CN113902101A - LSTM neural network algorithm-based non-invasive load identification method - Google Patents
LSTM neural network algorithm-based non-invasive load identification method Download PDFInfo
- Publication number
- CN113902101A CN113902101A CN202111214512.4A CN202111214512A CN113902101A CN 113902101 A CN113902101 A CN 113902101A CN 202111214512 A CN202111214512 A CN 202111214512A CN 113902101 A CN113902101 A CN 113902101A
- Authority
- CN
- China
- Prior art keywords
- neural network
- load identification
- load
- data
- invasive
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 25
- 230000000306 recurrent effect Effects 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims abstract description 7
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 6
- 125000004122 cyclic group Chemical group 0.000 claims abstract description 4
- 238000012806 monitoring device Methods 0.000 claims abstract description 4
- 230000009467 reduction Effects 0.000 claims description 13
- 210000002569 neuron Anatomy 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 description 13
- 238000012544 monitoring process Methods 0.000 description 6
- 230000005611 electricity Effects 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000010223 real-time analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Business, Economics & Management (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)
Abstract
The invention discloses a non-invasive load identification method based on an LSTM neural network algorithm, which comprises the steps of obtaining current waveform data of electric equipment in a steady state, which is monitored by a non-invasive monitoring device; carrying out N-order harmonic decomposition on the steady-state current waveform by using discrete Fourier transform to obtain a sample set, and introducing the training set into a cyclic neural network for training to obtain a load identification prediction model based on the neural network; and bringing the test set into a load identification model based on the recurrent neural network, comparing output results, and generating the optimized load identification model based on the recurrent neural network. The method can accurately identify the electric appliances, especially the low-power electric appliances, under different load conditions, has short identification time, and can monitor the load in real time.
Description
Technical Field
The invention relates to the technical field of non-invasive load identification, in particular to a non-invasive load identification method based on an LSTM neural network algorithm.
Background
Along with rapid energy consumption caused by rapid growth of world population, environmental protection and energy conservation are targets jointly pursued in the current energy field, wherein the residential electricity monitoring technology has wide application and research at home and abroad, the current residential electricity monitoring system mainly comprises an invasive type and a non-invasive type, and the power load monitoring and decomposition can refine the power consumption monitoring to each/type (main) of electric equipment in the total load so as to realize load electricity detail monitoring; the economic loss caused by the conservation of the load model at present is reduced. The residents can further adjust and optimize the electricity utilization behavior according to the monitoring information so as to save electric energy and electricity charge; the method helps a user to quickly and accurately detect, diagnose and clear the electric appliance fault, and replace the high-energy-consumption electric appliance with the high-energy-efficiency electric appliance; the trust and satisfaction of the user to the power supply service of the power company are improved; and a regulation and control basis is provided for the household automation control terminal equipment.
Disclosure of Invention
1. The technical problem to be solved is as follows:
aiming at the technical problems, the invention provides a non-invasive load identification method based on an LSTM neural network algorithm, which realizes real-time analysis of the acquired current signals
2. The technical scheme is as follows:
a non-invasive load identification method based on an LSTM neural network algorithm is characterized by comprising the following steps:
the method comprises the following steps: acquiring current waveform data of the electric equipment in a steady state monitored by a non-invasive monitoring device; the recording interval of the data is in the order of minutes, including between 1 and 60 minutes; the time span of the historical current waveform data is K times of the period in which the repeated change of the load can be observed, and K is an integer greater than or equal to 1;
step two: carrying out N-order harmonic decomposition on the steady-state current waveform by using discrete Fourier transform, and carrying out load identification by using the N-order harmonic characteristics of the obtained current as load characteristics; n is an integer not equal to 0;
step three: performing dimensionality reduction on the harmonic features of the current harmonic generated in the step two to generate a sample set for load identification, and dividing the sample set into a training set and a sample testing set according to a preset proportion; introducing the training set into a circulating neural network for training to obtain a load identification prediction model based on the neural network; the number of input layer neurons of the cyclic neural network is set to be 10, the number of hidden layer neurons is 30, the number of output layer neurons is 1, the number of hidden layers is 2, and a full-connection layer is added between the second hidden layer and the output layer;
step four: bringing the test set into a load identification model based on a recurrent neural network, comparing output results and determining an optimal identification model; and obtaining the deviation degree of the prediction output of the load identification model based on the recurrent neural network and the true value, establishing the mean square error of the fitting result and the true value as a loss function of the model, and generating the optimized load identification model based on the recurrent neural network.
Further, in the third step, a PCA dimension reduction method is adopted to perform dimension reduction processing on the harmonic features of the current harmonic waves, and main information of more than 99% of original data is selected and reserved; if the characteristic dimension of the data set is reduced to K, more than 99% of main information of the original data set can be reserved by satisfying the formula (1):
(1) in the formula x(i)Coordinates of the original data points; x is the number of(i) approxThe numerator represents the mean of the projection error squared and the denominator represents the data total variance for the coordinates of the projected points of the data in the selected direction.
Further, in step four, the loss function is specifically:
y is a sample label value;
and (4) predicting the value of the model.
3. Has the advantages that:
the PCA dimension reduction adopted for the dimension reduction of the current harmonic wave can visualize high-dimensional data in a low-dimensional space; the method has the advantages that linearly related components in the original data are removed, the obtained new data are linearly independent, the data interpretation is easier, the characteristic quantity of the data set is reduced, and the time required by model training and prediction is shortened.
The method can accurately identify the electric appliances, especially the low-power electric appliances, under different load conditions, has short identification time, and can monitor the load in real time.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The present invention will be described in detail with reference to the accompanying drawings.
Firstly, PCA dimension reduction introduction;
PCA is an effective method for reducing the number of features, linear related components in the features are removed, and the scale of feature data is reduced. PCA is a very powerful multidimensional data analysis method that has been well documented in many applications including power system utilities. PCA is mainly used to reduce data dimensionality and identify data patterns, and the main advantages of PCA are the following two:
firstly, high-dimensional data can be visualized in a low-dimensional space;
eliminating linearly related components in the original data, and linearly correlating the obtained new data to make data interpretation easier.
A matrix X of order n X m is given, where n denotes the number of samples observed and m denotes the number of features each sample possesses. The covariance matrix of matrix X is C, the covariance is a measure of how two variables change together, a positive covariance indicates that two variables change simultaneously in a certain direction, and a negative covariance indicates that two variables change oppositely in a certain direction. The covariance value is close to zero, indicating that the two variables are uncorrelated. The covariance matrix is given by expression (3):
in the formula xiIs the measured value of the ith sample;
is the mean of the samples.
And after obtaining a matrix C through the matrix X, processing the matrix C by using a singular value decomposition method to obtain a [ U, S, V ] matrix, and transforming the matrix X by using the first k components of the matrix U to obtain a new matrix Z with the order of nxk, wherein the matrix Z is a new characteristic matrix obtained after being processed by using a PCA method. PCA can be regarded as searching a group of new orthogonal coordinate systems to project original data into the new coordinate systems so as to achieve the purpose of dimension reduction, and therefore the calculation efficiency of the model is optimized.
Introduction of two-cycle neural network model
A recurrent neural network is a special type of neural network that not only takes into account the current inputs to the network, but also has a "memory" function of the input data, i.e. the current outputs of a certain sequence are related to the sequence output values. The recurrent neural network memorizes the historical input information and is applied to the calculation of the current output information.
As shown in the attached figure 1, the non-invasive load identification method based on the LSTM neural network algorithm is characterized by comprising the following steps of:
the method comprises the following steps: acquiring current waveform data of the electric equipment in a steady state monitored by a non-invasive monitoring device; the recording interval of the data is in the order of minutes, including between 1 and 60 minutes; the time span of the historical current waveform data is K times of the period in which the repeated change of the load can be observed, and K is an integer greater than or equal to 1;
step two: carrying out N-order harmonic decomposition on the steady-state current waveform by using discrete Fourier transform, and carrying out load identification by using the N-order harmonic characteristics of the obtained current as load characteristics; n is an integer not equal to 0;
step three: performing dimensionality reduction on the harmonic features of the current harmonic generated in the step two to generate a sample set for load identification, and dividing the sample set into a training set and a sample testing set according to a preset proportion; introducing the training set into a circulating neural network for training to obtain a load identification prediction model based on the neural network; the number of input layer neurons of the cyclic neural network is set to be 10, the number of hidden layer neurons is 30, the number of output layer neurons is 1, the number of hidden layers is 2, and a full-connection layer is added between the second hidden layer and the output layer;
step four: bringing the test set into a load identification model based on a recurrent neural network, comparing output results and determining an optimal identification model; and obtaining the deviation degree of the prediction output of the load identification model based on the recurrent neural network and the true value, establishing the mean square error of the fitting result and the true value as a loss function of the model, and generating the optimized load identification model based on the recurrent neural network.
Further, in the third step, a PCA dimension reduction method is adopted to perform dimension reduction processing on the harmonic features of the current harmonic waves, and main information of more than 99% of original data is selected and reserved; if the characteristic dimension of the data set is reduced to K, more than 99% of main information of the original data set can be reserved by satisfying the formula (1):
(1) in the formula x(i)Coordinates of the original data points; x is the number of(i) approxThe numerator represents the mean of the projection error squared and the denominator represents the data total variance for the coordinates of the projected points of the data in the selected direction.
Further, in step four, the loss function is specifically:
y is a sample label value;
and (4) predicting the value of the model.
Although the present invention has been described with reference to the preferred embodiments, it should be understood that various changes and modifications can be made therein by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (3)
1. A non-invasive load identification method based on an LSTM neural network algorithm is characterized in that: the method comprises the following steps:
the method comprises the following steps: acquiring current waveform data of the electric equipment in a steady state monitored by a non-invasive monitoring device; the recording interval of the data is in the order of minutes, including between 1 and 60 minutes; the time span of the historical current waveform data is K times of the period in which the repeated change of the load can be observed, and K is an integer greater than or equal to 1;
step two: carrying out N-order harmonic decomposition on the steady-state current waveform by using discrete Fourier transform, and carrying out load identification by using the N-order harmonic characteristics of the obtained current as load characteristics; n is an integer not equal to 0;
step three: performing dimensionality reduction on the harmonic features of the current harmonic generated in the step two to generate a sample set for load identification, and dividing the sample set into a training set and a sample testing set according to a preset proportion; introducing the training set into a circulating neural network for training to obtain a load identification prediction model based on the neural network; the number of input layer neurons of the cyclic neural network is set to be 10, the number of hidden layer neurons is 30, the number of output layer neurons is 1, the number of hidden layers is 2, and a full-connection layer is added between the second hidden layer and the output layer;
step four: bringing the test set into a load identification model based on a recurrent neural network, comparing output results and determining an optimal identification model; and obtaining the deviation degree of the prediction output of the load identification model based on the recurrent neural network and the true value, establishing the mean square error of the fitting result and the true value as a loss function of the model, and generating the optimized load identification model based on the recurrent neural network.
2. The LSTM neural network algorithm-based non-invasive load identification method of claim 1, wherein: in the third step, a PCA dimension reduction method is adopted to perform dimension reduction processing on the harmonic characteristics of the current harmonic, and main information of more than 99% of original data is selected and reserved; if the characteristic dimension of the data set is reduced to K, more than 99% of main information of the original data set can be reserved by satisfying the formula (1):
(1) in the formula x(i)Coordinates of the original data points; x is the number of(i) approxThe numerator represents the mean of the projection error squared and the denominator represents the data total variance for the coordinates of the projected points of the data in the selected direction.
3. The LSTM neural network algorithm-based non-invasive load identification method of claim 1, wherein: in the fourth step, the loss function is specifically:
y is a sample label value;
and (4) predicting the value of the model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111214512.4A CN113902101A (en) | 2021-10-19 | 2021-10-19 | LSTM neural network algorithm-based non-invasive load identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111214512.4A CN113902101A (en) | 2021-10-19 | 2021-10-19 | LSTM neural network algorithm-based non-invasive load identification method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113902101A true CN113902101A (en) | 2022-01-07 |
Family
ID=79192782
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111214512.4A Pending CN113902101A (en) | 2021-10-19 | 2021-10-19 | LSTM neural network algorithm-based non-invasive load identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113902101A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114355275A (en) * | 2022-03-21 | 2022-04-15 | 青岛鼎信通讯股份有限公司 | Electric energy meter load monitoring method, system and device and computer readable storage medium |
| CN118246770A (en) * | 2024-05-23 | 2024-06-25 | 华北电力科学研究院有限责任公司 | Method, system, medium and electronic equipment for predicting load harmonic current of power distribution network |
-
2021
- 2021-10-19 CN CN202111214512.4A patent/CN113902101A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114355275A (en) * | 2022-03-21 | 2022-04-15 | 青岛鼎信通讯股份有限公司 | Electric energy meter load monitoring method, system and device and computer readable storage medium |
| CN118246770A (en) * | 2024-05-23 | 2024-06-25 | 华北电力科学研究院有限责任公司 | Method, system, medium and electronic equipment for predicting load harmonic current of power distribution network |
| CN118246770B (en) * | 2024-05-23 | 2024-09-10 | 华北电力科学研究院有限责任公司 | Method, system, medium and electronic equipment for predicting load harmonic current of power distribution network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107025365B (en) | A kind of non-intruding load discrimination method for user side | |
| Liu et al. | Energy disaggregation of appliances consumptions using ham approach | |
| CN106786534B (en) | A kind of non-intrusive electrical load transient process discrimination method and system | |
| Lin et al. | Development of an improved time–frequency analysis-based nonintrusive load monitor for load demand identification | |
| Yu et al. | Nonintrusive appliance load monitoring for smart homes: Recent advances and future issues | |
| CN101567559B (en) | Tabular method of non-intrusive electrical load decomposition | |
| Andrean et al. | A hybrid method of cascade-filtering and committee decision mechanism for non-intrusive load monitoring | |
| CN113902101A (en) | LSTM neural network algorithm-based non-invasive load identification method | |
| CN109633301B (en) | Non-invasive electrical appliance load identification method based on quantum genetic optimization | |
| Kang et al. | Adaptive non-intrusive load monitoring based on feature fusion | |
| Gong et al. | A svm optimized by particle swarm optimization approach to load disaggregation in non-intrusive load monitoring in smart homes | |
| Li et al. | Non-intrusive load monitoring based on harmonic characteristics | |
| Gurbuz et al. | Comprehensive non-intrusive load monitoring process: Device event detection, device feature extraction and device identification using KNN, random forest and decision tree | |
| Bucci et al. | A non-intrusive load identification system based on frequency response analysis | |
| Gurbuz et al. | A brief review of non-intrusive load monitoring and its impact on social life | |
| Gowrienanthan et al. | Deep learning based non-intrusive load monitoring for a three-phase system | |
| Lee et al. | Novel monitoring system for low-voltage DC distribution network using deep-learning-based disaggregation | |
| Zheng et al. | Non-intrusive load monitoring based on the graph least squares reconstruction method | |
| Shang et al. | Effinformer: A deep-learning-based data-driven modeling of DC–DC bidirectional converters | |
| Sala et al. | Disaggregation of HVAC load profiles for the monitoring of individual equipment | |
| Lundstrom et al. | High-frequency, multiclass nonintrusive load monitoring for grid-interactive residential buildings | |
| Xiong et al. | Research on Industrial Load Decomposition Based on Correlation Analysis and DNN | |
| Zhang et al. | Research on Data-driven Self-diagnosis for Measurement Errors in Capacitor Voltage Transformers | |
| Choksi et al. | Novel computational-index as a representative feature for non-intrusive load monitoring | |
| Du et al. | Two-Step Non-Intrusive Load Monitoring Based on High-Frequency Current Data |
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 |