CN113036754A - Non-invasive load identification method based on improved DAG-SVMS - Google Patents
Non-invasive load identification method based on improved DAG-SVMS Download PDFInfo
- Publication number
- CN113036754A CN113036754A CN202110252776.2A CN202110252776A CN113036754A CN 113036754 A CN113036754 A CN 113036754A CN 202110252776 A CN202110252776 A CN 202110252776A CN 113036754 A CN113036754 A CN 113036754A
- Authority
- CN
- China
- Prior art keywords
- load
- transient
- identification
- module
- transient event
- 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
Images
Classifications
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2310/00—The network for supplying or distributing electric power characterised by its spatial reach or by the load
- H02J2310/70—Load identification
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/40—Display of information, e.g. of data or controls
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a non-intrusive load identification method based on an improved DAG-SVMS. The method comprises the following steps: the system comprises a data acquisition module, an event detection module, a feature extraction module, a load identification module and a data transmission module. The data acquisition module acquires total current data at an inlet of the power system; the event detection module is used for carrying out transient event detection on the current; after the transient event is detected, the characteristic extraction module performs transient waveform separation, and performs characteristic normalization processing on the separated transient current waveform characteristics; the load identification module is trained by using a DAG-SVMS algorithm to generate a load identification model, the classifier is used for performing parameter optimization by using a particle swarm algorithm, and the arrangement sequence of DAG-SVMS nodes is optimized by using a Gini index so as to reduce accumulated errors and improve the identification accuracy of the model. When the power system comprises various loads, particularly, variable frequency loads with load state changes, the method provided by the invention can accurately identify the types and time of the changed loads, and has high identification accuracy and high speed.
Description
Technical Field
The invention relates to the field of smart power grids, and particularly provides a non-intrusive load identification method based on improved DAG-SVMS
Background
The non-invasive load identification technology is embedded in a power supply inlet, so that power demand side management is facilitated, power consumption information exchange between power consumers and a power grid is realized, the power consumption of the consumers is optimized, the construction of a smart power grid system is perfected, and meanwhile, the development of smart home is promoted.
From the aspect of load feature extraction, non-intrusive load identification technologies can be divided into transient-based and steady-state-based technologies, the extraction of transient features needs high sampling frequency, the requirement on hardware is high, most of the current non-intrusive load identification technologies are based on the steady state, but with the development of large-scale integrated circuits, the hardware cost is reduced, and the development and popularization of the transient-based non-intrusive load identification technologies become possible. Compared with a steady state, the transient state characteristics can provide more detailed information, the variable frequency device identification with continuously changed states has better performance, the identification accuracy is higher, secondly, the characteristic extraction in the transient state process is used as input, the identification of multiple loads is converted into a single load identification model, and the identification complexity is reduced.
In conclusion, the invention discloses a non-intrusive load identification method based on an improved directed acyclic graph, which comprises the steps of firstly detecting a system transient process, carrying out transient waveform separation for transient characteristic extraction after detecting the transient process, training and generating a load identification model by using a DAG-SVMS algorithm after carrying out normalization processing on characteristic quantities, carrying out parameter optimization on a classifier in the model by using a PSO algorithm, finally optimizing the node arrangement sequence of the classification model, reducing accumulated errors and improving the identification accuracy of the model. Compared with the common method, the method has the advantages of high identification accuracy, high identification speed and strong practicability.
Disclosure of Invention
The invention discloses a non-invasive load identification method based on improved DAG-SVMS, which has the advantages of high identification accuracy, high identification speed and strong practicability compared with the common method.
One of the purposes of the sound generation is realized through the scheme that the non-intrusive load identification method based on the improved DAG-SVMS comprises a data acquisition module, an event detection module, a feature extraction module, a load identification module and a data transmission module.
The data acquisition module is used for acquiring total current data at an inlet of the power system, and the acquisition mode is that the input end of the current transformer is connected with a live wire and the output end of the current transformer is subjected to A/D conversion.
The event detection module detects the collected total current data by using an event detection algorithm, and stores the initial time of the transient event after the transient event is detected.
The characteristic extraction module is used for carrying out waveform separation on the transient current waveform after the transient event is detected by the event detection module, acquiring a target load transient waveform causing the transient event, and further extracting transient characteristics for distinguishing different loads.
The load identification module comprises a load identification model training stage and a load real-time identification stage, wherein the load identification model training stage is used for training a load identification model by taking sample library data as support; and in the load identification stage, target load characteristics causing the transient event are extracted in real time and input into an identification model for identification.
And the result data transmission module feeds the transient initial time detected by the event detection module and the identification result of the load identification module back to the user during real-time identification, so that the user can monitor the switching time and the working state of the power load in real time.
Furthermore, the data acquisition module performs high-frequency sampling and A/D conversion on the current data at the output end of the current transformer to acquire total current data and provide a data source for the event detection module.
Further, the event detection module uses an event detection algorithm to perform transient event monitoring, where the event detection algorithm is a transient event monitoring algorithm based on a heuristic method, and specifically includes:
the load current level defining the tth period is defined as:
in the above formula, K is the total number of sampling points of the current in one period; i (k) is the current value of the kth sampling point in the T period.
When the internal load running states of the power system are not changed, the current intensity difference value delta I of adjacent periodsintensityApproaching 0, the system is in transient state once, and Δ IintensityIf the delta is exceeded, the transient process of the load in the system is judged, and the process is represented as:
the system, upon detecting the generation of a transient event, executes a transient over-time algorithm, Δ IintensityIf the number smaller than epsilon is larger than gamma, the system transient process is judged to be finished, and the following is expressed:
and after the transient event is monitored, recording the current starting time of the second period as the transient event starting time, judging that the transient event is ended, subtracting gamma periods from the current time to obtain the transient event ending time, and subtracting the starting time and the ending time to obtain the transient duration which is integral multiple of the current period.
Further, the feature extraction module performs phase difference on the transient event current waveform and the current waveform before the transient event to obtain a target load transient current waveform; the current waveforms are in phase and are differentiated, namely transient event current data with the same period length are differentiated from transient event pre-steady state current data;
further, the feature extraction module performs feature extraction on the target load transient current waveform, where the feature extraction includes: duration (difference between the starting time and the ending time of the transient current waveform), maximum value (maximum value of the current waveform), peak-to-peak value (difference between the maximum value and the minimum value of the current waveform), and absolute average value (absolute value of the current waveform is taken first and then average) of the average value (current waveform intensity, the number of sampling points is the number of current waveform points), the dynamic load-shedding current-limiting method based on the transient event comprises the following steps of (1) a steady-state waveform effective value (a steady-state current intensity difference value before and after a transient event) peak-to-average ratio (a peak-to-current intensity ratio), a waveform factor (a ratio of current intensity to an absolute average value), a crest factor (a maximum value and a current intensity ratio), and a load zone bit (judging whether a load is put into or cut off according to current intensity change before and after the transient event, wherein the current intensity is increased to be put into, the current intensity is reduced to be cut.
Further, the load identification module comprises a load identification model training stage and a load real-time identification stage. The load identification model training stage, the sample library data as the data source, training the load identification model; and in the load identification stage, target load characteristics causing the transient event are extracted in real time and input into an identification model for identification.
In the load identification model training stage, after the transient characteristics of each power load in the power system are extracted by the characteristic extraction module, labeling is carried out, a sample database containing the characteristics of each power load is formed after normalization, and the samples are divided into an input type and a cut-off type according to the characteristic zone bits and used as data sources for training the identification model.
Further, the load identification module trains the load identification model by the specific steps of:
step 1: two types of data samples of a sample database are combined and grouped in pairs, wherein any two types of loads in each type of sample are combined;
step 2: optimizing a support vector machine error penalty parameter C and a Gaussian radial basis kernel parameter g by using a PSO algorithm, and training a classifier for each group respectively;
and step 3: respectively building directed acyclic graph models for classifiers obtained by training each group of data in the two types of samples, and performing node sequence optimization by using Gini indexes to obtain two identification models;
further, in the load identification model training stage, in step 2, the parameters of the classifier are optimized by using a particle swarm algorithm, the fitness function adopts K-cross validation classification accuracy, the kernel function uses a Gaussian radial basis kernel function, the speed and position range is set, the speed and position are initialized randomly, and the combination of the C and g parameters is searched when the performance of each classifier is optimal. The velocity and position are updated as follows:
vid(t+1)=w*vid(t)+c1r1(pid-xid(t))+c2r2(pgd-xid(t))
xid(t+1)=xid(t)+vid(t+1)
in the above formula: v. ofidIs the velocity of the particle; x is the number ofidIs the position of the particle; w is the inertial weight; c. C1、c2Is a learning factor; r is1,r2Is uniformly distributed in [0, 1 ]]Random numbers within a range.
Further, in the load identification model training stage, in step 3, the directed acyclic graph is respectively built on classifiers trained by two types of samples to form a multi-classification model, because the model adopts a hierarchical structure and error accumulation is a classification defect inevitable to the hierarchical structure, the high-level nodes have a larger influence on the classification structure, and a Gini index is adopted to perform node sequence optimization as follows:
wherein p isiThe probability that the sample in the D belongs to the ith class is calculated; d represents the entire training set of data samples.
If the sample set is SVM by a certain nodexyDivided into two subsets D1,D2Then, after the division, the Gini index of the sample is reduced to:
through this division, the Gini exponent difference produced is:
ΔGini(SVMxy)=Gini(D)-Ginixy(D)
the optimization of the node sequence is as follows: firstly, each classifier is utilized to divide a sample training set, a node with the maximum delta Gini is selected as a root node, and then a divided sample set D is used1,D2And respectively regarding the nodes as independent sets, selecting the classifier with the largest Gini index difference value in the next layer as a current node, and building a directed acyclic graph classification model by means of circulation.
In the load identification stage, the data acquisition module acquires total current data, the transient event module detects a transient event, the feature extraction module extracts features after separating a transient current waveform of a target load, and the features are input into a corresponding directed acyclic graph model for identification according to a load identification bit.
Further, the result data transmission module comprises the transient starting time detected by the event detection module and the load identification module identification result, and the result is sent to the user terminal through WIFI and GPRS or transmitted to a cloud storage for the user terminal to access.
The invention has the following advantages and beneficial effects:
from the perspective of saving hardware cost, most of the existing non-invasive load identification technical schemes are realized by low sampling based on steady state, the high-frequency sampling identification technology based on event monitoring has higher requirements on hardware, but the identification effect is greatly improved, and simultaneously, along with the development of large-scale integrated circuits, the hardware cost is reduced, and the popularization of the non-invasive load identification technology based on transient state becomes possible. The invention provides a non-intrusive load identification method based on improved DAG-SVMS, which converts multi-load identification into a single-load identification model, reduces the identification complexity, can effectively solve the defects of steady-state characteristics in the identification of variable frequency equipment with continuously changed states, can still keep better identification results when the load characteristics are similar to and overlapped because a classifier uses a kernel function, and simultaneously, compared with the current commonly used algorithm, the provided identification algorithm has higher load identification rate, higher identification speed and strong practicability.
Drawings
FIG. 1 is a flowchart of a non-intrusive load identification method based on an improved DAG-SVMS provided in the present invention
FIG. 2 is an example of transient current waveform separation
FIG. 3 is a flow chart of the particle swarm optimization algorithm for optimizing classifier parameters
FIG. 4 example of a directed acyclic graph support vector machine recognition model
Detailed description of the invention
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below with reference to the accompanying drawings in the embodiments of the present invention, which are only used for illustration and are not to be construed as limiting the present invention.
Example (b):
a non-intrusive load identification method based on an improved DAG-SVMS, as shown in fig. 1. A public data set BLUED is selected, an exemplary explanation is given to the non-intrusive load identification scheme disclosed by the invention, the data set shows a cycle of power utilization of a certain family in the United states, the power frequency is 60Hz, high-frequency current and voltage data (12kHz) at a bus are provided, two hundred sampling points in one current working period are provided, and a corresponding equipment state conversion list is provided. 5 electric equipment with more switching times are selected from the data set for experimental verification, and are shown in table 1:
TABLE 1 transient sample data composition
After the transient event monitoring module monitors a transient event, the transient event monitoring module performs target load transient current waveform separation, as shown in fig. 2, the feature extraction module performs feature extraction on a target load current waveform, and the transient current waveform further extracted features shown in fig. 2 are shown in table 2:
TABLE 2 load characteristics
When the algorithm is in a training stage, after the characteristics of each power load are extracted, training labels are respectively attached, normalization processing is carried out, a sample database containing the characteristics of each power load is formed, and a load identification model is trained.
And (3) optimizing parameters of the classifier C, g by using a particle swarm algorithm, dividing the classifier into an input part and a cut-off part according to a load zone bit as shown in fig. 3, respectively building a directed acyclic graph model, and optimizing the node sequence by using a Gini index. The directed acyclic graph support vector machine recognition model adopts the 'exclusion' concept, samples flow from top to bottom along with a root node to a leaf node, a branch node decides whether the samples are specifically classified into a left branch or a right branch, and finally the category of the leaf node is the category to which the samples belong. For example, the classification 3 is shown in fig. 4.
Respectively randomly selecting 60% of data samples from 10 transient processes for training, remaining 40% of the data samples for testing, and the execution results are shown in Table 3
TABLE 3 identification results
As can be seen from table 3, the input identification rates of the method for the load transient process are all 100%, and the load shedding identification rate is relatively low, because the duration of the transient process is relatively short when the load is shed, and meanwhile, the transient characteristic discrimination is weakened due to the noise interference in the power line, and the average identification rate of the algorithm is 97.69%, which indicates the effectiveness of the load identification method of the present invention.
TABLE 4 comparison before and after optimization
As can be seen from Table 4, after the classifier parameters are optimized by using the particle swarm optimization, the identification accuracy is improved by 8.91%, and the average identification rate after the node sequence is optimized is improved by 1.04%.
To illustrate the practicability of the algorithm in the text, the BP neural network, the K-NN algorithm and the CART algorithm which are commonly used at present are used for comparing and analyzing with the directed acyclic graph identification algorithm provided by the invention, and the result is shown in Table 4
TABLE 4 identification results
As can be seen from Table 4, the single sample recognition time of the invention is shortest, the real-time performance is strongest, and is only 0.22% of BP neural network and 0.26% of K-NN algorithm, and the average recognition rate is highest. The method has the longest average off-line training time of 20.35s, the off-line training time is acceptable in a certain range in practical application, and the method has the advantages of short on-line identification time, high identification accuracy and reflecting algorithm practicability by considering that the requirements of a system communication module, man-machine interaction and the like on the on-line identification speed of the system are higher.
Finally, the description is as follows: the above embodiments are only for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered thereby.
Claims (8)
1. A non-intrusive load identification method based on an improved DAG-SVMS is characterized by comprising a data acquisition module, an event detection module, a feature extraction module, a load identification module and a data transmission module.
The data acquisition module is used for acquiring total current data at an inlet of the power system, and the acquisition mode is that the input end of the current transformer is connected with a live wire and the output end of the current transformer is subjected to A/D conversion.
The event detection module detects the collected total current data by using an event detection algorithm, and stores the initial time of the transient event after the transient event is detected.
The characteristic extraction module is used for carrying out waveform separation on the transient current waveform after the transient event is detected by the event detection module, acquiring a target load transient waveform causing the transient event, and further extracting transient characteristics for distinguishing different loads.
The load identification module comprises a load identification model training stage and a load real-time identification stage, wherein the load identification model training stage is used for training a load identification model by taking sample library data as support; and in the load identification stage, target load characteristics causing the transient event are extracted in real time and input into an identification model for identification.
And the result data transmission module feeds the transient initial time detected by the event detection module and the identification result of the load identification module back to the user during real-time identification, so that the user can monitor the switching time and the working state of the power load in real time.
2. The improved DAG-SVMS-based non-intrusive load identification method of claim 1, wherein the data acquisition module employs high frequency sampling and A/D conversion on current data at the output terminal of the current transformer to obtain total current data and provide a data source for the event detection module.
3. The method as claimed in claim 1, wherein the event detection module performs the transient event monitoring using an event detection algorithm, where the event detection algorithm is to calculate the current intensity of each period by using the total current data as a unit of period, and when the current intensity of an adjacent period exceeds a threshold value at a certain time, it is determined that a transient event occurs, and the current start time of the second period is recorded as the start time of the transient event. After the transient event is detected, if the current intensity difference value of adjacent continuous periods is smaller than a threshold value, judging that the transient event is ended, and recording the ending moment of the transient event; subtracting the starting time of the transient event from the ending time of the transient event to obtain the duration of the transient event; the transient event duration is an integer multiple of the current duty cycle. After the transient event is detected, the transient event waveform and the current waveform before the transient event are subjected to same phase difference to obtain a target load transient current waveform; the current waveforms are differenced in phase, and transient event current data with the same period length is differenced with transient event pre-steady state current data.
4. The method as claimed in claim 1, wherein the feature extraction module performs feature extraction on a target load current waveform, and the feature extraction includes: and (3) extracting the maximum value, the peak-to-peak value, the average value, the absolute average value, the current intensity, the steady-state waveform, the peak-to-average ratio, the effective value, the form factor, the crest factor and the load zone bit from the transient current waveform.
5. The method as claimed in claim 1, wherein the feature extraction module includes a load recognition model training phase and a load recognition phase. After the characteristics of each electric load are extracted in the load identification model training stage, respectively attaching training labels, normalizing to form a sample database, and dividing the sample into an input type and a cut-off type according to the characteristic mark position to be used as a training set; and after the target load characteristics are acquired in the load identification stage, inputting the characteristics into the trained model for identification.
6. The improved DAG-SVMS-based non-intrusive load identification method of claim 1, wherein the load identification module comprises a stage of training a load identification model, and the specific steps are as follows:
step 1: two types of data samples of the sample database, wherein any two types of loads in each type of sample are combined to be grouped pairwise;
step 2: optimizing a support vector machine error penalty parameter C and a Gaussian radial basis kernel parameter g by using a PSO algorithm, and respectively training a classifier for each group;
and step 3: and (3) respectively building DAG-VMS recognition models for classifiers obtained by training each group of data in the two types of samples, and performing node sequence optimization by using Gini indexes to obtain two recognition models.
7. The improved DAG-SVMS-based non-intrusive load identification method of claim 1, wherein the load identification module further comprises a load identification phase, specifically: and detecting a transient event at an inlet of the power system, extracting target load characteristics, judging whether the transient event is input or cut off according to the load zone bit, and inputting the characteristics into a corresponding recognition model for recognition.
8. The method as claimed in claim 1, wherein the data transmission module sends the transient event start time (load switching time) and the identification result detected by the transient event detection module to the user terminal through WIFI or GPRS during real-time identification, or sends the transient event start time and the identification result to the cloud for the user terminal to access.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110252776.2A CN113036754A (en) | 2021-03-05 | 2021-03-05 | Non-invasive load identification method based on improved DAG-SVMS |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110252776.2A CN113036754A (en) | 2021-03-05 | 2021-03-05 | Non-invasive load identification method based on improved DAG-SVMS |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113036754A true CN113036754A (en) | 2021-06-25 |
Family
ID=76466971
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110252776.2A Pending CN113036754A (en) | 2021-03-05 | 2021-03-05 | Non-invasive load identification method based on improved DAG-SVMS |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113036754A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113610400A (en) * | 2021-08-09 | 2021-11-05 | 山东建筑大学 | Power utilization regulation and control system and regulation and control method |
CN117349639A (en) * | 2023-12-04 | 2024-01-05 | 国网江西省电力有限公司供电服务管理中心 | Identification method and device for charging behavior of electric bicycle |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106250904A (en) * | 2016-05-18 | 2016-12-21 | 国网新疆电力公司电力科学研究院 | Based on Power Disturbance analyser and the sorting technique of improving S-transformation |
CN110288048A (en) * | 2019-07-02 | 2019-09-27 | 东北大学 | A kind of submarine pipeline methods of risk assessment of SVM directed acyclic graph |
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 |
CN110956220A (en) * | 2019-12-11 | 2020-04-03 | 深圳市活力天汇科技股份有限公司 | Non-invasive household appliance load identification method |
CN111722028A (en) * | 2019-03-19 | 2020-09-29 | 华北电力大学 | Load identification method based on high-frequency data |
-
2021
- 2021-03-05 CN CN202110252776.2A patent/CN113036754A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106250904A (en) * | 2016-05-18 | 2016-12-21 | 国网新疆电力公司电力科学研究院 | Based on Power Disturbance analyser and the sorting technique of improving S-transformation |
CN111722028A (en) * | 2019-03-19 | 2020-09-29 | 华北电力大学 | Load identification method based on high-frequency data |
CN110288048A (en) * | 2019-07-02 | 2019-09-27 | 东北大学 | A kind of submarine pipeline methods of risk assessment of SVM directed acyclic graph |
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 |
CN110956220A (en) * | 2019-12-11 | 2020-04-03 | 深圳市活力天汇科技股份有限公司 | Non-invasive household appliance load identification method |
Non-Patent Citations (4)
Title |
---|
JIANMIN LI 等: ""Detection and classification of power quality disturbances using double resolution S-transform and DAGSVMs"", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
MRUTYUNJAYA SAHANI 等: ""FPGA-Based Online Power Quality Disturbances Monitoring Using Reduced-Sample HHT and Class-Specific Weighted RVFLN"", 《IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS》 * |
崔晨 等: ""基于小波包分析和有向无环图支持向量机的电机故障诊断研究"", 《煤矿机械》 * |
曹敏 等: ""基于暂态过程的非侵入式负荷监测研究"", 《水电能源科学》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113610400A (en) * | 2021-08-09 | 2021-11-05 | 山东建筑大学 | Power utilization regulation and control system and regulation and control method |
CN117349639A (en) * | 2023-12-04 | 2024-01-05 | 国网江西省电力有限公司供电服务管理中心 | Identification method and device for charging behavior of electric bicycle |
CN117349639B (en) * | 2023-12-04 | 2024-03-08 | 国网江西省电力有限公司供电服务管理中心 | Identification method and device for charging behavior of electric bicycle |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Veerasamy et al. | LSTM recurrent neural network classifier for high impedance fault detection in solar PV integrated power system | |
CN110956220B (en) | Non-invasive household appliance load identification method | |
CN110879917A (en) | Electric power system transient stability self-adaptive evaluation method based on transfer learning | |
CN112149510B (en) | Non-invasive load detection method | |
CN108616120B (en) | Non-invasive power load decomposition method based on RBF neural network | |
Munshi et al. | Unsupervised nonintrusive extraction of electrical vehicle charging load patterns | |
CN111830347B (en) | Two-stage non-invasive load monitoring method based on event | |
CN113036754A (en) | Non-invasive load identification method based on improved DAG-SVMS | |
CN111628494B (en) | Low-voltage distribution network topology identification method and system based on logistic regression method | |
Jiao et al. | A context-aware multi-event identification method for nonintrusive load monitoring | |
Song et al. | A negative selection algorithm-based identification framework for distribution network faults with high resistance | |
CN112101110A (en) | Non-invasive load identification method for user side of power system | |
Wu et al. | Online detection of events with low-quality synchrophasor measurements based on $ i $ forest | |
CN112067918A (en) | Non-invasive load identification method and system based on event detection | |
CN117559443A (en) | Ordered power utilization control method for large industrial user cluster under peak load | |
Chen et al. | Non-intrusive load monitoring based on feature extraction of change-point and xgboost classifier | |
CN114942344A (en) | Non-invasive electrical appliance identification method, system, medium, equipment and terminal | |
Jain et al. | Current peak based device classification in NILM on a low-cost embedded platform using extra-trees | |
Feng et al. | Identification of disturbance sources based on random forest model | |
CN113010985A (en) | Non-invasive load identification method based on parallel AANN | |
CN111898446A (en) | Single-phase earth fault studying and judging method based on multi-algorithm normalization analysis | |
CN116502149A (en) | Low-voltage power distribution network user-transformation relation identification method and system based on current characteristic conduction | |
CN115564324A (en) | Electric power fingerprint identification method, device and equipment based on event detection | |
CN113158446A (en) | Non-invasive electric load identification method | |
CN114818455A (en) | Power system multi-agent transient state stability judging method and system for small amount of PMU sampling |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20210625 |
|
WD01 | Invention patent application deemed withdrawn after publication |