CN106920008A - A kind of wind power forecasting method based on Modified particle swarm optimization BP neural network - Google Patents
A kind of wind power forecasting method based on Modified particle swarm optimization BP neural network Download PDFInfo
- Publication number
- CN106920008A CN106920008A CN201710113432.7A CN201710113432A CN106920008A CN 106920008 A CN106920008 A CN 106920008A CN 201710113432 A CN201710113432 A CN 201710113432A CN 106920008 A CN106920008 A CN 106920008A
- Authority
- CN
- China
- Prior art keywords
- particle
- neural network
- wind power
- method based
- swarm optimization
- 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
- 239000002245 particle Substances 0.000 title claims abstract description 129
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 64
- 238000005457 optimization Methods 0.000 title claims abstract description 32
- 238000013277 forecasting method Methods 0.000 title claims abstract description 17
- 230000006870 function Effects 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 14
- 230000013016 learning Effects 0.000 claims description 5
- 230000001149 cognitive effect Effects 0.000 claims description 3
- 230000009326 social learning Effects 0.000 claims description 3
- 230000005611 electricity Effects 0.000 description 12
- 238000000034 method Methods 0.000 description 9
- 230000006872 improvement Effects 0.000 description 7
- 230000007935 neutral effect Effects 0.000 description 6
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 description 4
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 238000010248 power generation Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 210000000653 nervous system Anatomy 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000001172 regenerating effect Effects 0.000 description 2
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000035800 maturation Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 235000001968 nicotinic acid Nutrition 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Biophysics (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network, (1):It is particle that weights in BP neural network and threshold value are compiled, and is initialized;(2):Train the difference of acquired results and desired value as fitness function using BP neural network, calculate each particle fitness value;(3):The fitness value of each particle is compared with individual optimal particle, global optimum's particle is obtained;(4):Update particle rapidity and position;(5):Judge whether gained global optimum particle meets end condition, if meeting setting end condition, stop computing output best initial weights threshold value, otherwise return to step (2) interative computation;(6):The best initial weights threshold value for utilizing (5) to obtain connects input layer, hidden layer and the output layer of BP neural network, and wind power prediction result is obtained according to BP neural network prediction of result.Convergence rate faster, high precision, and be difficult to be absorbed in local extremum.
Description
Technical field
The present invention relates to new energy PREDICTIVE CONTROL, more particularly to a kind of wind based on Modified particle swarm optimization BP neural network
Electrical power Forecasting Methodology.
Background technology
Wind energy as a kind of regenerative resource, with it is pollution-free, do not need fuel, be not take up the advantages such as arable land, increasingly
Paid close attention to by people, reserves are also very huge, counted according to related data, the radiation energy from outer space is 1.5* every year
1018KWh, therein 2.5%, i.e. 3.8*1016The energy of kWh produces about 4.3*l0 by Atmospheric Absorptionl2The wind energy of kWh.I
The wind energy content that state can actually develop is 2.53 hundred million kW, occupies the 7.8% of the total wind energy resources of 10m height layers.
The semi-annual report of WWEA's issue points out that global Wind Power Generation Industry increases installation 63013MW, year for 2015 newly
Degree growth rate of market is up to 22%.American market whole year reaches 8.6GW, and Germany exceedes is expected, and 6GW is increased newly, including 2.3GW
Marine project.To the end of the year in 2015, the accumulative installed capacity of global wind-powered electricity generation reached 432419MW, and accumulative annual growth reaches
17%.At the same time, by 2015, used as the first big country of wind-power electricity generation installed capacity, China's wind-powered electricity generation adds up installed capacity
145.1GW is reached, single year wind power generation capacity accounts for the 3.3% of whole generated energy up to 186,300,000,000 kilowatt hours.
With the increasingly maturation of wind power technology, the continuous reduction of its operating cost, wind-power electricity generation presents powerful increasing
The impetus long, the also constantly increase of proportion in electric power market demand, and the negative effect of Wind Power Generation on Power System is also therewith
Highlight.Because wind energy has stronger intermittence, randomness and uncontrollability, wind power also has larger fluctuation, and this will
Badly influence the safe and stable operation of power system.Wind-power electricity generation power is predicted, can be changed into unknown, it is known that having
Effect mitigates the influence that such electricity generation system accesses power network, improves the utilization rate of regenerative resource, obtains bigger economic benefit.Cause
This, carry out Wind power forecasting to study and step up its precision of prediction is a great far-reaching job.
BP neural network (Back Propagation Neural Network, BPNN) is with its good Nonlinear Mapping
Ability, adaptive learning ability, are widely used in the generating prediction of the electricity generation systems such as wind-force, yet with being substantially for BP algorithm
Gradient descent method, makes it not ideal enough with the aspect of precision two in the speed of service.
The content of the invention
The purpose of the present invention is exactly to solve the above problems, there is provided one kind is based on Modified particle swarm optimization BP neural network
Wind power forecasting method, predict the outcome with more preferable nonlinear fitting ability, faster convergence rate and Geng Gao's is pre-
Precision is surveyed, and is difficult to be absorbed in local optimum.
To achieve these goals, the present invention is adopted the following technical scheme that:
A kind of wind power forecasting method based on Modified particle swarm optimization BP neural network, comprises the following steps:
Step (1):It is particle that weights in BP neural network and threshold value are compiled, and it is initialized;
Step (2):With BP neural network training to the wind power prediction output result and wind-powered electricity generation obtained by training set training
The difference of power desired value calculates each particle fitness value as fitness function;
Step (3):The fitness value of each particle is compared with individual optimal particle, global optimum's particle is obtained;
Step (4):Particle rapidity and position are updated according to preset rules;
Step (5):Judge whether gained global optimum particle meets setting end condition, if meeting setting end condition,
Then stop computing output best initial weights threshold value, otherwise return to step (2) interative computation;
Step (6):The best initial weights threshold value connection input layer of BP neural network that is obtained using step (5), hidden layer with
Output layer, wind power prediction result is obtained according to BP neural network prediction of result.
Initialization content includes particle populations scale, initial position and initial velocity in the step (1).
Fitness function is in the step (2):
In formula:OkIt is k-th wind power desired output of node, ykIt is the k-th wind power prediction output of node.
In the step (3) if each particle fitness value better than it is individual it is optimal if the particle is set to it is individual optimal,
Again by individuality it is optimal be compared with global optimum, if be better than global optimum if the particle is set to global optimum.
The formula of particle rapidity renewal foundation is as follows in the step (4):
vi(t+1)=ω vi(t)+c1r1(pb(t)-xi(t))+c2r2(pg(t)-xi(t))
In formula:viT () is the speed of i-th particle in the t times iteration;xiY () is i-th particle in the t times iteration
Position;ω is inertia weight;c1、c2Respectively the cognitive learning factor and social learning's factor;r1With r2For between [0,1]
Random number, pgT () is the individual optimal value in the t times iteration;pgT () is the global optimum in the t times iteration.
c1、c2Value be c1=c2=2.
The formula of particle position renewal institute foundation is as follows in the step (4):
xi(t+1)=xi(t)+vi(t+1)
In formula:xi(t+1) it is the position of i-th particle in the t+1 times iteration.
Wherein ω is defined as follows:
Wherein n is total number of particles, and α is [0,1] interval random number, ft(Xi) for the t time iteration when i-th particle correspondingly
Fitness value, ft(Xbest) for the t times iteration when the corresponding fitness value of optimal particle.
Input layer is determined by input data in the step (6), input layer can select prediction time it is previous when
Carve, the wind power value at preceding two moment and first three moment.
The selection of node in hidden layer is determined by below equation in the step (6):
Wherein m is input layer number, and q is output layer nodes, and a makes up the number or amount for examination, and span is 1~10.
The present invention will improve particle cluster algorithm and is combined with BP neural network Power Output for Wind Power Field is predicted, improvement
The situation of particle cluster algorithm fitness value as obtained by each iteration judges the speed and step-length of next particle search, improves effect
Rate and precision, it is to avoid it is absorbed in local optimum, with the weights and threshold of the optimizing ability optimization neural network for improving particle cluster algorithm
Value, achieves predicting the outcome better than BP neural network.
Beneficial effects of the present invention:
The present invention with can Automatic adjusument weight improve particle cluster algorithm substitute BP neural network in by gradient decline
Error function to adjust weight threshold, can rapidly search best initial weights threshold value, improving particle cluster algorithm can then avoid
It is absorbed in local extremum in searching process, has using being predicted the outcome obtained by prediction algorithm of the invention preferably non-linear
The precision of prediction of capability of fitting, faster convergence rate and Geng Gao.
Brief description of the drawings
Fig. 1 is BP neural network structural representation;
Fig. 2 is the flow chart of Modified particle swarm optimization BP algorithm;
Fig. 3 is that BP neural network predicts wind power;
Fig. 4 is that particle cluster algorithm Optimized BP Neural Network predicts wind power;
Fig. 5 predicts wind power to improve particle cluster algorithm Optimized BP Neural Network;
Fig. 6 is three kinds of absolute error contrasts of model;
Fig. 7 is applied to convergence during BP networks for standard particle colony optimization algorithm;
Fig. 8 is applied to convergence during BP networks for Modified particle swarm optimization algorithm.
Specific embodiment
The invention will be further described with embodiment below in conjunction with the accompanying drawings.
The problem that pin prior art is present, the present invention proposes a kind of wind-powered electricity generation based on Modified particle swarm optimization BP neural network
Power forecasting method, compared with the BP neural network prediction algorithm of standard particle group's optimization, proposes the prediction of optimal prediction model
Result has the precision of prediction of more preferable nonlinear fitting ability, faster convergence rate and Geng Gao, and is difficult to be absorbed in part most
It is excellent, the validity and practicality of MATLAB simulating, verifyings carried algorithm of the invention and model.
Artificial neural network is to the simulation of biological nervous system based on bionics.It is similar with biological nervous system, people
There is substantial amounts of neuron and complete neutral net, the input of each neuron connected and composed by complicated in artificial neural networks
The single output of correspondence, so as to carry out distributed parallel information processing, while neutral net also has the energy of study and self adaptation
Power.With the various demands of engineering, scholars have pointed out various neural network models, and wherein back propagation network is that BP networks are
Propose various artificial network's models in most typically, most widely used one kind, it have good non-linear mapping capability, from
Adaptive learning ability, is well suited for the generating prediction of the electricity generation systems such as wind-force.
As shown in figure 1, BP neural network is general by up of three layers, including:Input layer, hidden layer, output layer, wherein being input into
Layer has m node, and hidden layer has n node, and output layer has q node, ωmnAnd ωnqRespectively input layer and hidden layer it
Between, the connection weight between hidden layer and output layer, bnAnd bqThreshold value respectively between each node layer of hidden layer and output layer.
If one group of input vector is X=(x1,x2,…xm), output layer output vector is Y=(y1,y2,…yq), expect defeated
Outgoing vector is O=(o1,o2,…oq).BP neural network workflow is as follows:
Neutral net is trained first, the process of training is divided into the positive transmission of information and two mistakes of error back propagation
Journey, has for the positive transmission of information:
N-th node input of hidden layer has:
It exports expression formula:
Wherein f (hn) for hidden layer excitation function (excitation function is set to Sigmoid type functions here) be used for describe
Relation between layers, the input of q-th node of output layer:
It exports expression formula:
When the reality output of neutral net is inconsistent with desired output or error is more than setting value, BP neural network will
Error back propagation is shared to all units of each layer, by the error signal amendment each unit weights and threshold value of each layer unit,
Traditional BP neural network adjusts weight threshold using the error function declined by gradient, for K training sample, error function
It is as follows:
(D is w to the correction of each weight thresholdmn、wnq、bn、bqIt is referred to as)
When the error of network output is reduced in claimed range or reaches study number of times set in advance, training stops.
It is worth noting that, the BP neural network for determining model, its input layer and output layer nodes it has been determined that
And hidden layer neuron start node number then needs rule of thumb equation below to determine, then incrementally increased using " trial and error procedure " and
Reduce hidden layer node number and cause that the error of neutral net is minimum, it is final to determine.
Wherein the span of a is 1~10.
Particle swarm optimization algorithm (Particle Swarm Optimization, PSO) by Dr.Eberhart with
Dr.Kennedy proposes that its basic thought is, from RANDOM SOLUTION, to look for by the cooperation between personal and group in nineteen ninety-five
The optimal solution realized with information sharing.When with PSO solving-optimizing problems, the random particle of algorithm initialization a group is each
Individual particle has speed and a position of oneself, and particle updates speed and the position of oneself by each iteration, each particle according to
Two simply rules update its speed and position respectively:
1) the preferred plan movement that particle finds in itself is followed, individual optimal value is thus obtained;
2) particle behaved oneself best in all particles is followed, global optimum is thus obtained.Calculated according to particle position
The fitness value of particle, so as to weigh the quality of the particle, by this method, all of particle reaches globally optimal solution.
I-th speed of particle is as follows with location updating formula:
vi(t+1)=ω vi(t)+c1r1(pb(t)-xi(t))+c2r2(pg(t)-xi(t))
xi(t+1)=xi(t)+vi(t+1)
In formula:viT () is the speed of i-th particle in the t times iteration;ω is inertia weight;c1、c2It is respectively cognitive
Studying factors and social learning's factor, typically take c1=c2=2;r1With r2It is the random number between [0,1];pgT () changes for the t times
Individual optimal value in generation;pgT () is the global optimum in the t times iteration.
Wherein, the size of ω determines that particle is more biased towards in global search or Local Search, c1Determine particle to certainly
The degree that body is relied on, c2Determine particle to the global degree for relying on.
The present invention carry Modified particle swarm optimization algorithm (Modified Particle Swarm Optimization,
MPSO):Particle cluster algorithm is mainly characterized by that principle is simple, and parameter is few, and convergence rate is very fast, it is easy to accomplish, but it is present such as
Easily be absorbed in locally optimal solution and the defect such as be difficult to jump out, currently the improvement of particle cluster algorithm is concentrated mainly on allow inertia weight ω with
Iterations linearly reduces, and which makes following improvement to particle cluster algorithm according to the defect of population:
Wherein, ωmax,ωminThe respectively maxima and minima of inertia weight, t is current iteration number of times, and T is maximum
Iterations.
This improvement makes particle search step-length increase with iterations and progressively reduce, and can to a certain extent avoid grain
Son is absorbed in local extremum, but is once absorbed in and remains difficult to jump out, and in this algorithm ω change only and iterations linear correlation,
The optimization problem with complex nonlinear variation characteristic can not be better adapted to.For this present invention is proposed below according to fitness value
The particle cluster algorithm of adjustment:
Wherein n is total number of particles, ft(Xi) for the t times iteration when the corresponding fitness value of i-th particle, ft(Xbest) be
Optimal particle corresponding fitness value during the t times iteration.
K in formulatFor judging i-th dispersion degree of particle fitness value in the t times iteration, KtIt is bigger to illustrate this
Particle gets over diverging in iteration, smaller, illustrates in current iteration that particle is got over and levels off to optimal particle.ωtWith e as the truth of a matter, using Kt
With Kt-1Ratio as reference, understand ω can be so limited in [0,1] by the character of exponential function it is interval, in order that particle
Group's algorithm is more efficient, and in global search, step-length is bigger, is added in formulaWherein α is [0,1] interval random number, is widened
ω scopes are interval to [0,1.5].
For ωtIf, Kt>Kt-1, i.e. Kt/Kt-1>1 the t times iteration of explanation dissipates relative to the t-1 times, now Kt/Kt-1
The bigger ω of ratiotSmaller, i.e., step-size in search is smaller, and population is got over and is intended to Local Search;If Kt<Kt-1, i.e. Kt/Kt-1<1 is said
Bright the t times iteration restrains relative to the t-1 times, now Kt/Kt-1The smaller ω of ratiotBigger, i.e., step-size in search is bigger, and population is got over
It is intended to global search.This causes that ω is not to reduce with the increase of iterations simply, but is adapted to according to specific function
Angle value determines.Adopt the particle cluster algorithm convergence rate after improving in this way to accelerate, and be difficult to be absorbed in local extremum.
BP neural network is existed and was learnt using the error function adjustment connection weight ω and threshold value b declined by gradient
Journey convergence rate is slow, network training is easily absorbed in local extremum, the structure of network is difficult to determine, network generalization is it is difficult to ensure that wait
Defect.Regarding to the issue above, BP neural network and improvement particle cluster algorithm are combined herein, uses particle swarm optimization algorithm generation
Improve the performance of BP algorithm for the parameter of the gradient descent method training neutral net in BP algorithm, make it be difficult to be absorbed in part
Minimum, strengthens generalization ability, reaches the purpose of optimization network.
As shown in Fig. 2 a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network, specific steps
It is as follows:
Step (1):It is particle that weights in BP neural network and threshold value are compiled, and it is initialized, in initialization
Appearance includes particle populations scale, initial position and initial velocity;
Step (2):Train the difference of acquired results and desired value as fitness function using BP neural network, calculate each grain
Sub- fitness value;
Step (3):The fitness value of each particle is compared with individual optimal particle, will if optimal better than individuality
The particle is set to individual optimal, then is compared individuality is optimal with global optimum, sets the particle if global optimum is better than
It is global optimum;
Step (4):According to Policy Updates particle rapidity and position;
Step (5):Judge whether gained global optimum particle meets end condition, if satisfaction imposes a condition, stop fortune
Calculate output optimum results, otherwise return to step (2) interative computation;
Step (6):Input layer, hidden layer and the output of BP neural network are connected using the best initial weights threshold value of above-mentioned acquisition
Layer, result is obtained according to BP neural network prediction of result.
Below emulating proof:
In order to prove the validity of put forward algorithm, herein by taking the wind power plant of state electricity Jinan Changqing as an example, using MATLAB language
Algorithm routine is write, three kinds of forecast models are constructed respectively:BP neural network forecast model, particle cluster algorithm optimization BP nerve nets
Network forecast model (PSO-BP), improves particle cluster algorithm Optimized BP Neural Network forecast model (MPSO-BP), to the one section time
Interior wind power output power is predicted emulation.From when power sample is for state's electricity Jinan Changqing wind power plant 1 day zero January in 2015,
Data to following 72 hours every collection in 15 minutes once.Network is trained using preceding 260 data, 28 afterwards
Data are used for predicting that i.e. 260 groups of training sample, test sample sum K is 28 groups, and by experimental debugging, hidden layer finally determines
It is 8 nodes.It is 2000 wherein to set BP networks maximum frequency of training, training precision 0.002, learning rate 0.01.To prediction
The performance evaluation of result uses mean absolute error MAPE, root-mean-square error MSE and degree of correlation R2。
Wherein ESS is regression sum of square, and TSS is overall quadratic sum.
Tri- kinds of models of BP, PSO-BP, MPSO-BP are emulated respectively, three kinds of such as Fig. 3 respectively that predict the outcome of model,
Shown in Fig. 4, Fig. 5.Fig. 6 is three kinds of contrasts of model absolute error.Fig. 7 and Fig. 8 show respectively particle when being applied to BP networks
The convergence of the particle cluster algorithm after group's algorithm and improvement.Three kinds of each results for running 20 times of model are as shown in table 1.
The interpretation of result of 1 three kinds of table model running 20 times
Fig. 3 is the result that BP neural network Forecasting Methodology is predicted to wind power, can be calculated its average absolute mistake
Difference MAPE is 0.0507, and root-mean-square error MSE is 0.0692, and prediction curve is with the degree of fitting R2 of desired output curve
0.8980;Fig. 4 is the result being predicted using particle cluster algorithm Optimized BP Neural Network method, can be calculated its average absolute
Error MAPE is 0.0327, and root-mean-square error MSE is 0.0469, and prediction curve is with the degree of fitting R2 of desired output curve
0.9655;Fig. 5 is the result being predicted using modified particle swarm optiziation Optimized BP Neural Network method, can be calculated it and puts down
Equal absolute error MAPE is 0.0248, and root-mean-square error MSE is 0.0346, the degree of fitting R2 of prediction curve and desired output curve
It is 0.9809.Obviously, the result for optimizing prediction using modified particle swarm optiziation Optimized BP Neural Network is better than population
The result that algorithm optimization BP neural network is predicted with single BP neural network.Calculated using improved population as can be seen from Figure 6
The result that method Optimized BP Neural Network optimizes prediction is overall relative to other two kinds of errors smaller, more steadily.And pass through Fig. 7
The improvement particle cluster algorithm for being carried herein is then illustrated with the contrast of Fig. 8 relative to basic particle group algorithm convergence rate faster.
By table 1 it can also be seen that be used for multiple times modified particle swarm optiziation Optimized BP Neural Network optimize the result of prediction relative to
Other two kinds more stable.
Although above-mentioned be described with reference to accompanying drawing to specific embodiment of the invention, not to present invention protection model
The limitation enclosed, one of ordinary skill in the art should be understood that on the basis of technical scheme those skilled in the art are not
Need the various modifications made by paying creative work or deformation still within protection scope of the present invention.
Claims (10)
1. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network, it is characterized in that, including following step
Suddenly:
Step (1):It is particle that weights in BP neural network and threshold value are compiled, and it is initialized;
Step (2):With BP neural network training to the wind power prediction output result and wind power obtained by training set training
The difference of desired value calculates each particle fitness value as fitness function;
Step (3):The fitness value of each particle is compared with individual optimal particle, global optimum's particle is obtained;
Step (4):Particle rapidity and position are updated according to preset rules;
Step (5):Judge whether gained global optimum particle meets setting end condition, if meeting setting end condition, stop
Only computing exports best initial weights threshold value, otherwise return to step (2) interative computation;
Step (6):The best initial weights threshold value obtained using step (5) connects input layer, hidden layer and the output of BP neural network
Layer, wind power prediction result is obtained according to BP neural network prediction of result.
2. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 1, it is special
Levy and be, initialization content includes particle populations scale, initial position and initial velocity in the step (1).
3. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 1, it is special
Levy and be, fitness function is in the step (2):
In formula:OkIt is k-th desired output of node, ykIt is the k-th prediction output of node.
4. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 1, it is special
Levy and be, in the step (3) if each particle fitness value better than it is individual it is optimal if the particle is set to it is individual optimal, then
By individuality it is optimal be compared with global optimum, if be better than global optimum if the particle is set to global optimum.
5. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 1, it is special
Levy and be, the formula of particle rapidity renewal foundation is as follows in the step (4):
vi(t+1)=ω vi(t)+c1r1(pb(t)-xi(t))+c2r2(pg(t)-xi(t))
In formula:viT () is the speed of i-th particle in the t times iteration;xiT () is the position of i-th particle in the t times iteration
Put;ω is inertia weight;c1、c2Respectively the cognitive learning factor and social learning's factor;r1With r2For random between [0,1]
Number, pgT () is the individual optimal value in the t times iteration;pgT () is the global optimum in the t times iteration.
6. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 5, it is special
Levy and be, c1、c2Value be c1=c2=2.
7. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 5, it is special
Levy and be, the formula of particle position renewal institute foundation is as follows in the step (4):
xi(t+1)=xi(t)+vi(t+1)
In formula:xi(t+1) it is the position of i-th particle in the t+1 times iteration.
8. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 5, it is special
Levy and be, wherein being defined as follows to ω:
Wherein n is total number of particles, and α is [0,1] interval random number, ft(Xi) for the t time iteration when i-th particle is corresponding fits
Answer angle value, ft(Xbest) for the t times iteration when the corresponding fitness value of optimal particle.
9. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 1, it is special
Levy and be, input layer is determined by input data in the step (6).
10. a kind of wind power forecasting method based on Modified particle swarm optimization BP neural network as claimed in claim 1, it is special
Levy and be, the selection of node in hidden layer is determined by below equation in the step (6):
Wherein m is input layer number, and q is output layer nodes, and a makes up the number or amount for examination, and span is 1~10.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710113432.7A CN106920008A (en) | 2017-02-28 | 2017-02-28 | A kind of wind power forecasting method based on Modified particle swarm optimization BP neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710113432.7A CN106920008A (en) | 2017-02-28 | 2017-02-28 | A kind of wind power forecasting method based on Modified particle swarm optimization BP neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106920008A true CN106920008A (en) | 2017-07-04 |
Family
ID=59454346
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710113432.7A Pending CN106920008A (en) | 2017-02-28 | 2017-02-28 | A kind of wind power forecasting method based on Modified particle swarm optimization BP neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106920008A (en) |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107665315A (en) * | 2017-10-31 | 2018-02-06 | 上海应用技术大学 | A kind of based role suitable for Hadoop and the access control method trusted |
CN108364071A (en) * | 2018-02-06 | 2018-08-03 | 南京国电南自电网自动化有限公司 | A kind of adaptive modeling wind power prediction method based on genetic programming algorithm |
CN108665112A (en) * | 2018-05-16 | 2018-10-16 | 东华大学 | Photovoltaic fault detection method based on Modified particle swarm optimization Elman networks |
CN108711874A (en) * | 2018-01-22 | 2018-10-26 | 南京理工大学 | A kind of electric vehicle participates in the power consumption control strategy process of public building routine peak regulation |
CN109002942A (en) * | 2018-09-28 | 2018-12-14 | 河南理工大学 | A kind of short-term load forecasting method based on stochastic neural net |
CN109118384A (en) * | 2018-07-16 | 2019-01-01 | 湖南优利泰克自动化系统有限公司 | A kind of Wind turbines healthy early warning method |
CN109308518A (en) * | 2018-09-13 | 2019-02-05 | 北京理工大学 | A kind of monitoring system and its smoothing parameter optimization method based on probabilistic neural network |
CN109472041A (en) * | 2018-09-10 | 2019-03-15 | 江苏大学 | A kind of double-suction pump impeller optimization method based on improvement particle swarm algorithm |
CN109559797A (en) * | 2018-10-31 | 2019-04-02 | 青岛大学附属医院 | A kind of haemodialysis rehabilitation exercise training method and system, terminal |
CN109886471A (en) * | 2019-01-22 | 2019-06-14 | 中国大唐集团科学技术研究院有限公司火力发电技术研究院 | Fired power generating unit load distribution method based on neural network and intelligent optimization algorithm |
CN110210657A (en) * | 2019-05-22 | 2019-09-06 | 国网冀北电力有限公司 | A kind of power of fan prediction technique, system and computer storage medium based on single machine model |
CN110533484A (en) * | 2019-09-05 | 2019-12-03 | 四川长虹电器股份有限公司 | A kind of product Method for Sales Forecast method based on PCA and improved BP |
CN110794308A (en) * | 2019-11-07 | 2020-02-14 | 交控科技股份有限公司 | Method and device for predicting train battery capacity |
CN111047002A (en) * | 2019-11-22 | 2020-04-21 | 浙江工业大学 | BP neural network-based straightening stroke prediction method |
WO2020082973A1 (en) * | 2018-10-22 | 2020-04-30 | 新智数字科技有限公司 | Load prediction method and apparatus based on neural network |
CN111489021A (en) * | 2020-03-31 | 2020-08-04 | 黑龙江大学 | Beet yield prediction method based on particle swarm optimization BP neural network |
CN111814826A (en) * | 2020-06-08 | 2020-10-23 | 武汉理工大学 | Rapid detection and rating method for residual energy of retired power battery |
CN111930844A (en) * | 2020-08-11 | 2020-11-13 | 罗忠明 | Financial prediction system based on block chain and artificial intelligence |
CN112368903A (en) * | 2018-07-05 | 2021-02-12 | 乌本产权有限公司 | Method for feeding electrical power into a power supply network |
CN112489733A (en) * | 2020-12-14 | 2021-03-12 | 郑州轻工业大学 | Octane number loss prediction method based on particle swarm algorithm and neural network |
CN112580855A (en) * | 2020-11-27 | 2021-03-30 | 国网上海市电力公司 | Cable group steady-state temperature rise prediction method based on self-adaptive variation PSO-BP neural network |
CN112819189A (en) * | 2019-11-15 | 2021-05-18 | 内蒙古电力(集团)有限责任公司内蒙古电力科学研究院分公司 | Wind power output prediction method based on historical predicted value |
CN112926767A (en) * | 2021-01-27 | 2021-06-08 | 天津大学 | Annular fog flow gas phase apparent flow velocity prediction method based on particle swarm BP neural network |
CN113095477A (en) * | 2021-03-17 | 2021-07-09 | 西安理工大学 | Wind power prediction method based on DE-BP neural network |
CN113139606A (en) * | 2021-04-27 | 2021-07-20 | 上海联创设计集团股份有限公司 | Heating and ventilation fan fault detection method and system based on artificial intelligence algorithm |
CN113486601A (en) * | 2021-09-07 | 2021-10-08 | 国网江西省电力有限公司电力科学研究院 | Feeder voltage calculation method and device based on CPSO-BP optimization model |
CN113553270A (en) * | 2021-07-28 | 2021-10-26 | 浙江网商银行股份有限公司 | Target object determination method and device |
CN113570120A (en) * | 2021-07-07 | 2021-10-29 | 上海电机学院 | Wind power prediction method and device based on improved particle swarm optimization |
CN113723534A (en) * | 2021-09-02 | 2021-11-30 | 南京润内克西信息科技有限公司 | Urban raise dust on-line monitoring system based on BP neural network |
CN114819313A (en) * | 2022-04-15 | 2022-07-29 | 南通大学 | Wind power prediction method based on improved PSO (particle swarm optimization) optimized double-weight neural network |
CN117074437A (en) * | 2023-10-11 | 2023-11-17 | 青岛金诺德科技有限公司 | Nondestructive testing method and application thereof in recovery of lithium battery of new energy automobile |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361393A (en) * | 2014-09-06 | 2015-02-18 | 华北电力大学 | Method for using improved neural network model based on particle swarm optimization for data prediction |
CN105223241A (en) * | 2015-09-18 | 2016-01-06 | 南京信息工程大学 | A kind of compensation method of humidity sensor |
CN106022471A (en) * | 2016-05-12 | 2016-10-12 | 大连海事大学 | Wavelet neural network model ship rolling real-time prediction method based on particle swarm optimization algorithm |
-
2017
- 2017-02-28 CN CN201710113432.7A patent/CN106920008A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104361393A (en) * | 2014-09-06 | 2015-02-18 | 华北电力大学 | Method for using improved neural network model based on particle swarm optimization for data prediction |
CN105223241A (en) * | 2015-09-18 | 2016-01-06 | 南京信息工程大学 | A kind of compensation method of humidity sensor |
CN106022471A (en) * | 2016-05-12 | 2016-10-12 | 大连海事大学 | Wavelet neural network model ship rolling real-time prediction method based on particle swarm optimization algorithm |
Non-Patent Citations (2)
Title |
---|
李操: "基于IPSO-BP神经网络模型的风电功率预测研究", 《万方数据HTTP://WWW.WANFANGDATA.COM.CN/DETAILS/DETAIL.DO?_TYPE=DEGREE&ID=Y3022004》 * |
董现等: "基于参数相关性和混合神经网络的结构随机灵敏度分析方法", 《建筑结构学报》 * |
Cited By (39)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107665315B (en) * | 2017-10-31 | 2020-12-15 | 上海应用技术大学 | Role and trust-based access control method suitable for Hadoop |
CN107665315A (en) * | 2017-10-31 | 2018-02-06 | 上海应用技术大学 | A kind of based role suitable for Hadoop and the access control method trusted |
CN108711874A (en) * | 2018-01-22 | 2018-10-26 | 南京理工大学 | A kind of electric vehicle participates in the power consumption control strategy process of public building routine peak regulation |
CN108711874B (en) * | 2018-01-22 | 2021-05-28 | 南京理工大学 | Power utilization control strategy method for electric vehicle participating in conventional peak regulation of public building |
CN108364071A (en) * | 2018-02-06 | 2018-08-03 | 南京国电南自电网自动化有限公司 | A kind of adaptive modeling wind power prediction method based on genetic programming algorithm |
CN108665112A (en) * | 2018-05-16 | 2018-10-16 | 东华大学 | Photovoltaic fault detection method based on Modified particle swarm optimization Elman networks |
CN112368903A (en) * | 2018-07-05 | 2021-02-12 | 乌本产权有限公司 | Method for feeding electrical power into a power supply network |
CN109118384A (en) * | 2018-07-16 | 2019-01-01 | 湖南优利泰克自动化系统有限公司 | A kind of Wind turbines healthy early warning method |
CN109472041A (en) * | 2018-09-10 | 2019-03-15 | 江苏大学 | A kind of double-suction pump impeller optimization method based on improvement particle swarm algorithm |
CN109308518B (en) * | 2018-09-13 | 2021-09-10 | 北京理工大学 | Monitoring system based on probabilistic neural network and smoothing parameter optimization method thereof |
CN109308518A (en) * | 2018-09-13 | 2019-02-05 | 北京理工大学 | A kind of monitoring system and its smoothing parameter optimization method based on probabilistic neural network |
CN109002942A (en) * | 2018-09-28 | 2018-12-14 | 河南理工大学 | A kind of short-term load forecasting method based on stochastic neural net |
WO2020082973A1 (en) * | 2018-10-22 | 2020-04-30 | 新智数字科技有限公司 | Load prediction method and apparatus based on neural network |
CN109559797A (en) * | 2018-10-31 | 2019-04-02 | 青岛大学附属医院 | A kind of haemodialysis rehabilitation exercise training method and system, terminal |
CN109886471A (en) * | 2019-01-22 | 2019-06-14 | 中国大唐集团科学技术研究院有限公司火力发电技术研究院 | Fired power generating unit load distribution method based on neural network and intelligent optimization algorithm |
CN110210657A (en) * | 2019-05-22 | 2019-09-06 | 国网冀北电力有限公司 | A kind of power of fan prediction technique, system and computer storage medium based on single machine model |
CN110533484A (en) * | 2019-09-05 | 2019-12-03 | 四川长虹电器股份有限公司 | A kind of product Method for Sales Forecast method based on PCA and improved BP |
CN110794308A (en) * | 2019-11-07 | 2020-02-14 | 交控科技股份有限公司 | Method and device for predicting train battery capacity |
CN112819189A (en) * | 2019-11-15 | 2021-05-18 | 内蒙古电力(集团)有限责任公司内蒙古电力科学研究院分公司 | Wind power output prediction method based on historical predicted value |
CN111047002A (en) * | 2019-11-22 | 2020-04-21 | 浙江工业大学 | BP neural network-based straightening stroke prediction method |
CN111047002B (en) * | 2019-11-22 | 2023-04-07 | 浙江工业大学 | BP neural network-based alignment stroke prediction method |
CN111489021A (en) * | 2020-03-31 | 2020-08-04 | 黑龙江大学 | Beet yield prediction method based on particle swarm optimization BP neural network |
CN111814826A (en) * | 2020-06-08 | 2020-10-23 | 武汉理工大学 | Rapid detection and rating method for residual energy of retired power battery |
CN111814826B (en) * | 2020-06-08 | 2022-06-03 | 武汉理工大学 | Rapid detection and rating method for residual energy of retired power battery |
CN111930844B (en) * | 2020-08-11 | 2021-09-24 | 肖岩 | Financial prediction system based on block chain and artificial intelligence |
CN111930844A (en) * | 2020-08-11 | 2020-11-13 | 罗忠明 | Financial prediction system based on block chain and artificial intelligence |
CN112580855A (en) * | 2020-11-27 | 2021-03-30 | 国网上海市电力公司 | Cable group steady-state temperature rise prediction method based on self-adaptive variation PSO-BP neural network |
CN112489733A (en) * | 2020-12-14 | 2021-03-12 | 郑州轻工业大学 | Octane number loss prediction method based on particle swarm algorithm and neural network |
CN112926767A (en) * | 2021-01-27 | 2021-06-08 | 天津大学 | Annular fog flow gas phase apparent flow velocity prediction method based on particle swarm BP neural network |
CN113095477A (en) * | 2021-03-17 | 2021-07-09 | 西安理工大学 | Wind power prediction method based on DE-BP neural network |
CN113095477B (en) * | 2021-03-17 | 2024-02-09 | 西安理工大学 | Wind power prediction method based on DE-BP neural network |
CN113139606A (en) * | 2021-04-27 | 2021-07-20 | 上海联创设计集团股份有限公司 | Heating and ventilation fan fault detection method and system based on artificial intelligence algorithm |
CN113570120A (en) * | 2021-07-07 | 2021-10-29 | 上海电机学院 | Wind power prediction method and device based on improved particle swarm optimization |
CN113553270A (en) * | 2021-07-28 | 2021-10-26 | 浙江网商银行股份有限公司 | Target object determination method and device |
CN113723534A (en) * | 2021-09-02 | 2021-11-30 | 南京润内克西信息科技有限公司 | Urban raise dust on-line monitoring system based on BP neural network |
CN113486601A (en) * | 2021-09-07 | 2021-10-08 | 国网江西省电力有限公司电力科学研究院 | Feeder voltage calculation method and device based on CPSO-BP optimization model |
CN114819313A (en) * | 2022-04-15 | 2022-07-29 | 南通大学 | Wind power prediction method based on improved PSO (particle swarm optimization) optimized double-weight neural network |
CN117074437A (en) * | 2023-10-11 | 2023-11-17 | 青岛金诺德科技有限公司 | Nondestructive testing method and application thereof in recovery of lithium battery of new energy automobile |
CN117074437B (en) * | 2023-10-11 | 2024-01-12 | 青岛金诺德科技有限公司 | Nondestructive testing method and application thereof in recovery of lithium battery of new energy automobile |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106920008A (en) | A kind of wind power forecasting method based on Modified particle swarm optimization BP neural network | |
Yang et al. | A novel competitive swarm optimized RBF neural network model for short-term solar power generation forecasting | |
Zhang et al. | A comprehensive wind speed prediction system based on Monte Carlo and artificial intelligence algorithms | |
Liu et al. | Short-term photovoltaic power prediction on modal reconstruction: A novel hybrid model approach | |
CN107316099A (en) | Ammunition Storage Reliability Forecasting Methodology based on particle group optimizing BP neural network | |
CN107704875A (en) | Based on the building load Forecasting Methodology and device for improving IHCMAC neutral nets | |
CN104636985A (en) | Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network | |
CN104820877A (en) | Photovoltaic system generation power prediction method based on cloud adaptive PSO-SNN | |
Tian et al. | Variable frequency wind speed trend prediction system based on combined neural network and improved multi-objective optimization algorithm | |
CN114784823A (en) | Micro-grid frequency control method and system based on depth certainty strategy gradient | |
Ning et al. | GA-BP air quality evaluation method based on fuzzy theory. | |
CN109858665A (en) | Photovoltaic short term power prediction technique based on Feature Selection and ANFIS-PSO | |
Wen et al. | Forecasting the annual household electricity consumption of Chinese residents using the DPSO-BP prediction model | |
Liu et al. | Wind power short-term forecasting based on LSTM neural network with dragonfly algorithm | |
CN116454863A (en) | Optimal weight determining method of wind power combination prediction model based on improved hawk optimization algorithm | |
CN115995810A (en) | Wind power prediction method and system considering weather fluctuation self-adaptive matching | |
CN109408896B (en) | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production | |
Zhao et al. | Super short term combined power prediction for wind power hydrogen production | |
CN106786499A (en) | Based on the short-term wind power forecast method for improving AFSA optimizations ELM | |
CN117132132A (en) | Photovoltaic power generation power prediction method based on meteorological data | |
CN116865343A (en) | Model-free self-adaptive control method, device and medium for distributed photovoltaic power distribution network | |
CN116681154A (en) | Photovoltaic power calculation method based on EMD-AO-DELM | |
Ren et al. | Time series prediction based on echo state network tuned by divided adaptive multi-objective differential evolution algorithm | |
CN110532057A (en) | A kind of resource usage amount prediction technique of container | |
Yin et al. | Quantum-inspired distributed policy-value optimization learning with advanced environmental forecasting for real-time generation control in novel power systems |
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: 20170704 |