CN110646708B - 10kV single-core cable early state identification method based on double-layer long-and-short-term memory network - Google Patents
10kV single-core cable early state identification method based on double-layer long-and-short-term memory network Download PDFInfo
- Publication number
- CN110646708B CN110646708B CN201910922549.9A CN201910922549A CN110646708B CN 110646708 B CN110646708 B CN 110646708B CN 201910922549 A CN201910922549 A CN 201910922549A CN 110646708 B CN110646708 B CN 110646708B
- Authority
- CN
- China
- Prior art keywords
- state
- identification
- layer
- double
- stm
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/08—Locating faults in cables, transmission lines, or networks
- G01R31/081—Locating faults in cables, transmission lines, or networks according to type of conductors
- G01R31/086—Locating faults in cables, transmission lines, or networks according to type of conductors in power transmission or distribution networks, i.e. with interconnected conductors
-
- 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
Abstract
The invention discloses a 10kV single-core cable early state identification method based on a double-layer long-short time memory (L STM) network, which is suitable for the use in the electrical field.A 5 current observable electrical quantities are selected from the observable electrical quantities, a time sequence pair is extracted from the 5 observable electrical quantities to construct a cable early state combined time sequence characteristic matrix, then a double-layer L STM network for processing time sequence input is constructed according to the size characteristics of an identification matrix, model training under supervised learning is carried out through an optimization algorithm of a self-adaptive learning rate to obtain a cable early state identification model.
Description
Technical Field
The invention relates to a method for identifying an early state of a 10kV single-core cable, in particular to a method for identifying an early state of a 10kV single-core cable based on a double-layer long-time memory network, which is used in the electrical field.
Background
The utilization rate of the large-section 10kV single-core cable in the power distribution network is higher and higher, but due to the particularity of the cable structure, when an early fault occurs, the cable structure has little influence on the system and is not easy to perceive. However, many times, when a cable fault is fully manifested, a serious complete electrical loop fault has developed, resulting in a protective action or even a power outage. Therefore, the state monitoring work of the 10kV single-core cable is well done, the early-stage fault of the cable is timely identified, and corresponding measures are taken, so that the system stability and the property and personal safety are ensured.
Currently, for the state identification of the cable, the conventional electrical quantity analysis mainly identifies the state of the cable by researching identification information contained in a certain electrical quantity and setting 1 or more threshold values. The following problems exist in the identification by using the traditional electric quantity: the identification of the cable state is realized by only depending on 1 or a few characteristics, and is susceptible to disturbance, so that the accuracy of identification may be reduced. The idea of artificial intelligence is applied to cable state diagnosis, and the following two problems exist: (1) the workload of artificial statistical feature selection is large, whether the statistical feature is properly selected determines the identification accuracy, and the identification accuracy has certain contingency; (2) no temporal relevance of the signals is considered.
Disclosure of Invention
Aiming at the defects of the prior art, the 10kV single-core cable early state identification method based on the double-layer long-time memory network is high in detection efficiency, high in accuracy and high in identification precision, accurate identification of the cable early state is achieved, and accidental identification results are avoided.
In order to achieve the technical purpose, the method for identifying the early state of the 10kV single-core cable based on the double-layer long-and-short-term memory network comprises the following steps:
s1, selecting 5 observable electric quantities of 10kV single-core cable current from the plurality of observable electric quantities of the cable;
s2, selecting a time window containing a state change point and front and rear state points thereof to intercept the state change signal according to the transient process duration time of the signal of the observable electric quantity of the 10kV single-core cable, and storing the state change signals in all the observable electric quantities of each 10kV single-core cable in groups;
s3, extracting time sequence pairs from the state change signals in each group of observable electric quantity to construct a combined time sequence characteristic identification matrix X, forming identification samples, and obtaining a plurality of identification samples by using a plurality of groups of state change signals;
s4, performing z-score normalization (z-score normalization) on the identification sample characteristics, and then adding a state label to the identification sample by adopting a thermal coding method.
S5, randomly disordering all the labeled identification samples, taking 80% of the labeled identification samples as a training sample group, and taking the remaining 20% as a test sample group;
s6, constructing a double-layer L STM network model according to the size characteristics of the combined time sequence characteristic identification matrix X;
s7, carrying out supervised model training on the constructed double-layer L STM network model by utilizing a training sample group, thereby obtaining a 10kV single-core cable early state identification model;
and S8, placing the test sample group into a trained 10kV single-core cable early state identification model for testing to obtain a classification result, counting the identification accuracy, and completing the identification of the cable early state.
In the step S1, the observable electric quantities of 5 10kV single-core cables are obtained from the plurality of observable electric quantities of the cables and are respectively the state-change phase current ILFirst end of sheath current IssEnd sheath current IseAnd other two-phase current IO1And IO2And storing the 5 observable electrical quantities reflecting the cable state as a group for state identification.
In step S2, a time window including an observable electrical quantity state change point and its preceding and following state points is selected according to the signal transient process duration to intercept the state change signal; the time window is determined experimentally from multiple comparison trials.
In step S3, the cable status at time t can be represented by ctIndicating that the state-change phase current, the ground lead head end current, the ground lead tail end current, and the other two-phase current timing pairs reflecting the cable state at this time can be expressed as
A sub-sample can be formed
The subsample reflects the instantaneous corresponding relation between the observable electric quantity and the cable state, and because each state point is not isolated, the correlation relation on the time sequence exists among the state points, and the occurrence of the state at the time t is influenced by the state at the previous time:
a plurality of continuous state points form a state process, namely a plurality of subsamples of the continuous state points form a main sample SiIt may be represented by the following formula:
the main sample represents the corresponding relation between the observable electric quantity and the cable state change process on the time sequence; wherein n denotes the length of the time window, hiIs a label, i.e., the change state of the cable, the combined timing characteristics matrix X can be represented as follows:
in the step S6, according to the characteristics of the combined timing feature matrix, a dual-layer L STM network composed of 2 hidden layers is constructed, where the 2 hidden layers include L STM-1 representing a first hidden layer and L STM-2 representing a second hidden layer;
timing signal x ═ x1,x2,…,xt,…xn]Inputting L STM-1 layer, and performing L STM standard module series operation to obtain output result
L STResults of M-1 layer h(1)Sending the data as input to L STM-2 layer, calculating to obtain output result
Finally, obtaining a classification result h through a softmax functionout；
The double-layer L STM network model parameter solving adopts a cross entropy function as a loss function, and is defined as follows:
where W is the weight of the double layer L STM network, b is the bias of the double layer L STM network, (y ═ k) is the actual value in the training sample belonging to label k, hout(k) Representing the prediction probability.
Has the advantages that:
the method expands the original analysis object from single electric quantity to multi-correlation electric quantity, expands the original input dimensionality, does not extract statistical characteristics, but utilizes the original data to construct a combined time sequence characteristic matrix, reduces the workload of a data processing stage and avoids the problem of poor identification caused by improper statistical characteristics; the method takes continuous time sequence pairs in a time window as input, takes relevance of the input on time into consideration, greatly improves reliability of the identification method, and has the identification precision of the early state of the cable up to 99.06%.
Drawings
FIG. 1 is a flow chart of a method for identifying the early state of a 10kV cable
FIG. 2 is a graph showing the amount of electricity that can be observed
FIG. 3 is an expanded view of a double-layer L STM network structure
FIG. 4 is L STM standard module
FIG. 5 is a 10kV unbranched simulation system diagram
FIG. 6 shows the identification result and the average identification accuracy of 5 experiments
Detailed Description
The embodiments of the present invention will be given below with reference to the accompanying drawings, and the technical solutions of the present invention will be further clearly and completely described by the embodiments. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention.
As shown in fig. 1, the method for identifying the early state of the 10kV single-core cable based on the double-layer long-and-short time memory network of the invention comprises the following steps:
s1, as shown in figure 2, selecting 5 observable electric quantities of 10kV single-core cable current from the plurality of observable electric quantities of the cable;
a10 kV system model is built in PSCAD/EMTDC software to perform simulation experiments on the early states of the four cables so as to obtain sample data required by the method. Simulation model as shown in fig. 5, the cable early fault sets 4 early states as follows: (1) at a certain point in the early stage of the cable, the insulation is slightly degraded; (2) the development is continued, so that the external insulation is damaged, and the metal protective layer is grounded; (3) the metal protective layer is grounded due to the damage of the outer insulating sheath caused by external force; (4) and (4) disturbance of sudden load change. In the simulation, the length of the cable is set to 1km, and the sampling frequency is 100 kHz;
obtaining 5 observable electric quantities of 10kV single-core cable from a plurality of observable electric quantities of cable, wherein the observable electric quantities are respectively state-change phase current ILFirst end of sheath current IssEnd sheath current IseAnd other two-phase current IO1And IO2Storing the 5 observable electrical quantities reflecting the cable state as a group for state identification;
s2, selecting a time window containing a state change point and front and rear state points thereof to intercept the state change signal according to the transient process duration time of the signal of the observable electric quantity of the 10kV single-core cable, and storing the state change signals in all the observable electric quantities of each 10kV single-core cable in groups; selecting a time window containing an observable electric quantity state change point and front and rear state points thereof according to the signal transient process duration to intercept a state change signal; according to multiple comparison trial experiments, specifically, 10 sample points before a mutation point and 50 sample points after the mutation point are selected, and 61 points in total are taken as a time window;
s3, extracting time sequence pairs from the state change signals in each group of observable electric quantity to construct a combined time sequence characteristic identification matrix X, forming identification samples, and obtaining a plurality of identification samples by using a plurality of groups of state change signals;
specifically, the method comprises the following steps: the cable state at time t may be represented by ctIndicating that the state-change phase current, the ground lead head end current, the ground lead tail end current, and the other two-phase current timing pairs reflecting the cable state at this time can be expressed as
A sub-sample can be formed
The subsample reflects the instantaneous corresponding relation between the observable electric quantity and the cable state, and because each state point is not isolated, the correlation relation on the time sequence exists among the state points, and the occurrence of the state at the time t is influenced by the state at the previous time:
a plurality of continuous state points form a state process, namely a plurality of subsamples of the continuous state points form a main sample SiIt may be represented by the following formula:
the main sample represents the corresponding relation between the observable electric quantity and the cable state change process on the time sequence; wherein n denotes the length of the time window, hiIs a label, i.e., the change state of the cable, the combined timing characteristics matrix X can be represented as follows:
s4, performing z-score normalization (z-score normalization) on the characteristics of the identification samples, and then adding a state label to the identification samples by adopting a thermal coding method to divide all the samples into 4 types: 1000, 0100, 0010, 0001;
s5, randomly disordering all the labeled identification samples, taking 80% of the labeled identification samples as a training sample group, and taking the remaining 20% as a test sample group; specifically, after the steps 1 to 4, 7037 total samples are obtained, then 7037 samples are randomly disordered, and the sample data are distributed as shown in the following table:
s6, constructing a double-layer L STM network model according to the size characteristics of the combined time sequence characteristic identification matrix X;
an expansion diagram of a double-layer L STM network model structure is shown in FIG. 3, a double-layer L STM network is composed of 2 hidden layers, the 2 hidden layers comprise L STM-1 representing a first hidden layer, L STM-2 representing a second hidden layer, and a L STM standard module comprises a forgetting gate, an input gate and an output gate, and is shown in FIG. 4, wherein C in the diagramt-1Is the cell state at time t-1, ht-1Is the cell output at time t-1, htCell output at time t, xtCell input at time t, ftTo forget the door, itIs an input gate otSigma is an activation function sigmoid, tanh is an activation function tanh, and,
Is a candidate value vector;
timing signal x ═ x1,x2,…,xt,…xn]Inputting L STM-1 layer, and performing L STM standard module series operation to obtain output result
The result h of L STM-1 layer is then added(1)Sending the data as input to L STM-2 layer, calculating to obtain output result
Finally, obtaining a classification result h through a softmax functionout；
The parameters of the double layer L STM network structure are set as follows:
wherein, time step n is sample time window length, and the number of hidden layer nodes refers to the number of nodes in forgetting gate, input gate, output gate in the L STM unit, and double-deck L STM network model parameter solution adopts the cross entropy function as loss function, defines as follows:
wherein W ═ Wf,Wi,Wc,Wo]Is the weight of the two-layer L STM network, b ═ bf,bi,bc,bo]Is the bias of the two-layer L STM network, (y ═ k) is the actual value in the training sample that belongs to label k, hout(k) Representing a prediction probability;
s7, carrying out supervised model training on the constructed double-layer L STM network model by utilizing a training sample group, thereby obtaining a 10kV single-core cable early state identification model;
and S8, placing the test sample group into a trained 10kV single-core cable early state identification model for testing to obtain a classification result, counting the identification accuracy, and completing the identification of the cable early state.
For the neural network model with the same structure, the identification accuracy rate is different due to different structure initialization parameter settings, but after training iteration, the final identification result oscillates within a certain cell, and the identification accuracy rate approaches to a certain value. By using the constructed sample data, classification experiments are carried out on the early states of the 4 cables, the iteration number of model training is 5000, and the identification result and the average identification accuracy of 5 experiments are shown in fig. 6. The average identification accuracy of the proposed method is 99.06%.
Claims (5)
1. A10 kV single-core cable early state identification method based on a double-layer long-and-short time memory network is characterized by comprising the following steps:
s1, selecting 5 observable electric quantities of 10kV single-core cable current from the plurality of observable electric quantities of the cable;
s2, selecting a time window containing a state change point and front and rear state points thereof to intercept the state change signal according to the transient process duration time of the signal of the observable electric quantity of the 10kV single-core cable, and storing the state change signals in all the observable electric quantities of each 10kV single-core cable in groups;
s3, extracting time sequence pairs from the state change signals in each group of observable electric quantity to construct a combined time sequence characteristic identification matrix X, forming identification samples, and obtaining a plurality of identification samples by using a plurality of groups of state change signals;
s4, performing z-score normalization (z-score normalization) on the characteristics of the identification sample, and then adding a state label to the identification sample by adopting a thermal coding method;
s5, randomly disordering all the labeled identification samples, taking 80% of the labeled identification samples as a training sample group, and taking the remaining 20% as a test sample group;
s6, constructing a double-layer L STM network model according to the size characteristics of the combined time sequence characteristic identification matrix X;
s7, carrying out supervised model training on the constructed double-layer L STM network model by utilizing a training sample group, thereby obtaining a 10kV single-core cable early state identification model;
and S8, placing the test sample group into a trained 10kV single-core cable early state identification model for testing to obtain a classification result, counting the identification accuracy, and completing the identification of the cable early state.
2. The 10kV single-core cable early state identification method based on the double-layer long-and-short-term memory network according to claim 1: in the step S1, the observable electric quantities of 5 10kV single-core cables are obtained from the plurality of observable electric quantities of the cables and are respectively the state-change phase current ILFirst end of sheath current IssEnd sheath current IseAnd other two-phase current IO1And IO2The 5 pieces are reflected on the cable stateThe observable electrical quantities are stored as a group for state identification.
3. The 10kV single-core cable early state identification method based on the double-layer long-and-short-term memory network according to claim 1: in step S2, a time window including an observable electrical quantity state change point and its preceding and following state points is selected according to the signal transient process duration to intercept the state change signal; the time window is determined experimentally from multiple comparison trials.
4. The 10kV single-core cable early state identification method based on the double-layer long-and-short-term memory network according to claim 1: in step S3, the cable status at time t can be represented by ctIndicating that the state-change phase current, the ground lead head end current, the ground lead tail end current, and the other two-phase current timing pairs reflecting the cable state at this time can be expressed as
A sub-sample can be formed
The subsample reflects the instantaneous corresponding relation between the observable electric quantity and the cable state, and because each state point is not isolated, the correlation relation on the time sequence exists among the state points, and the occurrence of the state at the time t is influenced by the state at the previous time:
a plurality of continuous state points form a state process, namely a plurality of subsamples of the continuous state points form a main sample SiIt may be represented by the following formula:
the main sample represents the corresponding relation between the observable electric quantity and the cable state change process on the time sequence; wherein n denotes the length of the time window, hiIs a label, i.e. a change of state of the cableThe combined timing characteristics matrix X can be expressed as follows:
5. the method for identifying the early state of the 10kV single-core cable based on the double-layer long-and-short-term memory network as claimed in claim 1, wherein in the step S6, according to the characteristics of the combined time sequence feature matrix, a double-layer L STM network consisting of 2 hidden layers is constructed, wherein the 2 hidden layers comprise a L STM-1 hidden layer and a L STM-2 hidden layer;
timing signal x ═ x1，x2，…，xt，…xn]Inputting L STM-1 layer, and performing L STM standard module series operation to obtain output result
The result h of L STM-1 layer is then added(1)Sending the data as input to L STM-2 layer, calculating to obtain output result
Finally, obtaining a classification result h through a softmax functionout；
The double-layer L STM network model parameter solving adopts a cross entropy function as a loss function, and is defined as follows:
where W is the weight of the double layer L STM network, b is the bias of the double layer L STM network, (y ═ k) is the actual value in the training sample belonging to label k, hout(k) Representing the prediction probability.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910922549.9A CN110646708B (en) | 2019-09-27 | 2019-09-27 | 10kV single-core cable early state identification method based on double-layer long-and-short-term memory network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910922549.9A CN110646708B (en) | 2019-09-27 | 2019-09-27 | 10kV single-core cable early state identification method based on double-layer long-and-short-term memory network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110646708A CN110646708A (en) | 2020-01-03 |
| CN110646708B true CN110646708B (en) | 2020-07-17 |
Family
ID=69011522
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910922549.9A Active CN110646708B (en) | 2019-09-27 | 2019-09-27 | 10kV single-core cable early state identification method based on double-layer long-and-short-term memory network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110646708B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111208392B (en) * | 2020-04-16 | 2020-07-31 | 武汉格蓝若智能技术有限公司 | Label-free cable collection trend identification method |
| CN112070322B (en) * | 2020-09-28 | 2022-05-13 | 国网河北省电力有限公司雄安新区供电公司 | High-voltage cable line running state prediction method based on long-short term memory network |
| CN113092933A (en) * | 2021-03-23 | 2021-07-09 | 武汉大学 | LSTM-based single-phase earth fault line selection method and system |
| CN117289075B (en) * | 2023-09-07 | 2024-03-12 | 东莞市明盛电气有限公司 | Cable performance detection method and system |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1896756A (en) * | 2006-03-16 | 2007-01-17 | 重庆大学 | Cable-fault on-line distance measurement based on artifical nerve network mode |
| CN102981104A (en) * | 2012-11-19 | 2013-03-20 | 中国能源建设集团广东省电力设计研究院 | On-line monitoring method for submarine cables |
| CN104360227A (en) * | 2014-10-31 | 2015-02-18 | 国家电网公司 | Substation cable outlet fault monitoring method based on traveling wave method and transient basic frequency method |
| CN106251059A (en) * | 2016-07-27 | 2016-12-21 | 中国电力科学研究院 | A kind of cable status appraisal procedure based on probabilistic neural network algorithm |
| CN106599777A (en) * | 2016-11-02 | 2017-04-26 | 华南理工大学 | Cable partial discharge signal identification method based on energy percentage |
| 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 |
| CN109459671A (en) * | 2018-09-27 | 2019-03-12 | 国网浙江省电力有限公司湖州供电公司 | A kind of switch cabinet state monitoring method of the improvement neural network based on genetic algorithm |
| CN109738776A (en) * | 2019-01-02 | 2019-05-10 | 华南理工大学 | Fan converter open-circuit fault recognition methods based on LSTM |
| CN110068748A (en) * | 2019-04-26 | 2019-07-30 | 中国矿业大学 | 10kV single-core cable early stage state identification method based on depth convolutional neural networks |
| CN110244205A (en) * | 2019-07-12 | 2019-09-17 | 西南石油大学 | Cable accessory Recognition of Partial Discharge based on convolutional neural networks under a kind of small data quantity |
-
2019
- 2019-09-27 CN CN201910922549.9A patent/CN110646708B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1896756A (en) * | 2006-03-16 | 2007-01-17 | 重庆大学 | Cable-fault on-line distance measurement based on artifical nerve network mode |
| CN102981104A (en) * | 2012-11-19 | 2013-03-20 | 中国能源建设集团广东省电力设计研究院 | On-line monitoring method for submarine cables |
| CN104360227A (en) * | 2014-10-31 | 2015-02-18 | 国家电网公司 | Substation cable outlet fault monitoring method based on traveling wave method and transient basic frequency method |
| CN106251059A (en) * | 2016-07-27 | 2016-12-21 | 中国电力科学研究院 | A kind of cable status appraisal procedure based on probabilistic neural network algorithm |
| CN106599777A (en) * | 2016-11-02 | 2017-04-26 | 华南理工大学 | Cable partial discharge signal identification method based on energy percentage |
| 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 |
| CN109459671A (en) * | 2018-09-27 | 2019-03-12 | 国网浙江省电力有限公司湖州供电公司 | A kind of switch cabinet state monitoring method of the improvement neural network based on genetic algorithm |
| CN109738776A (en) * | 2019-01-02 | 2019-05-10 | 华南理工大学 | Fan converter open-circuit fault recognition methods based on LSTM |
| CN110068748A (en) * | 2019-04-26 | 2019-07-30 | 中国矿业大学 | 10kV single-core cable early stage state identification method based on depth convolutional neural networks |
| CN110244205A (en) * | 2019-07-12 | 2019-09-17 | 西南石油大学 | Cable accessory Recognition of Partial Discharge based on convolutional neural networks under a kind of small data quantity |
Non-Patent Citations (3)
| Title |
|---|
| 10kv地下电缆早期故障检测与识别方法探讨;戴铭;《中国优秀硕士学位论文全文数据库工程科技Ⅱ辑》;20121015;全文 * |
| 考虑金属护层结构的电力电缆单端故障识别方法;孙建明;《电力科学与技术学报》;20171231;全文 * |
| 采用 LSTM 网络的电力变压器运行状态预测方法研究;代杰杰;《高压电技术》;20180430;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110646708A (en) | 2020-01-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110646708B (en) | 10kV single-core cable early state identification method based on double-layer long-and-short-term memory network | |
| CN111274395B (en) | Power grid monitoring alarm event identification method based on convolution and long-short term memory network | |
| CN109635928B (en) | Voltage sag reason identification method based on deep learning model fusion | |
| WO2020191800A1 (en) | Method for predicting remaining service life of lithium-ion battery employing wde-optimized lstm network | |
| CN111157899B (en) | Method for estimating SOC of battery based on model fusion idea | |
| CN108020781B (en) | Breaker fault diagnosis method | |
| CN109375116B (en) | Battery system abnormal battery identification method based on self-encoder | |
| CN110879373B (en) | Oil-immersed transformer fault diagnosis method with neural network and decision fusion | |
| CN116840720A (en) | Fuel cell remaining life prediction method | |
| CN116381517A (en) | Lithium battery remaining life probability prediction method based on time convolution attention mechanism | |
| Zhang et al. | Transient stability prediction of power systems based on deep belief networks | |
| Zhu et al. | Integrated data-driven power system transient stability monitoring and enhancement | |
| CN114460481A (en) | Energy storage battery thermal runaway early warning method based on Bi-LSTM and attention mechanism | |
| Xu et al. | An improved ELM-WOA–based fault diagnosis for electric power | |
| CN116310551A (en) | Power transformer fault diagnosis method and system of multi-mode information fusion network | |
| He et al. | Partial discharge pattern recognition algorithm based on sparse self-coding and extreme learning machine | |
| Zhao et al. | A new transient voltage stability prediction model using big data analysis | |
| CN115684835B (en) | Power distribution network fault diagnosis method, device, equipment and storage medium | |
| Bai et al. | Abnormal Detection Scheme of Substation Equipment based on Intelligent Fusion Terminal | |
| Kim et al. | Cyberattack Detection in Smart Grids based on Reservoir Computing | |
| Song et al. | An improved convolutional neural network-based approach for short-term wind speed forecast | |
| CN115616437B (en) | Intelligent debugging control method and system for EPC special power supply | |
| Chen et al. | CNN-Informer-Based Remaining Useful Life Prediction for Electrical Devices | |
| Zhang et al. | Grid monitoring alarm event recognition method integrating expert rules and deep learning | |
| Li et al. | Fault Diagnosis Method of Smart Substation Secondary System Based on Deep Reinforcement Learning |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |