CN106327357A - Load identification method based on improved probabilistic neural network - Google Patents
Load identification method based on improved probabilistic neural network Download PDFInfo
- Publication number
- CN106327357A CN106327357A CN201610683840.1A CN201610683840A CN106327357A CN 106327357 A CN106327357 A CN 106327357A CN 201610683840 A CN201610683840 A CN 201610683840A CN 106327357 A CN106327357 A CN 106327357A
- Authority
- CN
- China
- Prior art keywords
- neural network
- probabilistic neural
- load
- particle
- identification method
- 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.)
- Granted
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 56
- 238000000034 method Methods 0.000 title claims abstract description 42
- 239000002245 particle Substances 0.000 claims abstract description 56
- 238000012549 training Methods 0.000 claims abstract description 29
- 230000005611 electricity Effects 0.000 claims abstract description 13
- 238000005457 optimization Methods 0.000 claims abstract description 8
- 229910002056 binary alloy Inorganic materials 0.000 claims abstract description 5
- 238000013507 mapping Methods 0.000 claims abstract description 5
- 238000009499 grossing Methods 0.000 claims description 29
- 230000006872 improvement Effects 0.000 claims description 15
- 230000001537 neural effect Effects 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 3
- 230000000505 pernicious effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 26
- 230000007935 neutral effect Effects 0.000 description 8
- 238000007654 immersion Methods 0.000 description 7
- 210000002569 neuron Anatomy 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 2
- 239000000470 constituent Substances 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 230000000977 initiatory effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 241000220317 Rosa Species 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- GOLXNESZZPUPJE-UHFFFAOYSA-N spiromesifen Chemical compound CC1=CC(C)=CC(C)=C1C(C(O1)=O)=C(OC(=O)CC(C)(C)C)C11CCCC1 GOLXNESZZPUPJE-UHFFFAOYSA-N 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
- 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
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Feedback Control In General (AREA)
Abstract
The invention discloses a load identification method based on an improved probabilistic neural network. The load identification method comprises steps of adopting a binary system to perform coding on a load type in an electricity usage network, establishing a coding library of load types, collecting an electric parameter of each load type, establishing a non-linear mapping relation between the electric parameter and the code of each load type, creating a probabilistic neural network, training the probabilistic neural network to obtain an error function, using the error function as a fitness function of a particle swarm algorithm, adopting the particle swarm algorithm to perform optimization on a smooth factor of the probabilistic neural network to obtain an optimal smooth factor, updating the probabilistic neural network according to the obtained optimal smooth factor to obtain an improved probabilistic neural network and performing recognition on the load type in the electricity usage network on the basis of the probabilistic neural network. The load identification method based on the improved probabilistic neural network can simultaneously identify multiple loads, improves speed of identifying a pernicious load under a multi-load mode, and can better realize recognition of restricted electric appliances and electricity usage control under a condition of multiple electric appliances.
Description
Technical field
The present invention relates to electricity consumption remained capacity and control technical field, particularly relate to a kind of based on improving probabilistic neural network
Load identification method.
Background technology
At present, generally there is the problems such as power consumption is big, management is difficult, potential safety hazard is serious in enterprise, College Apartments.The nearest
Several years, along with gradually expanding and the continuous progress of Power Electronic Technique of high school scale, number of student steeply rose, and uses electric loading
Of a great variety, the management of power use is complicated.The improper uses such as some high-power electric appliances such as immersion heater, hair dryer, electric blanket, the most then
Cause tripping operation, heavy then initiation fire, jeopardize student's personal safety as well as the property safety.The most high-power resistive load belongs to colleges and universities and expressly provides
Violated electric equipment, referred to as bad load, forbid use.Compare legacy powered device, current electrical equipment type and
Characteristic parameter increasingly diversification and complication, use traditional load identification method, has been difficult to identify accurately novel
Electrical equipment.The most correctly identify the type of every kind of electrical equipment, and it is suitably controlled management, for the use of student dormitory
Electricity safety has great importance.Existing bad load recognition methods mainly has power abrupt climatic change, frequency analysis and small echo
Analyze, although these methods can the purpose of remained capacity, but still suffer from certain deficiency.Such as method of discrimination is the simplest
Single, idealization, does not accounts for the current/voltage fluctuation in side circuit and the interference existed, it is impossible to the problems such as networking.
Neutral net has the functions such as self adaptation, self study, parallel processing and associative memory, is applied to remained capacity
Become a kind of important method in intelligent mode identification technology.At remained capacity technical elements, have some at present based on god
Through the solution of network, for example with feed-forward neutral net (BP neutral net), it is anti-that usual employing has conventional forward direction
Feedback neutral net (BP neutral net) is as pattern classifier, but the remained capacity problem that multiple features is inputted, conventional forward direction
There is structure complexity in Feedback Neural Network (BP neutral net), training time length, convergence rate are absorbed in local optimum slowly and easily
Etc. problem, cause net training time length, loadtype discrimination the highest.
Summary of the invention
In view of the deficiencies in the prior art, the invention provides a kind of based on the remained capacity side improving probabilistic neural network
Method, the method can identify multiple load simultaneously, improves the distinguishing speed of bad load under multiple load modes, preferably realizes
Limit the use of electrical equipment identification and power consumption control under Multifunctional electric electrical equipment situation, effectively solve the electricity consumption control of enterprise, business, campus user
Problem, ensures the security of the lives and property.
To achieve these goals, present invention employs following technical scheme:
A kind of load identification method based on improvement probabilistic neural network, comprising:
Use binary system to encode the loadtype in power utilization network, set up the code database of loadtype;
Gather each loadtype electrical quantity in starting moment to a period of time;
Set up the nonlinear mapping relation between the electrical quantity of each loadtype and corresponding coding, create probabilistic neural net
Network;
Being trained described probabilistic neural network, it is thus achieved that the error function of network, described error function is expressed as:In formula,For the desired output corresponding with training set sample, ytK () is probability
Actual output after neural metwork training, q is training sample set number, and S is loadtype number;
Using described error function as the fitness function of particle cluster algorithm, use particle cluster algorithm to described probabilistic neural
The smoothing factor of network is optimized, and obtains the optimal smoothing factor of described probabilistic neural network;
Described probabilistic neural network is updated, it is thus achieved that the probabilistic neural network of improvement according to the acquired optimal smoothing factor;
Based on the probabilistic neural network improved, the loadtype in power utilization network is identified.
Wherein, use particle cluster algorithm that the smoothing factor of described probabilistic neural network is optimized to comprise the following steps:
(S1) span of smoothing factor is set, can random initializtion a group particle { σ in solution space1,σ2,L σN,
Set maximum iteration time and set current iteration number of times k=1;
(S2) using described error function as the fitness function of particle cluster algorithm, particle fitness value is calculated, and with this
Find individual extreme value and colony's extreme value;Wherein, in the search volume of a D dimension, corresponding to the smoothing factor of N number of pattern class
Constituent particle group σ '=(σ1,σ2,L σN), i-th particle is expressed as the vectorial σ of a D dimensioni=[σi1,σi2,L σiD]T, represent
I-th particle position in D dimension search volume, the speed of i-th particle is Vi=[Vi1,Vi2,L,ViD]T, its individual extreme value
For Pi=[Pi1,Pi2,L,PiD]T, the global extremum of population is Pg=[Pg1,Pg2,L,PgD]T;Wherein, N, D are respectively integer;
(S3) particle rapidity and location updating, recalculates particle fitness value, and individual extreme value and colony's extreme value update,
To population of lower generation;More new formula is as follows:
In formula, ω is inertia weight;D=1,2 ..., D;I=1,2 ..., N;K is current iteration number of times;VidFor particle
Speed;c1And c2For the constant of non-negative, referred to as acceleration factor;r1And r2For being distributed in the random number between [0,1];
(S4) current iteration number of times k=k+1 is set;
(S5) check whether and meet termination condition: if reaching maximum iteration time or Jm=0, then stop;Otherwise return
Step (S3).
Wherein, set maximum iteration time is 200~1000 times.
Wherein, set maximum iteration time is 500 times.
Wherein, described loadtype includes that disposable load, disabling load, the combination of multiple disposable load, multiple disabling are born
The combination that the combination carried and disposable load and disabling load.
Wherein, each loadtype electrical quantity in starting moment to 15 second time is gathered.
Wherein, described electrical quantity includes voltage, electric current, power, power factor and harmonic wave.
Compared to prior art, embodiments provide a kind of based on the remained capacity side improving probabilistic neural network
Method, the method can identify multiple load simultaneously, improves the distinguishing speed of bad load under multiple load modes, preferably realizes
Limit the use of electrical equipment identification and power consumption control under Multifunctional electric electrical equipment situation, effectively solve the electricity consumption control of enterprise, business, campus user
Problem, ensures the security of the lives and property.Wherein, probabilistic neural network (probabilistic neural networks, PNN) is
A kind of feed-forward type neutral net developed based on radial basis function network, theoretical foundation is Bayesian Smallest Risk criterion, therefore
PNN network is suitable for pattern classification.It is advantageous that and complete, with linear learning algorithm, the work that nonlinear learning algorithm is done,
Keeping the characteristic such as high accuracy of nonlinear algorithm, the weights that network is corresponding are exactly the distribution of pattern sample simultaneously, it is possible to meet instruction
The requirement processed in real time on white silk.In embodiments of the invention, utilize particle cluster algorithm (Particle Swarm
Optimization, PSO) smoothing factor of PNN network is optimized, it is thus achieved that the PNN network of the optimal smoothing factor will make
Complete the expressing of probability nature of sample space, and then improve the classifying quality of PNN network.
Accompanying drawing explanation
Fig. 1 is the load identification method flow chart based on improvement probabilistic neural network of the embodiment of the present invention;
Fig. 2 is the structural representation of probabilistic neural network (PNN);
Fig. 3 is the flow chart using particle cluster algorithm to be optimized the smoothing factor of PNN network in the embodiment of the present invention;
Fig. 4 is the effect schematic diagram in the specific embodiment of the invention after PNN network training;
Error schematic diagram after PNN network training in Fig. 5 specific embodiment of the invention;
Prediction effect schematic diagram after PNN network training in Fig. 6 specific embodiment of the invention.
Detailed description of the invention
For making the object, technical solutions and advantages of the present invention clearer, the concrete reality to the present invention below in conjunction with the accompanying drawings
The mode of executing is described in detail.The example of these preferred implementations is illustrated in the accompanying drawings.Shown in accompanying drawing and according to
The embodiments of the present invention that accompanying drawing describes are merely exemplary, and the present invention is not limited to these embodiments.
Here, also, it should be noted in order to avoid having obscured the present invention because of unnecessary details, the most only
Show and according to the closely-related structure of the solution of the present invention and/or process step, and eliminate little with relation of the present invention
Other details.
Embodiments provide a kind of load identification method based on improvement probabilistic neural network, as it is shown in figure 1, should
Method includes step:
A (), different loads type coding: use binary system to encode the loadtype in power utilization network, set up negative
Carry the code database of type.Wherein, described loadtype includes disposable load, disabling load, the combination of multiple disposable load, multiple
The combination that the combination of disabling load and disposable load and disabling load.
(b), gather the electrical quantity of each loadtype: gather each loadtype in starting moment to a period of time
Electrical quantity.Here, within a period of time is typically to be set to 20 seconds, most preferably within 15 seconds.Described electrical quantity includes
Voltage, electric current, power, power factor and harmonic wave etc., wherein harmonic wave can be again first harmonic and second harmonic etc..
(c), establishment PNN network: the nonlinear mapping set up between the electrical quantity of each loadtype and corresponding coding is closed
System, creates probabilistic neural network.
Probabilistic neural network (probabilistic neural networks, PNN) is based on radial basis function network development
A kind of feed-forward type neutral net come, as in figure 2 it is shown, the hierarchical model of PNN network is input layer I, mode layer II, summation respectively
Layer III, output layer IV totally four layers.
Input layer I receives the value from training sample, and characteristic vector is passed to network, and its neuron number and sample are vowed
The dimension of amount is equal.Mode layer II calculates input feature value and the matching relationship of each pattern in training set, mode layer II god
Through first number equal to each classification number of training sum, this layer of each mode unit is output as:In formula, WiThe weights connected for input layer I to mode layer II;For smooth because of
Son, it plays vital effect to classification.
Summation layer III is to be added up by the probability belonging to certain class, thus obtains the estimated probability density function of classification mode.Often
One class only one of which summation layer unit, summation layer unit is connected with the pattern layer units being pertaining only to oneself class, and and mode layer
In other unit do not connect, therefore the output of the pattern layer units of to one's name class is simply added by summation layer unit,
And it is unrelated with the output of the pattern layer units belonging to other classification.
The neuron of output layer IV is a kind of competition neurons, and each neuron corresponds respectively to a data type and i.e. divides
Quasi-mode, output layer neuron number equal to the kind number of training sample data, it receive from summation layer output all kinds of generally
Rate density function, that neuron of probability density function maximum is output as 1, and that i.e. corresponding class is sample to be identified
Pattern class, the output of other neuron is all 0.
Mode identification method based on PNN network is widely accepted a kind of decision method in probability statistics, can retouch
State for: assume to have type-scheme θ known to two kindsA、θB, for type feature sample X=(x to be judged1,x2,L,xn):
If hAlAfA(X)>hBlBfB(X), then X ∈ θA;
If hAlAfA(X) < hBlBfB(X), then X ∈ θB;
In formula, hA、hBFor type-scheme θA、θBPrior probability (hA=NA/ N, hB=NB/N);NA、NBFor type-scheme θA、
θBNumber of training;N is training sample sum;lAFor originally belonging to θAFeature samples X be divided into pattern θ mistakenlyBGeneration
The valency factor;lBFor originally belonging to θBFeature samples X be divided into pattern θ mistakenlyAWork factor;fA、fBFor type-scheme θA、
θBProbability density function (Probability Density Function, PDF), usual PDF can not accurately obtain, can only
Its statistical value is sought according to existing feature samples.1962, Parzen proposed a kind of estimated probability from known random sample
The method of density function, as long as number of samples is abundant, the function that the method obtains can with continuously smooth to approach former probability close
Degree function.The PDF estimator obtained by Parzen method is as follows:
In formula, XaiFor type-scheme θAI-th training vector;M is type-scheme θATraining sample number;
For smoothing factor, its value determines the width of the bell curve centered by sample point.
(d), training PNN network, it is thus achieved that error function: according to the most acquired sample data to described probabilistic neural network
Being trained, it is thus achieved that the error function of network, described error function is expressed as:Formula
In,For the desired output corresponding with training set sample, ytK () is the actual output after probabilistic neural network training, q is instruction
Practicing the number of sample set, S is loadtype number.
(e), PSO algorithm optimization smoothing factor: using described error function as the fitness function of particle cluster algorithm, use
The smoothing factor of described probabilistic neural network is optimized by particle cluster algorithm, obtains the optimal smoothing of described probabilistic neural network
The factor.
Particle cluster algorithm (Particle Swarm Optimization, PSO) is computational intelligence field, is widely used
A kind of Swarm Intelligent Algorithm.PSO algorithm is to prey on this population behavior feature from birds to gain enlightenment and for asking
Solving optimization problem, in algorithm, each particle represents a potential solution of problem, and each particle correspondence one is by fitness letter
The fitness value that number determines.The speed of particle determines direction and the distance that particle moves, and speed is with self and other particles
Mobile experience dynamically adjusts, thus realizes individual can optimizing in solution space.PSO algorithm is first at the beginning of can be in solution space
Beginningization a group particle, each example represents a potential optimal solution of extremal optimization problem, with position, speed and fitness value
Three these particle characteristicses of index expression, fitness value is calculated by fitness function, and the quality of its value represents the quality of particle.
Particle moves in solution space, updates a body position by following the tracks of individual extreme value Pbest and colony's extreme value Gbest;Individual extreme value
Pbest refers to calculated fitness value optimal location in individual experienced position, and colony's extreme value Gbest refers in population
The fitness optimal location that all particle search arrive.Particle often updates a position, just calculates a fitness value, and passes through
The relatively fitness value of new particle and the fitness value of individual extreme value, colony's extreme value updates individual extreme value Pbest and colony's extreme value
Gbest position.
Wherein, the process using particle cluster algorithm to be optimized the smoothing factor of described probabilistic neural network includes: as
Shown in Fig. 3, first calculate PNN network error (i.e. calculating particle fitness value) according to current smoothing factor, then update PNN
The smoothing factor of network, finally judges whether to meet termination condition: the most then with meet termination condition smoothing factor as optimum
Smoothing factor;If it is not, then recalculate PNN network error according to the smoothing factor updated, enter iterative cycles, until meeting knot
Bundle condition.Specifically, it comprises the following steps:
(S1) span of smoothing factor is set, can random initializtion a group particle { σ in solution space1,σ2,L σN,
Set maximum iteration time and set current iteration number of times k=1.Wherein, set maximum iteration time is generally 200~1000
In secondary scope, the present embodiment is specifically set as 500 times.
(S2) using described error function as the fitness function of particle cluster algorithm, particle fitness value is calculated, and with this
Find individual extreme value and colony's extreme value;Wherein, in the search volume of a D dimension, corresponding to the smoothing factor of N number of pattern class
Constituent particle group σ '=(σ1,σ2,L σN), i-th particle is expressed as the vectorial σ of a D dimensioni=[σi1,σi2,L σiD]T, represent
I-th particle position in D dimension search volume, the speed of i-th particle is Vi=[Vi1,Vi2,L,ViD]T, its individual extreme value
For Pi=[Pi1,Pi2,L,PiD]T, the global extremum of population is Pg=[Pg1,Pg2,L,PgD]T;Wherein, N, D are respectively integer.
(S3) particle rapidity and location updating, recalculates particle fitness value, and individual extreme value and colony's extreme value update,
To population of lower generation;More new formula is as follows:
In formula, ω is inertia weight;D=1,2 ..., D;I=1,2 ..., N;K is current iteration number of times;VidFor particle
Speed;c1And c2For the constant of non-negative, referred to as acceleration factor;r1And r2For being distributed in the random number between [0,1];
(S4) current iteration number of times k=k+1 is set;
(S5) check whether and meet termination condition: if reaching maximum iteration time or Jm=0, then stop, acquisition is
Excellent smoothing factor;Otherwise return step (S3).
(f), renewal PNN network: update described probabilistic neural network according to the optimal smoothing factor obtained, it is thus achieved that improvement
Probabilistic neural network.
(g), loadtype identification: based on the probabilistic neural network improved, the loadtype in power utilization network is known
Not.
It is described below one of the load identification method based on improvement probabilistic neural network that an example performed as described above provides
Concrete application case.
College Students ' Apartments personnel concentrate, and electricity consumption load type is various, and the management of power use is complicated.Some high-power electric appliances make
With improper easily initiation fire, threatening property and the personal safety of classmates, the most high-power resistive load belongs to violated electricity
Device equipment, referred to as bad load (disabling load), forbids to use.The most correctly identify the type of every kind of electrical equipment, it is entered
Row suitably controls management, and the Electrical Safety for student dormitory has great importance.Present case is chosen several allusion quotations on market
The resistive loads such as the electrical appliance of type, such as immersion heater, electric blanket, hair dryer, the nonlinear load such as computer, mobile phone tests,
The service condition of simulation university dormitory electric loading.
First to encoding with electric loading employing binary system in university dormitory, part electric loading encodes result such as table
Shown in 1.
Table 1 university dormitory electric loading is numbered
Appliance type | Coding |
Notebook computer | 0001 |
Electric cooker | 0010 |
Hair dryer gear I | 0011 |
Hair dryer gear II | 0100 |
Immersion heater type I | 0101 |
Immersion heater Type II | 0110 |
Immersion heater type-iii | 0111 |
Hot-water bottle | 1000 |
Mobile phone | 1001 |
Notebook computer+electric cooker | 1010 |
Notebook computer+immersion heater type I | 1011 |
Hot-water bottle+hair dryer gear II | 1100 |
Mobile phone+notebook computer | 1101 |
Mobile phone+immersion heater Type II | 1110 |
Notebook computer+hair dryer gear I | 1111 |
Use electrical parameters detection equipment to carry out data sampling, gather each loadtype and start the electricity in moment to 15 seconds
The electrical quantitys such as pressure, electric current, power, power factor, harmonic wave, as the input feature value of PNN network, thus set up load electricity ginseng
Nonlinear mapping relation between number with corresponding coding.Specifically, present case collects 22 kinds of different loadtypes combinations, often
Individual type gathers 5 groups of electrical quantitys, forms 110 groups of sample sets altogether.Sample set is the matrix of 110 × 7 dimensions, front 6 row respectively voltage,
Electric current, power, power factor, first harmonic, second harmonic, last is classified as loadtype coding.By 110 groups of sample set orders
Intersection is upset, and chooses 70 groups of sample sets training sample as PNN network, and 40 groups of sample sets are as the checking sample of PNN network.
Utilize the non-linear classification that probabilistic neural network is powerful, by load performance parameter space map to loadtype
Space, forms one and has relatively strong fault tolerance ability and the pattern classification system of structure adaptive ability, thus it is pernicious to realize apartment
Loadtype identification function.The probabilistic neural network utilizing the PSO algorithm optimization with TSP question carries out student dormitory evil
Property remained capacity result as Figure 4-Figure 6, show for convenience of result, loadtype is numbered and is converted into real number.Wherein, Fig. 4
Being the effect schematic diagram (including 70 groups of training samples) after the PNN network training of Optimal improvements, Fig. 5 corresponds to the error of Fig. 4
Schematic diagram, Fig. 6 is the prediction effect schematic diagram (including 40 groups of checking samples) after the PNN network training of Optimal improvements.
As seen from the figure, the PNN network after optimizing, there is certain adaptive learning ability.After training, will training
Data substitute into, as feature input, the PNN network trained, and recognition result is completely correct, and enters with prediction data sample
The when of row checking, recognition result is also the most correct, and the PNN network finally obtained can be used to carry out the prediction of more multisample.
In sum, the load identification method based on improvement probabilistic neural network that the embodiment of the present invention provides, can be simultaneously
Identify multiple load, improve the distinguishing speed of bad load under multiple load modes, preferably realize Multifunctional electric electrical equipment situation
Under limit the use of electrical equipment identification and power consumption control, effectively solve the electricity consumption control problem of enterprise, business, campus user, ensure life
Property safety.
It should be noted that in this article, the relational terms of such as first and second or the like is used merely to a reality
Body or operation separate with another entity or operating space, and deposit between not necessarily requiring or imply these entities or operating
Relation or order in any this reality.And, term " includes ", " comprising " or its any other variant are intended to
Comprising of nonexcludability, so that include that the process of a series of key element, method, article or equipment not only include that those are wanted
Element, but also include other key elements being not expressly set out, or also include for this process, method, article or equipment
Intrinsic key element.In the case of there is no more restriction, statement " including ... " key element limited, it is not excluded that
Including process, method, article or the equipment of described key element there is also other identical element.
The above is only the detailed description of the invention of the application, it is noted that for the ordinary skill people of the art
For Yuan, on the premise of without departing from the application principle, it is also possible to make some improvements and modifications, these improvements and modifications also should
It is considered as the protection domain of the application.
Claims (7)
1. a load identification method based on improvement probabilistic neural network, it is characterised in that including:
Use binary system to encode the loadtype in power utilization network, set up the code database of loadtype;
Gather each loadtype electrical quantity in starting moment to a period of time;
Set up the nonlinear mapping relation between the electrical quantity of each loadtype and corresponding coding, create probabilistic neural network;
Being trained described probabilistic neural network, it is thus achieved that the error function of network, described error function is expressed as:In formula,For the desired output corresponding with training set sample, ytK () is probability
Actual output after neural metwork training, q is training sample set number, and S is loadtype number;
Using described error function as the fitness function of particle cluster algorithm, use particle cluster algorithm to described probabilistic neural network
Smoothing factor be optimized, obtain the optimal smoothing factor of described probabilistic neural network;
Described probabilistic neural network is updated, it is thus achieved that the probabilistic neural network of improvement according to the acquired optimal smoothing factor;
Based on the probabilistic neural network improved, the loadtype in power utilization network is identified.
The most according to claim 1 based on the load identification method improving probabilistic neural network, it is characterised in that to use grain
The smoothing factor of described probabilistic neural network is optimized and comprises the following steps by swarm optimization:
(S1) span of smoothing factor is set, can random initializtion a group particle { σ in solution space1,σ2,LσN, set
Maximum iteration time also sets current iteration number of times k=1;
(S2) using described error function as the fitness function of particle cluster algorithm, calculate particle fitness value, and find with this
Individual extreme value and colony's extreme value;Wherein, in the search volume of a D dimension, the smoothing factor corresponding to N number of pattern class forms
Population σ '=(σ1,σ2,LσN), i-th particle is expressed as the vectorial σ of a D dimensioni=[σi1,σi2,LσiD]T, represent i-th grain
Son position in D dimension search volume, the speed of i-th particle is Vi=[Vi1,Vi2,L,ViD]T, its individual extreme value is Pi=
[Pi1,Pi2,L,PiD]T, the global extremum of population is Pg=[Pg1,Pg2,L,PgD]T;Wherein, N, D are respectively integer;
(S3) particle rapidity and location updating, recalculates particle fitness value, and individual extreme value and colony's extreme value update, under obtaining
For population;More new formula is as follows:
In formula, ω is inertia weight;D=1,2 ..., D;I=1,2 ..., N;K is current iteration number of times;VidSpeed for particle;
c1And c2For the constant of non-negative, referred to as acceleration factor;r1And r2For being distributed in the random number between [0,1];
(S4) current iteration number of times k=k+1 is set;
(S5) check whether and meet termination condition: if reaching maximum iteration time or Jm=0, then stop;Otherwise return step
(S3)。
The most according to claim 3 based on the load identification method improving probabilistic neural network, it is characterised in that set
Maximum iteration time be 200~1000 times.
The most according to claim 4 based on the load identification method improving probabilistic neural network, it is characterised in that set
Maximum iteration time be 500 times.
5. according to the arbitrary described load identification method based on improvement probabilistic neural network of claim 1-4, it is characterised in that
Described loadtype include disposable load, disabling load, the combination of multiple disposable load, multiple disabling load combination and can
With load and the combination of disabling load.
6. according to the arbitrary described load identification method based on improvement probabilistic neural network of claim 1-4, it is characterised in that
Gather each loadtype electrical quantity in starting moment to 15 second time.
The most according to claim 6 based on the load identification method improving probabilistic neural network, it is characterised in that described electricity
Parameter includes voltage, electric current, power, power factor and harmonic wave.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610683840.1A CN106327357B (en) | 2016-08-17 | 2016-08-17 | A kind of load identification method based on improvement probabilistic neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610683840.1A CN106327357B (en) | 2016-08-17 | 2016-08-17 | A kind of load identification method based on improvement probabilistic neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106327357A true CN106327357A (en) | 2017-01-11 |
CN106327357B CN106327357B (en) | 2019-10-01 |
Family
ID=57743237
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610683840.1A Active CN106327357B (en) | 2016-08-17 | 2016-08-17 | A kind of load identification method based on improvement probabilistic neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106327357B (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107240097A (en) * | 2017-06-27 | 2017-10-10 | 长春工业大学 | Lung neoplasm image processing method based on MKL SVM PSO algorithms |
CN107240026A (en) * | 2017-04-20 | 2017-10-10 | 西安电子科技大学 | A kind of community discovery method suitable for noise network |
CN107643088A (en) * | 2017-08-10 | 2018-01-30 | 中国科学院深圳先进技术研究院 | Navigation of Pilotless Aircraft method, apparatus, unmanned plane and storage medium |
CN109036385A (en) * | 2018-10-19 | 2018-12-18 | 北京旋极信息技术股份有限公司 | A kind of voice instruction recognition method, device and computer storage medium |
CN110909802A (en) * | 2019-11-26 | 2020-03-24 | 西安邮电大学 | Improved PSO (particle swarm optimization) based fault classification method for optimizing PNN (portable network) smoothing factor |
CN112396098A (en) * | 2020-11-12 | 2021-02-23 | 天津天大求实电力新技术股份有限公司 | Non-embedded apartment electrical appliance load identification and analysis method, system and application |
CN112529683A (en) * | 2020-11-27 | 2021-03-19 | 百维金科(上海)信息科技有限公司 | Method and system for evaluating credit risk of customer based on CS-PNN |
CN113990017A (en) * | 2021-11-21 | 2022-01-28 | 特斯联科技集团有限公司 | Forest and grassland fire early warning system and method based on PNN neural network |
CN115952447A (en) * | 2023-03-14 | 2023-04-11 | 广东云山能源科技有限公司 | Intelligent identification system and method for load types of electric appliances |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040249775A1 (en) * | 2003-05-13 | 2004-12-09 | Dingguo Chen | Very short term load prediction in an energy management system |
CN102570392A (en) * | 2012-01-17 | 2012-07-11 | 上海电力学院 | Method for identifying exciting inrush current of transformer based on improved probability neural network |
CN103336992A (en) * | 2013-06-27 | 2013-10-02 | 电子科技大学 | FNN learning algorithm |
CN104050517A (en) * | 2014-06-27 | 2014-09-17 | 哈尔滨工业大学 | Photovoltaic power generation forecasting method based on GRNN |
CN104852399A (en) * | 2015-05-18 | 2015-08-19 | 国家电网公司 | Method of dynamically optimizing energy storage capacity of optical storage micro-grid system |
-
2016
- 2016-08-17 CN CN201610683840.1A patent/CN106327357B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040249775A1 (en) * | 2003-05-13 | 2004-12-09 | Dingguo Chen | Very short term load prediction in an energy management system |
CN102570392A (en) * | 2012-01-17 | 2012-07-11 | 上海电力学院 | Method for identifying exciting inrush current of transformer based on improved probability neural network |
CN103336992A (en) * | 2013-06-27 | 2013-10-02 | 电子科技大学 | FNN learning algorithm |
CN104050517A (en) * | 2014-06-27 | 2014-09-17 | 哈尔滨工业大学 | Photovoltaic power generation forecasting method based on GRNN |
CN104852399A (en) * | 2015-05-18 | 2015-08-19 | 国家电网公司 | Method of dynamically optimizing energy storage capacity of optical storage micro-grid system |
Non-Patent Citations (1)
Title |
---|
胡航海等: "以直供用电量预测经营区域及全国用电量的方法研究", 《华东电力》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107240026A (en) * | 2017-04-20 | 2017-10-10 | 西安电子科技大学 | A kind of community discovery method suitable for noise network |
CN107240026B (en) * | 2017-04-20 | 2021-01-29 | 西安电子科技大学 | Community discovery method suitable for noise network |
CN107240097A (en) * | 2017-06-27 | 2017-10-10 | 长春工业大学 | Lung neoplasm image processing method based on MKL SVM PSO algorithms |
CN107643088A (en) * | 2017-08-10 | 2018-01-30 | 中国科学院深圳先进技术研究院 | Navigation of Pilotless Aircraft method, apparatus, unmanned plane and storage medium |
CN109036385A (en) * | 2018-10-19 | 2018-12-18 | 北京旋极信息技术股份有限公司 | A kind of voice instruction recognition method, device and computer storage medium |
CN110909802A (en) * | 2019-11-26 | 2020-03-24 | 西安邮电大学 | Improved PSO (particle swarm optimization) based fault classification method for optimizing PNN (portable network) smoothing factor |
CN112396098A (en) * | 2020-11-12 | 2021-02-23 | 天津天大求实电力新技术股份有限公司 | Non-embedded apartment electrical appliance load identification and analysis method, system and application |
CN112529683A (en) * | 2020-11-27 | 2021-03-19 | 百维金科(上海)信息科技有限公司 | Method and system for evaluating credit risk of customer based on CS-PNN |
CN113990017A (en) * | 2021-11-21 | 2022-01-28 | 特斯联科技集团有限公司 | Forest and grassland fire early warning system and method based on PNN neural network |
CN113990017B (en) * | 2021-11-21 | 2022-04-29 | 特斯联科技集团有限公司 | Forest and grassland fire early warning system and method based on PNN neural network |
CN115952447A (en) * | 2023-03-14 | 2023-04-11 | 广东云山能源科技有限公司 | Intelligent identification system and method for load types of electric appliances |
Also Published As
Publication number | Publication date |
---|---|
CN106327357B (en) | 2019-10-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106327357A (en) | Load identification method based on improved probabilistic neural network | |
CN105930901B (en) | A kind of Diagnosis Method of Transformer Faults based on RBPNN | |
Gill et al. | Training back propagation neural networks with genetic algorithm for weather forecasting | |
CN109598386A (en) | A kind of accurate analysis method of distribution optimization and system based on deep learning | |
CN109142946A (en) | Transformer fault detection method based on ant group algorithm optimization random forest | |
CN107423839A (en) | A kind of method of the intelligent building microgrid load prediction based on deep learning | |
CN103926526A (en) | Analog circuit fault diagnosis method based on improved RBF neural network | |
CN104636801A (en) | Transmission line audible noise prediction method based on BP neural network optimization | |
CN103995237A (en) | Satellite power supply system online fault diagnosis method | |
CN106485089B (en) | The interval parameter acquisition methods of harmonic wave user's typical condition | |
CN104636985A (en) | Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network | |
CN110837915B (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
CN108563863B (en) | Energy consumption calculation and scheduling method for urban rail transit system | |
CN106779177A (en) | Multiresolution wavelet neutral net electricity demand forecasting method based on particle group optimizing | |
CN108280998A (en) | Short-time Traffic Flow Forecasting Methods based on historical data dynamic select | |
CN106803135A (en) | The Forecasting Methodology and device of a kind of photovoltaic power generation system output power | |
CN111325315A (en) | Distribution transformer power failure and power loss prediction method based on deep learning | |
CN113780684A (en) | Intelligent building user energy consumption behavior prediction method based on LSTM neural network | |
CN109829627A (en) | A kind of safe confidence appraisal procedure of Electrical Power System Dynamic based on integrated study scheme | |
CN111242161A (en) | Non-invasive non-resident user load identification method based on intelligent learning | |
Xiao et al. | Network security situation prediction method based on MEA-BP | |
Cao et al. | Currency recognition modeling research based on BP neural network improved by gene algorithm | |
CN111767675A (en) | Transformer vibration fault monitoring method and device, electronic equipment and storage medium | |
CN113688960A (en) | Grey wolf optimization GHFCM-based residential power data clustering method and device | |
Tang et al. | The research on BP neural network model based on guaranteed convergence particle swarm optimization |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |