CN109145949A - Non-intrusive electrical load monitoring and decomposition method and system based on integrated study - Google Patents
Non-intrusive electrical load monitoring and decomposition method and system based on integrated study Download PDFInfo
- Publication number
- CN109145949A CN109145949A CN201810796286.7A CN201810796286A CN109145949A CN 109145949 A CN109145949 A CN 109145949A CN 201810796286 A CN201810796286 A CN 201810796286A CN 109145949 A CN109145949 A CN 109145949A
- Authority
- CN
- China
- Prior art keywords
- integrated study
- load
- data
- machine learning
- integrated
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention discloses a kind of non-intrusive electrical load monitoring and decomposition method and system based on integrated study.Wherein, this method comprises: obtaining the voltage and current data of electric load inlet；Active power, current harmonics, power-factor angle, the Current harmonic distortion rate feature of electric load inlet are calculated, sample set is constructed；Sample set is randomly divided into training set and forecast set, integrated learner is trained using training set；After the completion of the training of integrated study device, using forecast set come classification belonging to sample in test sample collection.It has the effect of that test accuracy is high, highly reliable and more stable.
Description
Technical field
The invention belongs to electric power data excavation applications more particularly to a kind of non-intrusive electrical loads based on integrated study
Monitoring and decomposition method and system.
Background technique
Load monitoring is of great significance for the reliability of electric system, and the application of metering separate technology can be real
The electric energy management mode of existing scientific quantitative analysis.There are two types of typical implementations for the monitoring of load electricity consumption details: intrusive and non-intruding
Formula two schemes.Wherein, non-intrusive electrical load monitoring is not necessarily to enter inside load with decomposition technique, only by power load
The information such as voltage, electric current and the power of lotus inlet are measured, are analyzed, and it is real-time can to obtain different electrical equipments inside load
Power consumption proportion, thus realize electric load decompose.This method is simple, economical, reliably, is easy to promote rapidly, Neng Goubang
Help Utilities Electric Co.'s Accurate Prediction electric load, science formulate dispatching of power netwoks scheme, then improve electric system stability and can
By property.Meanwhile and family can be used to understand the service condition of electrical equipment in different periods, help user to formulate reasonable energy conservation meter
It draws, the service condition of electrical equipment is adjusted, to reduce power consumption.On the other hand, metering separate can be directed to energy consumption data
Realize the function of ageing equipment, fault pre-alarming, we production, have a very actual meaning in life.
Currently, most researchs using the basic algorithms such as k nearest neighbor, neural network, support vector machine to single load condition into
Row detection, the feature used is more single, and specific electric appliance that can only be common to part has detection effect, same for a variety of electrical appliances
The complex situations detection effect of Shi Yunhang is poor.
Summary of the invention
In order to solve the deficiencies in the prior art, the first object of the present invention is to provide a kind of non-intruding based on integrated study
The monitoring of formula electric load and decomposition method, have the effect of that test accuracy is high, highly reliable and more stable.
A kind of non-intrusive electrical load monitoring and decomposition method based on integrated study of the invention, comprising:
Obtain the voltage and current data of electric load inlet；
Active power, current harmonics, power-factor angle, the Current harmonic distortion rate feature of electric load inlet are calculated,
Construct sample set；
Sample set is randomly divided into training set and forecast set, integrated learner is trained using training set；Work as integrated study
After the completion of device training, using forecast set come classification belonging to sample in test sample collection；
Wherein, integrated study device is made of at least two machine learning model parallel connections, and each machine learning model is to pre-
It surveys collection and carries out classification belonging to sample in tentative prediction sample set, according to the accuracy of each machine learning model power corresponding to its
The sum of value is greater than preset threshold to determine the trained completion of integrated study device, otherwise continues the integrated learner of training；Wherein, respectively
Accuracy of the corresponding weight value of a machine learning model equal to each machine learning model classification results and all machine learning moulds
The ratio of the sum of cumulative obtained total accuracy of type；The preset threshold predicts classification results probability value by each machine learning
Average value determine.
Further, after the voltage and current data for obtaining electric load inlet, further includes: right in time series
Measurement missing values in the voltage and current data of electric load inlet are fitted.
The present invention by time series to the measurement missing values in the voltage and current data of electric load inlet
It is fitted, improves the continuity of data, and then data base has been established to the reliability and stability of the integrated learner of training
Plinth.
Further, during constructing sample set, further includes:
Mode segmentation is carried out to all samples in sample set by sliding window slope fit method, by total load and single negative
Lotus is divided into from opening to several transient state sections and stable state section in shut-down process, then never extracts required load mark with section
Note.
The present invention has more been bonded the practical feelings of load operation by carrying out mode segmentation to all samples in sample set
Condition, and then data basis has been established to the reliability of the integrated learner of training.
Further, by sliding window slope fit method all samples in sample set are carried out with the process packet of mode segmentation
It includes:
The threshold value for setting the variation of a power data, by the maximum value calculation slope threshold value of slope fit value, readout power
Data initialization detection data window；
According to the size relation of slope fit value and slope threshold value, whether judging the generation of load event, load thing is marked
The generation point and end point of part:
By the terminal for the previous stable state section that the previous data point markers of starting point are the transient state, or label transient state section
Starting point, by the latter data point markers of transient state section be stable state section starting point.
Further, parallel combination decision tree, k neighbour, artificial neural network and support vector machines in integrated study device
Model, and voted using Nearest Neighbor with Weighted Voting method tentative prediction result, to obtain final result, and introduce naive Bayesian
Model promotes the noiseproof feature of integrated study device.
The second object of the present invention be to provide it is a kind of based on integrated study non-intrusive electrical load monitoring and resolving system
System, has the effect of that test accuracy is high, highly reliable and more stable.
A kind of non-intrusive electrical load monitoring and decomposing system based on integrated study of the invention, comprising:
Electric power detection device, is configured as: obtaining the voltage and current data of electric load inlet；
Processor comprising:
Characteristic extracting module is configured as: calculate the active power of electric load inlet, current harmonics, power because
Number angle, Current harmonic distortion rate feature, construct sample set；
Integrated study device training module, is configured as: sample set being randomly divided into training set and forecast set, utilizes training
Collection is to train integrated learner；
Wherein, integrated study device is made of at least two machine learning model parallel connections, and each machine learning model is to pre-
It surveys collection and carries out classification belonging to sample in tentative prediction sample set, according to the accuracy of each machine learning model power corresponding to its
The sum of value is greater than preset threshold to determine the trained completion of integrated study device, otherwise continues the integrated learner of training；Wherein, respectively
Accuracy of the corresponding weight value of a machine learning model equal to each machine learning model classification results and all machine learning moulds
The ratio of the sum of cumulative obtained total accuracy of type；The preset threshold predicts classification results probability value by each machine learning
Average value determine；
Test module is configured as: after the completion of the training of integrated study device, using forecast set come sample in test sample collection
Classification belonging to this.
Further, the processor further include:
Data fitting module, is configured as: after the voltage and current data for obtaining electric load inlet, in the time
The measurement missing values in the voltage and current data of electric load inlet are fitted in sequence.
Further, the processor further includes mode segmentation module, is configured as:
Mode segmentation is carried out to all samples in sample set by sliding window slope fit method, by total load and single negative
Lotus is divided into from opening to several transient state sections and stable state section in shut-down process, then never extracts required load mark with section
Note.
Further, in mode segmentation module, by sliding window slope fit method to all samples in sample set
The process of this progress mode segmentation includes:
The threshold value for setting the variation of a power data, by the maximum value calculation slope threshold value of slope fit value, readout power
Data initialization detection data window；
According to the size relation of slope fit value and slope threshold value, whether judging the generation of load event, load thing is marked
The generation point and end point of part:
By the terminal for the previous stable state section that the previous data point markers of starting point are the transient state, or label transient state section
Starting point, by the latter data point markers of transient state section be stable state section starting point.
Further, parallel combination decision tree, k neighbour, artificial neural network and support vector machines in integrated study device
Model, and voted using Nearest Neighbor with Weighted Voting method tentative prediction result, to obtain final result, and introduce naive Bayesian
Model promotes the noiseproof feature of integrated study device.
Compared with prior art, the beneficial effects of the present invention are:
(1) the present invention is based on integrated study device, that non-intrusive electrical load detection is passed through with decomposing system is certain
Statistical law automatically identify the operating status of electrical equipment, and total load is decomposed into each electrical equipment, has test quasi-
High, the highly reliable and more stable beneficial effect of exactness.
(2) the present invention is based on integrated study device, make only can will use be known by obtaining electric load inlet total load
Electrical appliance state is divided into each mode class of electrical appliance, and then determines accurate electrical equipment state, have test accuracy it is high, can
The strong and more stable beneficial effect by property.
Detailed description of the invention
The accompanying drawings constituting a part of this application is used to provide further understanding of the present application, and the application's shows
Meaning property embodiment and its explanation are not constituted an undue limitation on the present application for explaining the application.
Fig. 1 is a kind of non-intrusive electrical load monitoring and decomposition method flow chart based on integrated study of the invention；
Fig. 2 is non-intrusion type system monitoring scheme schematic diagram；
Fig. 3 is electrical equipment power diagram；
Fig. 4 is electrical appliance mode segmentation flow chart；
Fig. 5 (a) is the first state segmentation figure by taking oaks fan as an example；
Fig. 5 (b) is the second state segmentation figure by taking oaks fan as an example；
Fig. 5 (c) is the third state segmentation figure by taking oaks fan as an example；
Fig. 5 (d) is the 4th state segmentation figure by taking oaks fan as an example；
Fig. 5 (e) is the 5th state segmentation figure by taking oaks fan as an example；
Fig. 6 is artificial neural network module diagram；
Fig. 7 is a kind of non-intrusive electrical load monitoring and decomposing system structural representation based on integrated study of the invention
Figure.
Specific embodiment
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
Fig. 1 is a kind of non-intrusive electrical load monitoring and decomposition method flow chart based on integrated study of the invention.
As shown in Figure 1, a kind of non-intrusive electrical load monitoring and decomposition method based on integrated study of the invention, packet
It includes:
Step 1: obtaining the voltage and current data of electric load inlet.
After the voltage and current data for obtaining electric load inlet, further includes: to electric load in time series
Measurement missing values in the voltage and current data of inlet are fitted.
After the voltage and current data for obtaining electric load inlet, further includes: to the electric power after screening and fitting
Electric current, the voltage data measured value of load inlet format.
Specifically, data are pre-processed first, total data is checked, calculate its missing values ratio, determined
The range of missing values.According to missing ratio and field importance, different processing strategies is taken.
The feature high for importance, miss rate is low is filled by experience or professional knowledge estimation；For importance
Feature high, miss rate is high calculates completion using other more complicated models.
Data are all based on time series, and contain a large amount of implicit missing values.First in seconds by time series into
After row polishing, according to different operational circumstances, accordingly and harmonic data is in the time by the device data of different electrical appliances, cycle
It is divided into transient state and steady state data in sequence, while according to its distinct device type, Lagrange's interpolation is taken to carry out at polishing
Reason.
The case where partially arranging the problem of not being aligned existing for data to importing, and having more column, carries out artificial treatment.
Step 2: calculating active power, current harmonics, power-factor angle, the Current harmonic distortion rate of electric load inlet
Feature constructs sample set.
This example non-invasive system monitoring scheme is by taking Fig. 2 as an example:
Data totally 6264 of this example use, every record contain having time, electric current, voltage, active power, idle function
Rate, power factor, total active power, total power factor, and in one cycle on (0.02 second) 128 time points of acquisition
Electric current and voltage cycle data, the harmonic data of 51 primary currents and voltage.These data are from including fan, micro-wave oven, heat
5 kinds of electrical equipments such as kettle, each equipment and its running parameter are as shown in table 1.Fig. 3 illustrates pair of each electrical appliance power waveform
Than.
Each equipment of table 1 and its running parameter
Step 3: sample set being randomly divided into training set and forecast set, integrated learner is trained using training set；Work as collection
After the completion of learner training, using forecast set come classification belonging to sample in test sample collection；
Wherein, integrated study device is made of at least two machine learning model parallel connections, and each machine learning model is to pre-
It surveys collection and carries out classification belonging to sample in tentative prediction sample set, according to the accuracy of each machine learning model power corresponding to its
The sum of value is greater than preset threshold to determine the trained completion of integrated study device, otherwise continues the integrated learner of training；Wherein, respectively
Accuracy of the corresponding weight value of a machine learning model equal to each machine learning model classification results and all machine learning moulds
The ratio of the sum of cumulative obtained total accuracy of type；The preset threshold predicts classification results probability value by each machine learning
Average value determine.
During constructing sample set, further includes:
Mode segmentation is carried out to all samples in sample set by sliding window slope fit method, by total load and single negative
Lotus is divided into from opening to several transient state sections and stable state section in shut-down process, then never extracts required load mark with section
Note.
That is the form of " ..., TSLn, SSLn, TSLn+1, SSLn+1 ... ", i.e. transient state, stable state, transient state ..., then from difference
The load marking needed for section extracts, detailed process are shown in that Fig. 4, Fig. 5 (a)-Fig. 5 (e) are microwave power under different conditions after segmentation
Waveform.
The present invention has more been bonded the practical feelings of load operation by carrying out mode segmentation to all samples in sample set
Condition, and then data basis has been established to the reliability of the integrated learner of training.
Include: by the process that sliding window slope fit method carries out mode segmentation to all samples in sample set
The threshold value for setting the variation of a power data, by the maximum value calculation slope threshold value of slope fit value, readout power
Data initialization detection data window；
According to the size relation of slope fit value and slope threshold value, whether judging the generation of load event, load thing is marked
The generation point and end point of part:
By the terminal for the previous stable state section that the previous data point markers of starting point are the transient state, or label transient state section
Starting point, by the latter data point markers of transient state section be stable state section starting point.
Specifically, in mode cutting procedure, pass through the threshold value Δ P of setting one power data variationth, by slope fit
The maximum value calculation slope threshold value K of valueth, readout power data initialization detection data window.In formula, WdFor data window length.
According to the slope fit value K of data in linear least-squares theoretical calculation data windowi:
In above formula, PiFor the starting point of current detection data window, i ∈ 1,2,3 ... }, { Pj| j=i, i+1 ..., i+Wd-
It 1 } is power data sequence contained by current detection data window；Judge that the size of slope fit value and slope threshold value is closed according to the following formula
System, whether judging the generation of load event:
The decision condition that load event occurs are as follows:
|Ki-1|≤Kth
|Ki| > Kth (4)
The decision condition that load event terminates are as follows:
|Ki-1| > Kth
|Ki|≤Kth (5)
In above formula, j=i-1 or i+Wd- 1, power number strong point P at this timeiThe as end point of load event；If meeting load
Event occurs or the decision condition that terminates of load event, then marks the generation point (transient state starting point) of load event, while by starting point
Previous data point markers be the transient state previous stable state section terminal, or label load event end point (transient state
Terminal), while being the starting point of stable state section by the latter data point markers of transient state section.
Power swing measurement is calculated, is measured according to gained power swing, slope threshold value is updated by following formula.
In formula,It is the inverse function of standard normal distribution function, σpFor power swing measurement.Real event is missed general
Rate is αk,It is slope value corresponding with the minimum power change absolute value reliably detected is required.
Wherein, parallel combination decision tree, k neighbour, artificial neural network and supporting vector machine model in integrated study device,
And voted using Nearest Neighbor with Weighted Voting method tentative prediction result, to obtain final result, and introduce model-naive Bayesian
To promote the noiseproof feature of integrated study device.
Integrated study device includes decision tree module, k neighbour module, artificial neural network module, support vector machines multilayer film
Block for obtaining each Weak Classifier to the prediction classification results of each unknown state, and finally uses Nearest Neighbor with Weighted Voting module, by right
The result of each classifier is unitized, calculates the accuracy of each classifier result, and the accuracy for being added each classifier obtains always correctly
The value of rate obtains the weight of single classifier by classifier accuracy compared with total accuracy, according to class selected by each classifier
The average value of probability of outcome value determines optimal threshold, and carries out final vote processing to classifier.
Classifier in the present embodiment refers to machine learning model.
Artificial neural network module can extract hiding information relevant to mode, to create the expression of itself.Learn rank
Section includes the renewal process of setting network parameter, such as hides the number of plies, concealed nodes number, Model Weight, and the signal of input passes through
Network is successively propagated forward, and back-propagation algorithm undated parameter is passed through.
The neuron for designing neural network input layer is 32, constitutes input matrix P by input electrical appliance active powerL=
[PL (1)..., PL (32)].Three-layer neural network can approach nonlinear-load decomposition model substantially, therefore, the multilayer designed herein
Neural network is three layers.In general, neural network is made of input layer, hidden layer and output layer.
Decision-making module uses Nearest Neighbor with Weighted Voting module, unitized by the result to each classifier, calculates each classifier result
Accuracy, the accuracy for being added each classifier obtains the value of total accuracy, by classifier accuracy compared with total accuracy,
Obtain the weight of single classifier, optimal threshold determined according to the average value of class probability of outcome value selected by each classifier, and to point
Class device carries out final vote processing.As a result it is shown below:
In formula, λiFor the weight of single classifier, i is classifier number, and P (i) is the accuracy of single classifier, and θ is
Optimal threshold.
Decision tree module, use information ratio of profit increase select optimal dividing attribute, increase the processing capacity to connection attribute.
It is assumed that currently there is a sample set to be denoted as D, ratio shared by kth class sample is P in sample setk(k=1,2 ... ..., C),
C is class number total in sample then sample set comentropy is defined as:
It is assumed that being divided now according to attribute A to sample set, if there are V possible values in attribute A, can generate
V branch node, wherein V (v=1,2,3 ... ..., n) a branch node contains all in sample set takes on attribute A
Value is AvSample, be denoted as Dv.Then there is information gain:
And then obtain the information gain-ratio of attribute A:
IV (A) indicates the eigenvalue of attribute A in formula:
By calculating the ratio of profit increase of different attribute, the division category for selecting the maximum attribute of ratio of profit increase to divide as this
Property, then adopt the ratio of profit increase for calculating other attributes in a like fashion.Gradually divided, until all loads are distinguished or
All samples of person value on all properties is identical, until can not divide.
In the module, select the feature active power of said extracted, current harmonics, power-factor angle, current harmonics abnormal
Variability constructs decision tree as input.
Each point is assigned to by k mean cluster module using power in the state of 32 kinds of states and harmonic wave as 32 mass centers
Nearest mass center forms K cluster, and the mass center of each cluster is recalculated with Euclidean distance, until cluster does not change or reaches
Maximum number of iterations minimizes mean square deviation in the hope of optimal solution.
Support vector machines multilayer module recognizes electrical appliance state using the feature in feature set as the input of model,
Acquire classification results.
In this experiment, 4 classifiers are constructed, using different characteristic as inputting, decision is carried out using weighted voting algorithm,
Training is done to base classifier.Experimental result such as table 2:
The experimental result of 2 embodiment 1 of table
As can be seen from the results, the non-intrusion type load recognizer proposed by the present invention based on integrated study is negative to non-intrusion type
Lotus identification has relatively good beneficial effect, there is great value in production.
Embodiment 2:
In order to verify the reliability of model, we are with oaks fan, the micro-wave oven of beauty, nine positive hot-water bottles, ThinkPad
5 kinds of notebook, incandescent lamp electrical equipments carry out research and analysis, and totally 25The vector power of kind transient state situation, 32 kinds of situations is made
For the input of neural network, the neuron for designing neural network input layer is 32, and input matrix P is made of input variableL=
[PL (1)..., PL (32)]。
Three-layer neural network can approach nonlinear-load decomposition model substantially, therefore, the multilayer nerve net designed herein
Network is three layers.In general, neural network is made of input layer, hidden layer and output layer, as shown in Figure 6.
Generally, the neuron number of hidden layer is designed as between 4 to 60.By repeatedly training, and rule of thumb really
Determining hidden layer neuron number is 40.The transfer function of hidden layer need to meet condition that everywhere can be micro-, be typically chosen logarithm
Sigmoid function.Similarly, output layer neuron transfer function also selects logarithm Sigmoid function.The multilayer of this module design
The structure of neural network is 32-50-32, i.e. input layer has 32 nodes, and hidden layer has 50 nodes, and output layer has 32 sections
Point.Final experimental result such as table 3:
The experimental result of 3 embodiment 2 of table
Integrated model known to examining is more opposite than single model to have more certain reliability and robustness, compares and makes us full
Meaning.
The present invention by time series to the measurement missing values in the voltage and current data of electric load inlet
It is fitted, improves the continuity of data, and then data base has been established to the reliability and stability of the integrated learner of training
Plinth.
Fig. 7 is a kind of non-intrusive electrical load monitoring and decomposing system structural representation based on integrated study of the invention
Figure.
As shown in fig. 7, a kind of non-intrusive electrical load monitoring and decomposing system based on integrated study of the invention, packet
It includes:
(1) electric power detection device is configured as: obtaining the voltage and current data of electric load inlet；
(2) processor comprising:
Characteristic extracting module is configured as: calculate the active power of electric load inlet, current harmonics, power because
Number angle, Current harmonic distortion rate feature, construct sample set；
Integrated study device training module, is configured as: sample set being randomly divided into training set and forecast set, utilizes training
Collection is to train integrated learner；
Wherein, integrated study device is made of at least two machine learning model parallel connections, and each machine learning model is to pre-
It surveys collection and carries out classification belonging to sample in tentative prediction sample set, according to the accuracy of each machine learning model power corresponding to its
The sum of value is greater than preset threshold to determine the trained completion of integrated study device, otherwise continues the integrated learner of training；Wherein, respectively
Accuracy of the corresponding weight value of a machine learning model equal to each machine learning model classification results and all machine learning moulds
The ratio of the sum of cumulative obtained total accuracy of type；The preset threshold predicts classification results probability value by each machine learning
Average value determine；
Test module is configured as: after the completion of the training of integrated study device, using forecast set come sample in test sample collection
Classification belonging to this.
In specific implementation, the processor further include:
Data fitting module, is configured as: after the voltage and current data for obtaining electric load inlet, in the time
The measurement missing values in the voltage and current data of electric load inlet are fitted in sequence.
In specific implementation, the processor further includes mode segmentation module, is configured as:
Mode segmentation is carried out to all samples in sample set by sliding window slope fit method, by total load and single negative
Lotus is divided into from opening to several transient state sections and stable state section in shut-down process, then never extracts required load mark with section
Note.
In specific implementation, in mode segmentation module, by sliding window slope fit method to the institute in sample set
Have sample carry out mode segmentation process include:
The threshold value for setting the variation of a power data, by the maximum value calculation slope threshold value of slope fit value, readout power
Data initialization detection data window；
According to the size relation of slope fit value and slope threshold value, whether judging the generation of load event, load thing is marked
The generation point and end point of part:
By the terminal for the previous stable state section that the previous data point markers of starting point are the transient state, or label transient state section
Starting point, by the latter data point markers of transient state section be stable state section starting point.
In specific implementation, in integrated study device parallel combination decision tree, k neighbour, artificial neural network and support to
Amount machine model, and voted using Nearest Neighbor with Weighted Voting method tentative prediction result, to obtain final result, and introduce simple shellfish
This model of leaf promotes the noiseproof feature of integrated study device.
A kind of non-intrusive electrical load monitoring and decomposing system based on integrated study of the invention, further includes: display
Unit is connected with processor, the result for output processor.
The present invention is based on integrated study device enable non-intrusive electrical load detection with decomposing system pass through it is certain
Statistical law automatically identifies the operating status of electrical equipment, and total load is decomposed each electrical equipment, has test accurate
Spend high, highly reliable and more stable beneficial effect.
The present invention is based on integrated study device, make only can will electricity consumption be known by obtaining electric load inlet total load
Device state demarcation determines accurate electrical equipment state to each mode class of electrical appliance, has test accuracy high, reliable
The strong and more stable beneficial effect of property.
It should be understood by those skilled in the art that, the embodiment of the present invention can provide as method, system or computer program
Product.Therefore, the shape of hardware embodiment, software implementation or embodiment combining software and hardware aspects can be used in the present invention
Formula.Moreover, the present invention, which can be used, can use storage in the computer that one or more wherein includes computer usable program code
The form for the computer program product implemented on medium (including but not limited to magnetic disk storage and optical memory etc.).
The present invention be referring to according to the method for the embodiment of the present invention, the process of equipment (system) and computer program product
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a computer-readable storage medium
In, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random
AccessMemory, RAM) etc..
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects model to the present invention
The limitation enclosed, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art are not
Need to make the creative labor the various modifications or changes that can be made still within protection scope of the present invention.
Claims (10)
1. a kind of non-intrusive electrical load monitoring and decomposition method based on integrated study characterized by comprising
Obtain the voltage and current data of electric load inlet；
Calculate active power, current harmonics, power-factor angle, the Current harmonic distortion rate feature of electric load inlet, building
Sample set；
Sample set is randomly divided into training set and forecast set, integrated learner is trained using training set；When integrated study device is instructed
After the completion of white silk, using forecast set come classification belonging to sample in test sample collection；
Wherein, integrated study device is made of at least two machine learning model parallel connections, and each machine learning model is to forecast set
Carry out classification belonging to sample in tentative prediction sample set, according to the accuracy of each machine learning model and its corresponding weight value it
The trained completion of integrated study device is determined with preset threshold is greater than, and otherwise continues the integrated learner of training；Wherein, Ge Geji
The corresponding weight value of device learning model is equal to the accuracy and all machine learning models of each machine learning model classification results
The ratio of the sum of cumulative obtained total accuracy；The preset threshold is flat by each machine learning prediction classification results probability value
Mean value determines.
2. a kind of non-intrusive electrical load monitoring and decomposition method based on integrated study as described in claim 1, special
Sign is, after the voltage and current data for obtaining electric load inlet, further includes: enter in time series to electric load
The measurement missing values in voltage and current data at mouthful are fitted.
3. a kind of non-intrusive electrical load monitoring and decomposition method based on integrated study as described in claim 1, special
Sign is, during constructing sample set, further includes:
Mode segmentation is carried out to all samples in sample set by sliding window slope fit method, by total load and single load point
For from open to several transient state sections and stable state section in shut-down process, then never with section extract needed for load label.
4. a kind of non-intrusive electrical load monitoring and decomposition method based on integrated study as claimed in claim 3, special
Sign is, includes: by the process that sliding window slope fit method carries out mode segmentation to all samples in sample set
The threshold value for setting the variation of a power data, by the maximum value calculation slope threshold value of slope fit value, readout power data
Initialize detection data window；
Load event is marked whether judging the generation of load event according to the size relation of slope fit value and slope threshold value
Point and end point occurs:
By of the terminal for the previous stable state section that the previous data point markers of starting point are the transient state or label transient state section
Point, while being the starting point of stable state section by the latter data point markers of transient state section.
5. a kind of non-intrusive electrical load monitoring and decomposition method based on integrated study as described in claim 1, special
Sign is, parallel combination decision tree, k neighbour, artificial neural network and supporting vector machine model in integrated study device, and uses
Nearest Neighbor with Weighted Voting method votes to tentative prediction result, to obtain final result, and introduces model-naive Bayesian to be promoted
The noiseproof feature of integrated study device.
6. a kind of non-intrusive electrical load monitoring and decomposing system based on integrated study characterized by comprising
Electric power detection device, is configured as: obtaining the voltage and current data of electric load inlet；
Processor comprising:
Characteristic extracting module is configured as: calculate the active power of electric load inlet, current harmonics, power-factor angle,
Current harmonic distortion rate feature constructs sample set；
Integrated study device training module, is configured as: sample set is randomly divided into training set and forecast set, using training set come
The integrated learner of training；
Wherein, integrated study device is made of at least two machine learning model parallel connections, and each machine learning model is to forecast set
Carry out classification belonging to sample in tentative prediction sample set, according to the accuracy of each machine learning model and its corresponding weight value it
The trained completion of integrated study device is determined with preset threshold is greater than, and otherwise continues the integrated learner of training；Wherein, Ge Geji
The corresponding weight value of device learning model is equal to the accuracy and all machine learning models of each machine learning model classification results
The ratio of the sum of cumulative obtained total accuracy；The preset threshold is flat by each machine learning prediction classification results probability value
Mean value determines；
Test module is configured as: after the completion of the training of integrated study device, using forecast set come sample institute in test sample collection
The classification of category.
7. a kind of non-intrusive electrical load monitoring and decomposing system based on integrated study as claimed in claim 6, special
Sign is, the processor further include:
Data fitting module, is configured as: after the voltage and current data for obtaining electric load inlet, in time series
On the measurement missing values in the voltage and current data of electric load inlet are fitted.
8. a kind of non-intrusive electrical load monitoring and decomposing system based on integrated study as claimed in claim 6, special
Sign is that the processor further includes mode segmentation module, is configured as:
Mode segmentation is carried out to all samples in sample set by sliding window slope fit method, by total load and single load point
For from open to several transient state sections and stable state section in shut-down process, then never with section extract needed for load label.
9. a kind of non-intrusive electrical load monitoring and decomposing system based on integrated study as claimed in claim 8, special
Sign is, in mode segmentation module, carries out mode to all samples in sample set by sliding window slope fit method
The process of segmentation includes:
The threshold value for setting the variation of a power data, by the maximum value calculation slope threshold value of slope fit value, readout power data
Initialize detection data window；
Load event is marked whether judging the generation of load event according to the size relation of slope fit value and slope threshold value
Point and end point occurs:
By the terminal for the previous stable state section that the previous data point markers of starting point are the transient state, or of label transient state section
The latter data point markers of transient state section are the starting point of stable state section by point.
10. a kind of non-intrusive electrical load monitoring and decomposing system based on integrated study as claimed in claim 6, special
Sign is, parallel combination decision tree, k neighbour, artificial neural network and supporting vector machine model in integrated study device, and uses
Nearest Neighbor with Weighted Voting method votes to tentative prediction result, to obtain final result, and introduces model-naive Bayesian to be promoted
The noiseproof feature of integrated study device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810796286.7A CN109145949A (en) | 2018-07-19 | 2018-07-19 | Non-intrusive electrical load monitoring and decomposition method and system based on integrated study |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810796286.7A CN109145949A (en) | 2018-07-19 | 2018-07-19 | Non-intrusive electrical load monitoring and decomposition method and system based on integrated study |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109145949A true CN109145949A (en) | 2019-01-04 |
Family
ID=64800975
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810796286.7A Pending CN109145949A (en) | 2018-07-19 | 2018-07-19 | Non-intrusive electrical load monitoring and decomposition method and system based on integrated study |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109145949A (en) |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109387712A (en) * | 2018-10-09 | 2019-02-26 | 厦门理工学院 | Non-intrusion type cutting load testing and decomposition method based on state matrix decision tree |
| CN110445653A (en) * | 2019-08-12 | 2019-11-12 | 灵长智能科技(杭州)有限公司 | Network state prediction technique, device, equipment and medium |
| CN110458195A (en) * | 2019-07-10 | 2019-11-15 | 河海大学 | A kind of abnormal data discriminating method based on the fusion of more criterions |
| CN110488112A (en) * | 2019-05-23 | 2019-11-22 | 杭州海兴电力科技股份有限公司 | Classification metering method non-intrusion type load identification and its realized based on recognition result |
| CN110514889A (en) * | 2019-07-19 | 2019-11-29 | 浙江万胜智能科技股份有限公司 | A kind of method and system of non-intrusion type household electricity remained capacity |
| CN110909795A (en) * | 2019-11-22 | 2020-03-24 | 国网湖南省电力有限公司 | Method for judging transient voltage stability of power grid |
| CN111009893A (en) * | 2019-11-08 | 2020-04-14 | 广西电网有限责任公司电力科学研究院 | Household user short-term load prediction method based on load decomposition technology |
| CN111092486A (en) * | 2019-11-20 | 2020-05-01 | 内蒙古电力(集团)有限责任公司包头供电局 | Non-invasive load acquisition system and method and storage medium |
| CN111242161A (en) * | 2019-12-27 | 2020-06-05 | 国网山西省电力公司大同供电公司 | Non-invasive non-resident user load identification method based on intelligent learning |
| CN111242391A (en) * | 2020-03-06 | 2020-06-05 | 云南电网有限责任公司电力科学研究院 | Machine learning model training method and system for power load identification |
| CN111382789A (en) * | 2020-03-06 | 2020-07-07 | 云南电网有限责任公司电力科学研究院 | Power load identification method and system based on machine learning |
| CN111428755A (en) * | 2020-02-29 | 2020-07-17 | 国网(苏州)城市能源研究院有限责任公司 | Non-invasive load monitoring method |
| CN111553444A (en) * | 2020-05-15 | 2020-08-18 | 东南大学 | Load identification method based on non-invasive load terminal data |
| CN111582567A (en) * | 2020-04-28 | 2020-08-25 | 昆明理工大学 | Wind power probability prediction method based on hierarchical integration |
| CN111665390A (en) * | 2020-06-15 | 2020-09-15 | 威胜集团有限公司 | Non-invasive load detection method, terminal device and readable storage medium |
| CN111898694A (en) * | 2020-08-07 | 2020-11-06 | 广东电网有限责任公司计量中心 | Non-invasive load identification method and device based on random tree classification |
| CN113033821A (en) * | 2021-03-29 | 2021-06-25 | 上海思创电器设备有限公司 | Load identification machine learning system applied to algorithm core unit |
| CN113050486A (en) * | 2021-03-12 | 2021-06-29 | 南京工程学院 | Electric power system edge calculation and data distribution device based on industrial personal computer |
| CN113094931A (en) * | 2021-04-29 | 2021-07-09 | 杭州电力设备制造有限公司 | Non-invasive load decomposition method, device and equipment |
| CN113219272A (en) * | 2021-04-06 | 2021-08-06 | 国网江苏省电力有限公司营销服务中心 | Method and system for predicting household appliance aging based on non-invasive monitoring |
| CN113285442A (en) * | 2021-04-21 | 2021-08-20 | 广东工业大学 | Single electric equipment identification method and device based on non-invasive load characteristics |
| CN113378102A (en) * | 2021-05-31 | 2021-09-10 | 国网上海市电力公司 | Data missing preprocessing method, medium and application for short-term load prediction |
| CN113723495A (en) * | 2021-08-25 | 2021-11-30 | 广西大学 | Electromagnetic equipment multi-parameter non-invasive identification method, system, equipment and storage medium based on integrated algorithm |
| CN113899944A (en) * | 2021-09-30 | 2022-01-07 | 广东电网有限责任公司 | Detection method and device for power load switching point |
| CN117081246A (en) * | 2023-08-16 | 2023-11-17 | 北京市计量检测科学研究院 | Indoor electric bicycle identification system that charges and computer equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103716204A (en) * | 2013-12-20 | 2014-04-09 | 中国科学院信息工程研究所 | Abnormal intrusion detection ensemble learning method and apparatus based on Wiener process |
| CN104483575A (en) * | 2014-12-22 | 2015-04-01 | 天津求实智源科技有限公司 | Self-adaptive load event detection method for noninvasive power monitoring |
| CN104573013A (en) * | 2015-01-09 | 2015-04-29 | 上海大学 | Category weight combined integrated learning classifying method |
| CN107122790A (en) * | 2017-03-15 | 2017-09-01 | 华北电力大学 | Non-intrusion type load recognizer based on hybrid neural networks and integrated study |
-
2018
- 2018-07-19 CN CN201810796286.7A patent/CN109145949A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103716204A (en) * | 2013-12-20 | 2014-04-09 | 中国科学院信息工程研究所 | Abnormal intrusion detection ensemble learning method and apparatus based on Wiener process |
| CN104483575A (en) * | 2014-12-22 | 2015-04-01 | 天津求实智源科技有限公司 | Self-adaptive load event detection method for noninvasive power monitoring |
| CN104573013A (en) * | 2015-01-09 | 2015-04-29 | 上海大学 | Category weight combined integrated learning classifying method |
| CN107122790A (en) * | 2017-03-15 | 2017-09-01 | 华北电力大学 | Non-intrusion type load recognizer based on hybrid neural networks and integrated study |
Non-Patent Citations (2)
| Title |
|---|
| 文武: "《基于混合神经网络和集成学习的非侵入式负荷识别算法》", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 李佳 等: "《基于多分类器加权投票法的越南语组合歧义消歧》", 《计算机科学》 * |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109387712A (en) * | 2018-10-09 | 2019-02-26 | 厦门理工学院 | Non-intrusion type cutting load testing and decomposition method based on state matrix decision tree |
| CN110488112A (en) * | 2019-05-23 | 2019-11-22 | 杭州海兴电力科技股份有限公司 | Classification metering method non-intrusion type load identification and its realized based on recognition result |
| CN110458195A (en) * | 2019-07-10 | 2019-11-15 | 河海大学 | A kind of abnormal data discriminating method based on the fusion of more criterions |
| CN110514889A (en) * | 2019-07-19 | 2019-11-29 | 浙江万胜智能科技股份有限公司 | A kind of method and system of non-intrusion type household electricity remained capacity |
| CN110445653A (en) * | 2019-08-12 | 2019-11-12 | 灵长智能科技(杭州)有限公司 | Network state prediction technique, device, equipment and medium |
| CN111009893A (en) * | 2019-11-08 | 2020-04-14 | 广西电网有限责任公司电力科学研究院 | Household user short-term load prediction method based on load decomposition technology |
| CN111009893B (en) * | 2019-11-08 | 2022-07-01 | 广西电网有限责任公司电力科学研究院 | Household user short-term load prediction method based on load decomposition technology |
| CN111092486A (en) * | 2019-11-20 | 2020-05-01 | 内蒙古电力(集团)有限责任公司包头供电局 | Non-invasive load acquisition system and method and storage medium |
| CN110909795A (en) * | 2019-11-22 | 2020-03-24 | 国网湖南省电力有限公司 | Method for judging transient voltage stability of power grid |
| CN111242161A (en) * | 2019-12-27 | 2020-06-05 | 国网山西省电力公司大同供电公司 | Non-invasive non-resident user load identification method based on intelligent learning |
| CN111242161B (en) * | 2019-12-27 | 2023-08-25 | 国网山西省电力公司大同供电公司 | Non-invasive non-resident user load identification method based on intelligent learning |
| CN111428755A (en) * | 2020-02-29 | 2020-07-17 | 国网(苏州)城市能源研究院有限责任公司 | Non-invasive load monitoring method |
| CN111428755B (en) * | 2020-02-29 | 2023-09-08 | 国网(苏州)城市能源研究院有限责任公司 | Non-invasive load monitoring method |
| CN111382789B (en) * | 2020-03-06 | 2023-11-14 | 云南电网有限责任公司电力科学研究院 | Power load identification method and system based on machine learning |
| CN111382789A (en) * | 2020-03-06 | 2020-07-07 | 云南电网有限责任公司电力科学研究院 | Power load identification method and system based on machine learning |
| CN111242391A (en) * | 2020-03-06 | 2020-06-05 | 云南电网有限责任公司电力科学研究院 | Machine learning model training method and system for power load identification |
| CN111582567A (en) * | 2020-04-28 | 2020-08-25 | 昆明理工大学 | Wind power probability prediction method based on hierarchical integration |
| CN111582567B (en) * | 2020-04-28 | 2022-07-01 | 昆明理工大学 | Wind power probability prediction method based on hierarchical integration |
| CN111553444A (en) * | 2020-05-15 | 2020-08-18 | 东南大学 | Load identification method based on non-invasive load terminal data |
| CN111665390A (en) * | 2020-06-15 | 2020-09-15 | 威胜集团有限公司 | Non-invasive load detection method, terminal device and readable storage medium |
| CN111898694A (en) * | 2020-08-07 | 2020-11-06 | 广东电网有限责任公司计量中心 | Non-invasive load identification method and device based on random tree classification |
| CN111898694B (en) * | 2020-08-07 | 2021-09-17 | 广东电网有限责任公司计量中心 | Non-invasive load identification method and device based on random tree classification |
| CN113050486A (en) * | 2021-03-12 | 2021-06-29 | 南京工程学院 | Electric power system edge calculation and data distribution device based on industrial personal computer |
| CN113050486B (en) * | 2021-03-12 | 2022-06-17 | 南京工程学院 | Electric power system edge calculation and data distribution device based on industrial personal computer |
| CN113033821A (en) * | 2021-03-29 | 2021-06-25 | 上海思创电器设备有限公司 | Load identification machine learning system applied to algorithm core unit |
| CN113219272A (en) * | 2021-04-06 | 2021-08-06 | 国网江苏省电力有限公司营销服务中心 | Method and system for predicting household appliance aging based on non-invasive monitoring |
| CN113285442A (en) * | 2021-04-21 | 2021-08-20 | 广东工业大学 | Single electric equipment identification method and device based on non-invasive load characteristics |
| CN113094931B (en) * | 2021-04-29 | 2022-04-22 | 杭州电力设备制造有限公司 | Non-invasive load decomposition method, device and equipment |
| CN113094931A (en) * | 2021-04-29 | 2021-07-09 | 杭州电力设备制造有限公司 | Non-invasive load decomposition method, device and equipment |
| CN113378102A (en) * | 2021-05-31 | 2021-09-10 | 国网上海市电力公司 | Data missing preprocessing method, medium and application for short-term load prediction |
| CN113723495A (en) * | 2021-08-25 | 2021-11-30 | 广西大学 | Electromagnetic equipment multi-parameter non-invasive identification method, system, equipment and storage medium based on integrated algorithm |
| CN113899944A (en) * | 2021-09-30 | 2022-01-07 | 广东电网有限责任公司 | Detection method and device for power load switching point |
| CN113899944B (en) * | 2021-09-30 | 2023-11-10 | 广东电网有限责任公司 | Detection method and device for power load switching point |
| CN117081246A (en) * | 2023-08-16 | 2023-11-17 | 北京市计量检测科学研究院 | Indoor electric bicycle identification system that charges and computer equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109145949A (en) | Non-intrusive electrical load monitoring and decomposition method and system based on integrated study | |
| Sun et al. | Using Bayesian deep learning to capture uncertainty for residential net load forecasting | |
| Kong et al. | A practical solution for non-intrusive type II load monitoring based on deep learning and post-processing | |
| Liu et al. | Energy disaggregation of appliances consumptions using ham approach | |
| Garcia-Perez et al. | Fully-convolutional denoising auto-encoders for NILM in large non-residential buildings | |
| Berges et al. | User-centered nonintrusive electricity load monitoring for residential buildings | |
| CN111368904B (en) | Electrical equipment identification method based on electric power fingerprint | |
| CN109492667A (en) | A kind of feature selecting discrimination method for non-intrusive electrical load monitoring | |
| CN106786534A (en) | A kind of non-intrusive electrical load transient process discrimination method and system | |
| CN108051660A (en) | A kind of transformer fault combined diagnosis method for establishing model and diagnostic method | |
| CN109359665B (en) | Household appliance load identification method and device based on support vector machine | |
| Li et al. | A nonintrusive load identification model based on time-frequency features fusion | |
| CN103020459A (en) | Method and system for sensing multiple-dimension electric utilization activities | |
| Verma et al. | A comprehensive review on the nilm algorithms for energy disaggregation | |
| CN106780108A (en) | A kind of distribution transformer state evaluating method based on improvement evidential reasoning fusion | |
| CN113659565B (en) | Online prediction method for frequency situation of new energy power system | |
| CN111369120A (en) | Non-invasive load monitoring method based on equipment transfer learning | |
| CN111242161A (en) | Non-invasive non-resident user load identification method based on intelligent learning | |
| CN109934469A (en) | Based on the heterologous power failure susceptibility method for early warning and device for intersecting regression analysis | |
| CN106295877B (en) | Method for predicting electric energy consumption of smart power grid | |
| Jia et al. | Data compression approach for the home energy management system | |
| CN108122173A (en) | A kind of conglomerate load forecasting method based on depth belief network | |
| Ullah et al. | Adaptive data balancing method using stacking ensemble model and its application to non-technical loss detection in smart grids | |
| CN116842459B (en) | Electric energy metering fault diagnosis method and diagnosis terminal based on small sample learning | |
| Han et al. | Non‐intrusive load monitoring by voltage–current trajectory enabled asymmetric deep supervised hashing |
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 |
Application publication date: 20190104 |
|
| RJ01 | Rejection of invention patent application after publication |