the fitness function
Figure FDA0002787310800000061
And
Figure FDA0002787310800000062
step 4, according to the optimal solution
Figure FDA0002787310800000063
And
Figure FDA0002787310800000064
respectively calculating the optimal lower bound of the prediction interval of the (i + 1) th time point
Figure FDA0002787310800000065
Optimal upper bound of prediction interval of i +1 th time point
Figure FDA0002787310800000066
And by an optimal lower bound of the prediction interval
Figure FDA0002787310800000067
And an optimal upper bound
Figure FDA0002787310800000068
And forming a final prediction interval, and substituting the test set data into the Bayesian network model so as to complete the probabilistic prediction of the wind power.
CN201910541439.8A 2019-06-21 2019-06-21 Wind power probabilistic prediction method based on chaotic firefly algorithm and Bayesian network Active CN110264006B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910541439.8A CN110264006B (en) 2019-06-21 2019-06-21 Wind power probabilistic prediction method based on chaotic firefly algorithm and Bayesian network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910541439.8A CN110264006B (en) 2019-06-21 2019-06-21 Wind power probabilistic prediction method based on chaotic firefly algorithm and Bayesian network

Publications (2)

Publication Number Publication Date
CN110264006A CN110264006A (en) 2019-09-20
CN110264006B true CN110264006B (en) 2021-01-15

Family

ID=67920203

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910541439.8A Active CN110264006B (en) 2019-06-21 2019-06-21 Wind power probabilistic prediction method based on chaotic firefly algorithm and Bayesian network

Country Status (1)

Country Link
CN (1) CN110264006B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112598883B (en) * 2021-02-07 2021-12-31 国网江苏省电力有限公司苏州供电分公司 Power transmission line lightning-fall probability early warning method and early warning system based on Bayesian network
CN113411216B (en) * 2021-06-21 2022-11-04 国网宁夏电力有限公司信息通信公司 Network flow prediction method based on discrete wavelet transform and FA-ELM
CN117744893B (en) * 2024-02-19 2024-05-17 西安热工研究院有限公司 Wind speed prediction method and system for energy storage auxiliary black start
CN118141392B (en) * 2024-05-09 2024-07-09 西南医科大学附属医院 Electroencephalogram identification method

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102562469A (en) * 2011-12-27 2012-07-11 华北电力大学 Short-term wind driven generator output power predicting method based on correction algorithm
CN106251027A (en) * 2016-08-17 2016-12-21 合肥工业大学 Electric load probability density Forecasting Methodology based on fuzzy support vector quantile estimate
CN106295899A (en) * 2016-08-17 2017-01-04 合肥工业大学 Based on genetic algorithm and the wind power probability density Forecasting Methodology supporting vector quantile estimate
CN108828436A (en) * 2018-06-27 2018-11-16 桂林电子科技大学 Analog-circuit fault diagnosis method based on the adaptive glowworm swarm algorithm of chaos cloud
CN109242200A (en) * 2018-09-29 2019-01-18 华北电力大学 A kind of wind power interval prediction method of Bayesian network forecasting model
CN109816177A (en) * 2019-02-22 2019-05-28 广东电网有限责任公司 A kind of Load aggregation quotient short-term load forecasting method, device and equipment
CN109886452A (en) * 2019-01-04 2019-06-14 国网山东省电力公司电力科学研究院 A kind of ultrashort-term wind power probability forecasting method and system based on experience dynamic modeling

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102562469A (en) * 2011-12-27 2012-07-11 华北电力大学 Short-term wind driven generator output power predicting method based on correction algorithm
CN106251027A (en) * 2016-08-17 2016-12-21 合肥工业大学 Electric load probability density Forecasting Methodology based on fuzzy support vector quantile estimate
CN106295899A (en) * 2016-08-17 2017-01-04 合肥工业大学 Based on genetic algorithm and the wind power probability density Forecasting Methodology supporting vector quantile estimate
CN108828436A (en) * 2018-06-27 2018-11-16 桂林电子科技大学 Analog-circuit fault diagnosis method based on the adaptive glowworm swarm algorithm of chaos cloud
CN109242200A (en) * 2018-09-29 2019-01-18 华北电力大学 A kind of wind power interval prediction method of Bayesian network forecasting model
CN109886452A (en) * 2019-01-04 2019-06-14 国网山东省电力公司电力科学研究院 A kind of ultrashort-term wind power probability forecasting method and system based on experience dynamic modeling
CN109816177A (en) * 2019-02-22 2019-05-28 广东电网有限责任公司 A kind of Load aggregation quotient short-term load forecasting method, device and equipment

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Chaotic Bayesian optimal prediction method and its application in hydrological time series;Xiao-Hua Yang等;《Computers and Mathematics with Applications》;20110430;第61卷;第1975-1978页 *
基于混合粒子群优化的贝叶斯网络结构学习方法;尉永清 等;《小型微型计算机系统》;20180930;第39卷(第9期);第2060-2065页 *
基于混合自适应Memetic算法的贝叶斯网络结构学习;沈佳杰 等;《系统工程与电子技术》;20120630;第34卷(第6期);第1293-1297页 *

Also Published As

Publication number Publication date
CN110264006A (en) 2019-09-20

Similar Documents

Publication Publication Date Title
CN110264006B (en) Wind power probabilistic prediction method based on chaotic firefly algorithm and Bayesian network
CN110807554B (en) Generation method and system based on wind power/photovoltaic classical scene set
CN109447394A (en) A kind of earth observation satellite mission effectiveness assessment indicator system and analysis method
US20230037193A1 (en) Wind power output interval prediction method
CN107886160B (en) BP neural network interval water demand prediction method
CN115374995A (en) Distributed photovoltaic and small wind power station power prediction method
CN112418495A (en) Building energy consumption prediction method based on longicorn stigma optimization algorithm and neural network
CN110084426A (en) A kind of wind power interval prediction method based on industry and mining city
CN114781723A (en) Short-term photovoltaic output prediction method based on multi-model fusion
CN114578087B (en) Wind speed uncertainty measurement method based on non-dominant sorting and stochastic simulation algorithm
CN115146542A (en) Photovoltaic power ultra-short term prediction method, system, equipment and medium
CN115310355A (en) Multi-energy coupling-considered multi-load prediction method and system for comprehensive energy system
CN107292434A (en) A kind of intelligent Forecasting of wind power output
CN117910668A (en) Power system evolution path planning method considering multiple uncertainty factors
CN110555566B (en) B-spline quantile regression-based photoelectric probability density prediction method
CN111582567B (en) Wind power probability prediction method based on hierarchical integration
CN113297805A (en) Wind power climbing event indirect prediction method
CN111476437A (en) Method and device for predicting ultra-short term interval of photovoltaic power generation power
CN110059972B (en) Daily solar radiation resource assessment method based on functional deep belief network
CN112001531A (en) Wind power short-term operation capacity credibility evaluation method based on effective load capacity
CN116565850A (en) Wind power ultra-short-term prediction method based on QR-BLSTM
CN110188967B (en) Power load probabilistic prediction method based on chaos crowd algorithm and Bayesian network
CN113449466B (en) Solar radiation prediction method and system for optimizing RELM based on PCA and chaos GWO
CN114372615A (en) Short-term photovoltaic power prediction method and system considering spatial correlation
CN113177675A (en) Air conditioner cold load prediction method based on optimization neural network of longicorn group algorithm

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant