CN112598303A - Non-invasive load decomposition method based on combination of 1D convolutional neural network and LSTM - Google Patents
Non-invasive load decomposition method based on combination of 1D convolutional neural network and LSTM Download PDFInfo
- Publication number
- CN112598303A CN112598303A CN202011578297.1A CN202011578297A CN112598303A CN 112598303 A CN112598303 A CN 112598303A CN 202011578297 A CN202011578297 A CN 202011578297A CN 112598303 A CN112598303 A CN 112598303A
- Authority
- CN
- China
- Prior art keywords
- layer
- neural network
- lstm
- load
- active power
- 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
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 19
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 19
- 238000003062 neural network model Methods 0.000 claims abstract description 29
- 238000012549 training Methods 0.000 claims abstract description 20
- 230000006870 function Effects 0.000 claims description 29
- 230000004913 activation Effects 0.000 claims description 28
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 238000012886 linear function Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 4
- 238000012544 monitoring process Methods 0.000 abstract description 8
- 230000008569 process Effects 0.000 abstract description 6
- 238000001514 detection method Methods 0.000 abstract description 2
- 230000007246 mechanism Effects 0.000 abstract description 2
- 230000005611 electricity Effects 0.000 description 10
- 238000013135 deep learning Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000001052 transient effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000012567 pattern recognition method Methods 0.000 description 1
- 239000011541 reaction mixture Substances 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- 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)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Educational Administration (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a non-invasive load decomposition method based on combination of a 1D convolutional neural network and LSTM, which comprises the following steps: s1, collecting active power time sequence data of total load and various single loads in a period of time as sample data; s2, respectively training corresponding neural network models for various single loads, specifically: training a neural network model combining a 1D convolutional neural network and an LSTM to obtain neural network models corresponding to various single loads by taking active power time sequence data of a total load as input and active power time sequence data of various single loads as output; and S3, inputting the active power data of the total load into the trained neural network model, and decomposing the active power data of the corresponding single load. The invention does not need to utilize an event monitoring mechanism to carry out variable point detection on the load curve, thereby greatly simplifying the load decomposition process.
Description
Technical Field
The invention belongs to the technical field of electric load decomposition, and particularly relates to a non-invasive load decomposition method based on combination of a 1D convolutional neural network and an LSTM.
Background
In recent years, with the construction of the ubiquitous power internet of things and the coming of the intelligent electricity utilization era, the intelligent electric meter is expected to become a mainstream product in various terminals of the national power grid. Through the analysis of the electricity consumption data acquired by the intelligent electric meter end in real time, the electricity consumption law and the electricity consumption of the load can be accurately mastered in real time, so that the electricity consumption of a user is more reasonably guided, and the aim of response of a demand side is fulfilled. In addition, real-time monitoring is carried out on the use of the power resources through the intelligent electric meter, and measures are taken to reduce energy waste when necessary, so that the intelligent electric meter has great research significance for reasonably utilizing the power resources.
The non-intrusive load decomposition is to realize the decomposition of total power data into the power consumption of a single electric appliance without installing a large number of sensors and monitoring equipment. The non-intrusive load decomposition only needs to add a non-intrusive load decomposition module into a total electric meter of a user, and decomposes the running states and related parameter information of different loads of the user according to a measured time sequence value, thereby realizing the online monitoring of the power consumption of the load. This has not only saved a large amount of monitoring devices and loaded down with trivial details sensor device, greatly reduced economic cost, effectively avoided the intrusive load monitoring to bring actual maneuverability poor, implementation cost and maintenance cost are all high, the problem such as the equipment maintenance repair also can be very inconvenient, has still improved the stability and the reliability of system monitoring simultaneously, and can not produce too much disturbance to user's production and life.
The non-intrusive load decomposition is mainly oriented to power grid companies, and the technology is more inclined to know the power consumption of various electrical appliances and does not care about the very accurate on/off time of the electrical appliances. At present, the conventional method combines transient and steady-state characteristics of a power system and adopts a pattern recognition method to decompose non-invasive electric loads, but the accuracy of decomposition of electric loads with similar characteristics is generally low by the methods. The load resolution technology is still a very challenging problem, the research in this field is still in an immature stage, and a significant technical problem is still to be overcome.
Disclosure of Invention
The invention aims to provide a non-invasive load decomposition method based on the combination of a 1D convolutional neural network and an LSTM (least squares TM). the method can be used for decomposing total power information acquired by a sensor arranged at a load inlet to obtain active power information of each electric load at different moments.
The invention provides a non-invasive load decomposition method based on combination of a 1D convolutional neural network and an LSTM, which comprises the following steps:
s1, collecting active power time sequence data of total load and various single loads in a period of time as sample data;
s2, respectively training corresponding neural network models for various single loads, specifically: training a neural network model combining a 1D convolutional neural network and an LSTM to obtain neural network models corresponding to various single loads by taking active power time sequence data of a total load as input and active power time sequence data of various single loads as output;
the neural network model combining the 1D convolutional neural network and the LSTM comprises an input layer, a hidden layer and an output layer;
the input layer is used for receiving a one-dimensional vector, namely active power time sequence data of a total load;
the hidden layer comprises 3 convolutional layers, 2 LSTM layers and 2 full-connection layers; the first convolution layer is composed of 32 convolution kernels with the dimensionality of 5, the convolution step size of the first convolution layer is 1, and the input of an activation function is normalized by adopting a Batchnormalization algorithm; the second convolution layer is composed of 32 convolution kernels with the dimensionality of 5, the convolution step length of the second convolution kernel is 1, and the input of the activation function is normalized by adopting a Batch Normalization algorithm; the third convolution layer is composed of 16 convolution kernels with 4 dimensionalities, the convolution step length is 1, and the input of the activation function is normalized by adopting a Batchnormalization algorithm; the first LSTM layer is output by adopting 64 dimensions; the second LSTM layer is output with the size of 128 dimensions; the second LSTM layer is followed by a full connection layer with a dimension of 64, the result obtained by the last layer is fully connected, then a ReLU activation function is called, full connection with a dimension of 1 is carried out again, and after linear change for one time, final output, namely active power information of a certain specific load is obtained;
the output layer outputs the active power information of the load corresponding to the neural network;
and S3, inputting the active power data of the total load into the trained neural network model, and decomposing the active power data of the corresponding single load.
Furthermore, the activation function in the first fully-connected layer adopts a ReLU activation function, and the activation functions in the convolutional layer, the LSTM layer and the second fully-connected layer are all linear functions.
Preferably, a Dropout layer with the parameter of 0.2 is arranged between every two convolution layers in the hidden layer;
there is also a Dropout layer with parameter 0.2 between the first LSTM layer and the second LSTM layer;
there is a Dropout layer with a parameter of 0.5 between the second LSTM layer and the fully connected layer, and between the two fully connected layers.
The invention has the following advantages and beneficial effects:
the invention does not need to utilize an event monitoring mechanism to carry out variable point detection on the load curve, thereby greatly simplifying the load decomposition process.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
fig. 2 is a schematic diagram of a network model used in the embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Before describing the embodiments, it is necessary to explain, hereinafter and throughout, that the load is a ready-to-use electrical device.
Active power can be one of the main features that distinguish different loads. The active power of the total load of the user side and the active power of each single load are respectively used as the input and the output of the neural network model, namely the input layer is the active power data of the total load of the user side, and the output layer is the decomposed active power data of the single load. And training the constructed neural network model combining the 1D convolutional neural network and the LSTM by adopting the active power of the total load and the active power of each single load, and outputting the power consumption information of a specific single load only by inputting the active power data of the total load into the trained neural network model after the training is finished.
Assuming that the single-load electricity consumption data of the family in the time period t isAssuming that m loads are used in total as the types of loads in the monitoring environment, the electricity consumption data of each load is accumulated as the total electricity consumption PtI.e. active power data of the total load. Time series (P) of total load active power of family1,P2,…,Pt),(Pt+1,Pt+2,…,P2t),…,(P(n-1)×t+1,P(n-1)×t+2,…,Pn×t) As input samples of the constructed neural network model, each active power subscript in the time series of the active power represents a time period, i.e., PiRepresents the total load active power corresponding to the time interval i, i is 1, 2, … n × t; the active power of the single load to be decomposed in the corresponding time interval n multiplied by tAs output samples of the neural network model. And continuously training the constructed neural network model by adopting sample data until the accuracy requirement is met. And for each power consumption load, training a corresponding neural network model according to the same method, wherein each trained network model is equivalent to the characteristic extractor of the specific load, and for any total load power consumption curve, as long as the total load power consumption curve is input, the network model can separate the power consumption information of the specific load from the total power consumption curve.
The invention realizes the aim that the active power data of the total load in a certain period of time is taken as input data, and an active power curve of a certain specific load is output through a trained neural network combining a 1D convolutional neural network and an LSTM. For load decomposition, the designed neural network model comprises an input layer, a hidden layer and an output layer. The input layer adopts the active power of the total load as the input of the neural network. The output layer has only one node, which corresponds to a certain load. And giving electricity utilization data in a certain time period, and outputting the information of the electricity utilization quantity of the specific load in the time period as an output result. The invention needs to train each load individually, one neural network model only corresponds to one family load, and the output of the neuron is the power consumption information of the load.
The following provides a method for constructing a neural network model combining a 1D convolutional neural network and an LSTM.
The problem to be solved by non-invasive load splitting is to identify the load components from the collected total load power usage curve, and the present invention utilizes an improved neural network to solve this problem.
As shown in fig. 1, the constructed neural network model structure of the combination of the 1D convolutional neural network and the LSTM. And training the active power of the load by using the neural network model, and extracting the corresponding active power curve characteristics. And further constructing a neural network model structure of each class of load as a recognizer for extracting the corresponding class of load. After training is finished, the model is applied to an inference scene, and an input total load power utilization curve is decomposed, so that the purpose of decomposing a certain load from each signal is achieved.
Specifically, a time window x (t) with the length t from a certain time point is given, the time window x (t) comprises total active power information of K loads, and a model g is adopted(j)(x (t), θ) estimating a power curve of the jth load in the corresponding time series, wherein θ is a network model parameter which is automatically determined by model training:
in the formula (1), the reaction mixture is,representing the subnet of activation signals from the input layer to the output layer of the neural network model,the meaning is that the output signal is obtained after the active power of the jth load under the time window x (t) passes through the neural network. The output layer is activated by an activation function sigma, mapping the activated output into an effective data range. The activation function σ uses a ReLU function, that is, σ (γ) ═ max (0, γ), and γ represents an input of the activation function. Substituting equation (1) with:
ω(j)(t) represents the output of the activation function.
In the training phase, (x (t), ω (t)) is referred to as a training segment with K data samples. The training segments are extracted from the mixed-signal time series (x (t), ω (t)) using non-overlapping windows such that the inputs to the training set are:
x=((x1,x2,...,xk),(xk+1,xk+2,...,x2k),...,(x(n-1)*k+1,x(n-1)*k+2,...,xnk)) (3)
in the formula (3), (x)1,x2,...,xk),(xk+1,xk+2,...,x2k),...,(x(n-1)*k+1,x(n-1)*k+2,...,xnk) I.e. the aforementioned time series (P) of the total load active power of the household1,P2,…,Pt),(Pt+1,Pt+2,…,P2t),…,(P(n-1)×t+1,P(n-1)×t+2,…,Pn×t)。
The corresponding single load output power Ω is noted as:
Ω=(ω(k),ω(2k),...,ω(n*k)) (4)
then, the output error of real active power relative to a single load is calculated, and a network model is trained by using Mean Square Error (MSE) as a loss function.
The structure of the neural network model based on the combination of the 1D convolutional neural network and the LSTM constructed by the invention is described as follows.
As shown in fig. 2, the neural network includes an input layer and a hidden layer and an output layer. The input layer is a total active power time sequence of the load in the designated timestamp, which is a one-dimensional vector and contains the information of the total active power sampled according to a certain frequency.
The hidden layer comprises 3 convolutional layers, 2 LSTM layers and 2 fully-connected layers, the first convolutional layer is composed of 32 convolutional kernels with 5 dimensions, the convolution step size 1 is that the input of the activation function is normalized by adopting a Batch Normalization algorithm (namely BN in figure 1) so as to solve the influence of offset and increase of input data. The activation function is a linear function. The second convolutional layer is also composed of 32 convolutional kernels with 5 dimensions, the convolution step is 1, the input of the activation function is normalized by the same Batch Normalization algorithm, and the activation function is a linear function. The third convolution layer is composed of 16 convolution kernels with 4 dimensions, the convolution step size is 1, the input of the activation function is normalized by adopting a Batch Normalization algorithm, and the activation function adopts a linear function. The first LSTM layer is output with a size of 64 dimensions and the activation function is a linear function. The second LSTM layer is output with a size of 128 dimensions and the activation function is a linear function. The second LSTM layer is followed by a full connection layer with a dimension of 64, the result obtained by the last LSTM layer is fully connected, then a ReLU activation function is called, full connection with a dimension of 1 is carried out again, and after linear change once, the final output, namely the active power information of a certain specific load is obtained.
And the output layer outputs the active power information of the load corresponding to the neural network.
There is also a Dropout layer with parameter 0.2 between every two convolutional layers in the hidden layer. There is also a Dropout layer with a parameter of 0.2 between the first LSTM layer and the second LSTM layer. There is a Dropout layer with a parameter of 0.5 between the second LSTM layer and the fully connected layer, and between the two fully connected layers. The purpose of the Dropout layer is to remove the training unit of the neural network from the network according to a certain probability in the training process of deep learning, and because the training unit is randomly discarded, a different network is trained each time the batch _ size samples are randomly selected.
The deep learning model has strong learning and expression capabilities, although the data collected from the smart meter is composite data obtained by overlapping each load, the relationship between the data and each decomposition data is very complex, and no objective rule can be followed, the deep learning model uses a plurality of nonlinear hidden layers to re-express the input data through continuously learning deep features of an object. Furthermore, these abstract features also express well the relationship between inputs and outputs. Compared with the traditional method, when the time series data is processed by adopting the deep learning method, the transient state and the steady state process of the load do not need to be distinguished, and the end-to-end load decomposition process can be realized without detecting the switching event of the load.
Those skilled in the art will appreciate that, in the embodiments of the methods of the present invention, the sequence numbers of the steps are not used to limit the sequence of the steps, and it is within the scope of the present invention for those skilled in the art to change the sequence of the steps without inventive work. The examples described herein are intended to aid the reader in understanding the practice of the invention and it is to be understood that the scope of the invention is not limited to such specific statements and examples. Those skilled in the art can make various other specific changes and combinations based on the teachings of the present invention without departing from the spirit of the invention, and these changes and combinations are within the scope of the invention.
Claims (3)
1. The non-invasive load decomposition method based on the combination of the 1D convolutional neural network and the LSTM is characterized by comprising the following steps:
s1, collecting active power time sequence data of total load and various single loads in a period of time as sample data;
s2, respectively training corresponding neural network models for various single loads, specifically: training a neural network model combining a 1D convolutional neural network and an LSTM to obtain neural network models corresponding to various single loads by taking active power time sequence data of a total load as input and active power time sequence data of various single loads as output;
the neural network model combining the 1D convolutional neural network and the LSTM comprises an input layer, a hidden layer and an output layer;
the input layer is used for receiving a one-dimensional vector, namely active power time sequence data of a total load;
the hidden layer comprises 3 convolutional layers, 2 LSTM layers and 2 full-connection layers; the first convolution layer is composed of 32 convolution kernels with the dimensionality of 5, the convolution step size of the first convolution layer is 1, and the input of an activation function is normalized by adopting a Batch Normalization algorithm; the second convolution layer is composed of 32 convolution kernels with the dimensionality of 5, the convolution step length of the second convolution kernel is 1, and the input of the activation function is normalized by adopting a Batch Normalization algorithm; the third convolution layer is composed of 16 convolution kernels with 4 dimensionalities, the convolution step length is 1, and the input of the activation function is normalized by adopting a Batch Normalization algorithm; the first LSTM layer is output by adopting 64 dimensions; the second LSTM layer is output with the size of 128 dimensions; the second LSTM layer is followed by a full connection layer with a dimension of 64, the result obtained by the last layer is fully connected, then a ReLU activation function is called, full connection with a dimension of 1 is carried out again, and after linear change for one time, final output, namely active power information of a certain specific load is obtained;
the output layer outputs the active power information of the load corresponding to the neural network;
and S3, inputting the active power data of the total load into the trained neural network model, and decomposing the active power data of the corresponding single load.
2. The method of claim 1 for non-invasive load splitting based on a combination of a 1D convolutional neural network and LSTM, wherein:
the activation function in the first fully-connected layer is a ReLU activation function, and the activation functions in the convolutional layer, the LSTM layer and the second fully-connected layer are linear functions.
3. The method of claim 1 for non-invasive load splitting based on a combination of a 1D convolutional neural network and LSTM, wherein:
a Dropout layer with the parameter of 0.2 is arranged between every two convolution layers in the hidden layer;
there is also a Dropout layer with parameter 0.2 between the first LSTM layer and the second LSTM layer;
there is a Dropout layer with a parameter of 0.5 between the second LSTM layer and the fully connected layer, and between the two fully connected layers.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011578297.1A CN112598303A (en) | 2020-12-28 | 2020-12-28 | Non-invasive load decomposition method based on combination of 1D convolutional neural network and LSTM |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011578297.1A CN112598303A (en) | 2020-12-28 | 2020-12-28 | Non-invasive load decomposition method based on combination of 1D convolutional neural network and LSTM |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112598303A true CN112598303A (en) | 2021-04-02 |
Family
ID=75203715
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011578297.1A Pending CN112598303A (en) | 2020-12-28 | 2020-12-28 | Non-invasive load decomposition method based on combination of 1D convolutional neural network and LSTM |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112598303A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN113408210A (en) * | 2021-07-12 | 2021-09-17 | 内蒙古电力(集团)有限责任公司乌兰察布电业局 | Deep learning based non-intrusive load splitting method, system, medium, and apparatus |
CN113592671A (en) * | 2021-07-30 | 2021-11-02 | 上海电力大学 | Long-time neural network-based resident load curve decomposition method |
CN113837894A (en) * | 2021-08-06 | 2021-12-24 | 国网江苏省电力有限公司南京供电分公司 | Non-invasive resident user load decomposition method based on residual convolution module |
CN113970667A (en) * | 2021-10-10 | 2022-01-25 | 上海梦象智能科技有限公司 | Non-invasive load monitoring method based on midpoint value of prediction window |
CN114511058A (en) * | 2022-01-27 | 2022-05-17 | 国网江苏省电力有限公司泰州供电分公司 | Load element construction method and device for power consumer portrait |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106096726A (en) * | 2016-05-31 | 2016-11-09 | 华北电力大学 | A kind of non-intrusion type load monitoring method and device |
CN107730003A (en) * | 2017-10-23 | 2018-02-23 | 华中科技大学 | One kind supports more high-precision NILM implementation methods of appliance type |
CN108899892A (en) * | 2018-06-08 | 2018-11-27 | 上海电力学院 | One kind being based on CNN non-intrusion type power load decomposition method |
CN112001486A (en) * | 2020-08-28 | 2020-11-27 | 河北工业大学 | Load decomposition method based on deep learning |
-
2020
- 2020-12-28 CN CN202011578297.1A patent/CN112598303A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106096726A (en) * | 2016-05-31 | 2016-11-09 | 华北电力大学 | A kind of non-intrusion type load monitoring method and device |
CN107730003A (en) * | 2017-10-23 | 2018-02-23 | 华中科技大学 | One kind supports more high-precision NILM implementation methods of appliance type |
CN108899892A (en) * | 2018-06-08 | 2018-11-27 | 上海电力学院 | One kind being based on CNN non-intrusion type power load decomposition method |
CN112001486A (en) * | 2020-08-28 | 2020-11-27 | 河北工业大学 | Load decomposition method based on deep learning |
Non-Patent Citations (2)
Title |
---|
姜波: "基于深度学习的非侵入式住宅用电负荷分解方法", 《中国优秀硕士论文全文数据库 工程科技Ⅱ辑》, no. 02, 15 February 2018 (2018-02-15), pages 042 - 852 * |
梁家阁: "基于可变输入LSTM和迁移学习的非侵入式电力负荷分解", 《中国优秀硕士论文全文数据库 工程科技Ⅱ辑》, no. 06, 15 June 2020 (2020-06-15), pages 042 - 627 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
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 |
CN113408210A (en) * | 2021-07-12 | 2021-09-17 | 内蒙古电力(集团)有限责任公司乌兰察布电业局 | Deep learning based non-intrusive load splitting method, system, medium, and apparatus |
CN113592671A (en) * | 2021-07-30 | 2021-11-02 | 上海电力大学 | Long-time neural network-based resident load curve decomposition method |
CN113592671B (en) * | 2021-07-30 | 2024-04-26 | 上海电力大学 | Resident load curve decomposition method based on long-short time neural network |
CN113837894A (en) * | 2021-08-06 | 2021-12-24 | 国网江苏省电力有限公司南京供电分公司 | Non-invasive resident user load decomposition method based on residual convolution module |
CN113837894B (en) * | 2021-08-06 | 2023-12-19 | 国网江苏省电力有限公司南京供电分公司 | Non-invasive resident user load decomposition method based on residual convolution module |
CN113970667A (en) * | 2021-10-10 | 2022-01-25 | 上海梦象智能科技有限公司 | Non-invasive load monitoring method based on midpoint value of prediction window |
CN113970667B (en) * | 2021-10-10 | 2024-04-05 | 上海梦象智能科技有限公司 | Non-invasive load monitoring method based on predicted window midpoint value |
CN114511058A (en) * | 2022-01-27 | 2022-05-17 | 国网江苏省电力有限公司泰州供电分公司 | Load element construction method and device for power consumer portrait |
CN114511058B (en) * | 2022-01-27 | 2023-06-02 | 国网江苏省电力有限公司泰州供电分公司 | Load element construction method and device for electric power user portrait |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112598303A (en) | Non-invasive load decomposition method based on combination of 1D convolutional neural network and LSTM | |
CN109635928B (en) | Voltage sag reason identification method based on deep learning model fusion | |
CN104655423B (en) | A kind of Fault Diagnosis of Roller Bearings based on time-frequency domain multi-dimensional vibration Fusion Features | |
CN110516788B (en) | Non-invasive load detection method and system based on Bi-LSTM algorithm | |
CN103616635A (en) | Method and device for diagnosing mechanical characteristic failures of high-voltage circuit-breaker | |
CN112434799B (en) | Non-invasive load identification method based on full convolution neural network | |
CN116381517A (en) | Lithium battery remaining life probability prediction method based on time convolution attention mechanism | |
CN107807860B (en) | Power failure analysis method and system based on matrix decomposition | |
CN111242276B (en) | One-dimensional convolutional neural network construction method for load current signal identification | |
CN113822418A (en) | Wind power plant power prediction method, system, device and storage medium | |
CN112215405A (en) | Non-invasive type residential electricity load decomposition method based on DANN domain adaptive learning | |
CN102262211A (en) | Analog circuit failure diagnosis method based on dynamic classification of echo state network | |
CN111639586A (en) | Non-invasive load identification model construction method, load identification method and system | |
Zu et al. | A simple gated recurrent network for detection of power quality disturbances | |
Yoon et al. | Deep learning-based method for the robust and efficient fault diagnosis in the electric power system | |
Zhu et al. | New feature extraction method for photovoltaic array output time series and its application in fault diagnosis | |
CN110866840A (en) | Database modeling method for power load characteristic quantity training based on knowledge graph | |
Xiang et al. | State-of-health prognosis for lithium-ion batteries considering the limitations in measurements via maximal information entropy and collective sparse variational gaussian process | |
CN115902557A (en) | Switch cabinet fault diagnosis processing method and device and nonvolatile storage medium | |
CN115293265A (en) | Non-invasive monitoring method and system for building electricity utilization | |
Bing et al. | Machinery fault diagnosis method of HV circuit breaker based on EEMD and RBF neural network | |
Lei et al. | Signal feature extract based on dual-channel wavelet convolutional network mixed with hypergraph convolutional network for fault diagnosis | |
Wang et al. | Real-time fault diagnosis method for low-voltage power line carrier communication network based on network topology | |
Liu et al. | SD2SDFNet: A novel deep fusion network based on statistical denoising and dual-dimension self-attention model for power transformer hybrid-space fault prognosis | |
Sun et al. | A new fault diagnosis method for HV circuit breakers based on wavelet packet-neural network |
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 |