CN117077065B - High-voltage direct-current transmission grounding electrode detection and early warning method and system - Google Patents
High-voltage direct-current transmission grounding electrode detection and early warning method and system Download PDFInfo
- Publication number
- CN117077065B CN117077065B CN202311330062.4A CN202311330062A CN117077065B CN 117077065 B CN117077065 B CN 117077065B CN 202311330062 A CN202311330062 A CN 202311330062A CN 117077065 B CN117077065 B CN 117077065B
- Authority
- CN
- China
- Prior art keywords
- grounding electrode
- detection
- early warning
- training data
- training
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 291
- 238000000034 method Methods 0.000 title claims abstract description 69
- 230000005540 biological transmission Effects 0.000 title claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 191
- 238000010801 machine learning Methods 0.000 claims abstract description 93
- 230000005856 abnormality Effects 0.000 claims abstract 5
- 230000002159 abnormal effect Effects 0.000 claims description 73
- 239000013598 vector Substances 0.000 claims description 38
- 230000000737 periodic effect Effects 0.000 claims description 37
- 230000008569 process Effects 0.000 claims description 13
- 238000004364 calculation method Methods 0.000 claims description 12
- 238000013480 data collection Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 6
- 238000012545 processing Methods 0.000 claims description 4
- 230000015654 memory Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 5
- 239000002689 soil Substances 0.000 description 4
- 238000005260 corrosion Methods 0.000 description 3
- 230000007797 corrosion Effects 0.000 description 3
- 238000006073 displacement reaction Methods 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 239000003990 capacitor Substances 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000001028 reflection method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2433—Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- 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
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E60/00—Enabling technologies; Technologies with a potential or indirect contribution to GHG emissions mitigation
- Y02E60/60—Arrangements for transfer of electric power between AC networks or generators via a high voltage DC link [HVCD]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Computational Mathematics (AREA)
- Algebra (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Emergency Alarm Devices (AREA)
Abstract
The invention discloses a high-voltage direct-current transmission grounding electrode detection and early warning method and system, which relate to the technical field of grounding electrode detection early warning, and are characterized in that historical early warning training data are collected in advance, a first machine learning model for predicting the grounding electrode abnormality probability and a second machine learning model for predicting the detection cost are trained, early warning training data are collected, the predicted grounding electrode abnormality probability is obtained based on the early warning training data and the first machine learning model, whether early warning is needed or not is judged based on the predicted grounding electrode abnormality probability, the predicted detection cost is obtained when early warning is not needed, and a decision result for carrying out second updating on an initial detection period is output by using an Actor-Critic network model based on the initial detection period, the predicted grounding electrode abnormality probability and the predicted detection cost; the detection cost is reduced, and the detection efficiency is improved.
Description
Technical Field
The invention relates to the technical field of grounding electrode detection and early warning, in particular to a high-voltage direct-current transmission grounding electrode detection and early warning method and system.
Background
In high voltage direct current transmission (HVDC) systems, a Ground Electrode (Ground Electrode) is the place to provide a current loop to ensure safe operation of the system and balance of current. In HVDC systems, high voltage direct current needs to be returned to the source and not be able to flow back through the earth as alternating current. The ground electrode is to realize such a current loop.
The grounding electrode in HVDC systems is typically located at an end station of the power transmission line, i.e. where the power transmission line is connected to an ac or dc grid. The grounding electrode is contacted with the underground layer through the underground electrode rod and releases current into the underground to form a safe loop. In this way, the high voltage direct current can be safely returned to the source end, and the balance of power transmission is maintained.
Thus, as an important part of HVDC, it is a very important task to detect the operating state of the earth electrode; however, because the grounding electrode has larger corrosiveness and the step voltage near the grounding electrode address brings personal safety problems, the grounding electrode is generally arranged at a more remote position, and therefore, larger manpower and time cost are required for detecting the grounding electrode every time;
the current mode of detecting the grounding electrode mainly comprises periodic detection, but lacks a balance method for monitoring the running state of the grounding electrode in real time and dynamically adjusting the detection period of the grounding electrode based on the running state and the detection cost of the grounding electrode;
the Chinese patent with the authorized bulletin number of CN106443329B discloses a system and a method for detecting faults of a direct current grounding electrode line, wherein the fault discrimination and the fault ranging of the direct current grounding electrode line are realized by using a time domain reflection method; the fault detection device sends out a common mode pulse signal and a differential mode pulse signal, and the common mode pulse signal and the differential mode pulse signal are injected to two leads of a direct current grounding electrode line through the impedance matching module and the high-voltage capacitor, and fault reflection traveling waves are detected through the loops. A high-voltage reactor is arranged at the split position of the grounding electrode line to block the interference of the direct-current field on common-mode signal traveling waves; however, the method fails to solve the balance problem between the detection cost and the detection period;
therefore, the invention provides a high-voltage direct-current transmission grounding electrode detection and early warning method and system.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems existing in the prior art. Therefore, the invention provides the high-voltage direct-current transmission grounding electrode detection and early warning method and system, which reduce the detection cost and improve the detection efficiency.
In order to achieve the above objective, embodiment 1 of the present invention provides a method for detecting and early warning a hvth power transmission grounding electrode, comprising the following steps:
step one: the method comprises the steps of collecting historical early warning training data in a periodic detection history in advance, and training a first machine learning model for predicting the abnormal probability of the grounding electrode and a second machine learning model for predicting the detection cost based on the historical early warning training data;
step two: acquiring a detection period in the periodic detection history as an initial detection period;
in the non-periodic detection process, early warning training data are collected, the predicted abnormal probability of the grounding electrode is obtained based on the early warning training data and the first machine learning model, and whether early warning is needed or not is judged based on the predicted abnormal probability of the grounding electrode;
if the early warning is needed, the early warning is carried out, and the initial detection period is updated for the first time;
if early warning is not needed, based on early warning training data and a second machine learning model, obtaining predicted detection cost, based on an initial detection period, predicted abnormal grounding electrode probability and predicted detection cost, using an Actor network output in an Actor-Critic network model to carry out a decision result of second updating on the initial detection period, and training the Actor-Critic network model;
step three: judging whether to detect the grounding electrode or not based on the initial detection period;
the periodic detection history is a period of time for periodic detection of the grounding electrode in the past high-voltage direct-current transmission process;
the historical early warning training data comprise ground electrode abnormal training data and detection cost training data;
the grounding electrode abnormal training data comprises collected grounding electrode training feature vectors and corresponding detection result labels determined by detection personnel when periodic detection is carried out each time;
wherein, the elements in the grounding electrode training feature vector comprise physical feature quantities of the grounding electrode, which influence the detection result;
the detection result label is the result of each periodic detection of a detection person, and is one of 0 or 1;
the detection cost training data are collected grounding electrode training feature vectors and corresponding detection cost labels determined by detection personnel;
the detection cost label is the comprehensive cost required to be spent for each detection;
the calculation mode of the comprehensive cost is as follows:
marking the time length of the detection personnel going to the grounding electrode as t1, and marking the time length of the detection personnel detecting the grounding electrode as t2;
the total cost spent in going to the ground and in the detection process is marked as；
The comprehensive cost is marked as H, and the calculation formula of the comprehensive cost H isThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>、/>Respectively preset proportional coefficients;
the first machine learning model for predicting the abnormal probability of the grounding electrode is trained in the following way:
taking each group of grounding electrode training feature vectors in the grounding electrode abnormal training data as input of a first machine learning model, wherein the first machine learning model takes the predicted grounding electrode abnormal probability of each group of grounding electrode training feature vectors as output, takes a detection result label corresponding to the grounding electrode training feature vectors in the grounding electrode abnormal training data as a prediction target, and takes the sum of the prediction errors between the minimum grounding electrode abnormal probability and the detection result label as a training target; training the first machine learning model until the sum of the prediction errors reaches convergence, and stopping training to obtain a first machine learning model which outputs predicted abnormal grounding electrode probability according to the grounding electrode training feature vector; the first machine learning model is any one of a polynomial regression model or an SVR model;
the second machine learning model that predicts the cost of detection is trained in the following manner:
taking each group of grounding electrode training feature vectors in the detection cost training data as input of a second machine learning model, wherein the second machine learning model takes the predicted detection cost labels of each group of grounding electrode training feature vectors as output, takes the detection cost labels corresponding to the grounding electrode training feature vectors in the detection cost training data as prediction targets, and takes the sum of all prediction errors of the detection cost labels as a training target; training the second machine learning model until the sum of the prediction errors reaches convergence, and stopping training to obtain the second machine learning model which outputs the predicted detection cost label according to the grounding electrode training feature vector; the second machine learning model is any one of a polynomial regression model or an SVR model;
the mode of collecting early warning training data is as follows:
presetting a sample collection period, collecting each physical characteristic quantity corresponding to each element in a grounding electrode training characteristic vector of a grounding electrode every sample collection period, and forming each physical characteristic quantity into early warning training data;
the predicted ground electrode anomaly probability is obtained by:
taking the early warning training data as the input of a first machine learning model, and obtaining the predicted abnormal probability of the grounding electrode output by the first machine learning model;
based on the predicted abnormal probability of the grounding electrode, judging whether the early warning is needed or not is as follows:
presetting an abnormal probability threshold, and judging that early warning is needed if the predicted abnormal probability of the grounding electrode is larger than the abnormal probability threshold;
the first update of the initial detection period is performed in the following manner:
updating the initial detection period to be: starting from the last detection end to the time when the early warning is judged to be needed;
the method for obtaining the predicted detection cost comprises the following steps:
taking the early warning training data as the input of a second machine learning model, and obtaining a predicted detection cost label output by the second machine learning model;
the method for outputting a decision result for carrying out second updating on the initial detection period by using an Actor network in the Actor-Critic network model and training the Actor-Critic network model comprises the following steps:
initializing parameters of an Actor network and a Critic network;
when the early warning is judged not to be needed, the following steps are executed:
step 11: taking early warning training data corresponding to the current sample collection period as a current state;
step 12: the Actor network outputs a decision result of a second update of the initial detection period;
taking early warning training data of a next round of sample collection period as a next state;
step 13: calculating an actual rewarding value Q; the actual rewarding value Q is obtained after making a second updating decision on the initial detection period each time;
the calculation mode of the reward value Q is as follows:
marking the value of the initial detection period before the second update as R0, and marking the value of the initial detection period after the second update as R;
marking the predicted abnormal probability of the grounding electrode as P, and marking the predicted detection cost label as D;
the calculation formula of the reward value Q isThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>And->Respectively preset proportional coefficients;
step 14: updating the value of the prize value function using an update formula of the Critic network to adjust the estimate of the actual prize value Q for the decision result;
step 15: updating parameters of the Actor network by using an updating formula of the Actor network so as to improve the probability of selecting a high rewarding decision result under a given current state;
based on the initial detection period, the mode of judging whether to detect the grounding electrode is as follows:
marking the time length from the last detection end to the current time as T;
if the duration T is smaller than the initial detection period, not processing;
if the duration T is greater than or equal to the initial detection period, notifying a detector to go to the grounding electrode for detection.
According to the embodiment 2 of the invention, the detection and early warning system for the high-voltage direct-current transmission grounding electrode comprises a training data collection module, a model training module, an early warning training data collection module and a detection period decision module; wherein, each module is electrically connected with each other;
the training data collection module is used for pre-collecting historical early warning training data in the periodic detection history and sending the historical early warning training data to the model training module;
the model training module is used for training a first machine learning model for predicting the abnormal probability of the grounding electrode and a second machine learning model for predicting the detection cost based on the historical early warning training data, and sending the first machine learning model and the second machine learning model to the detection period decision module;
the early warning training data collection module is used for obtaining a detection period in the periodic detection history as an initial detection period, collecting early warning training data in the non-periodic detection process and sending the initial detection period and the early warning training data to the detection period decision module;
the detection period decision module is used for updating the initial detection period based on the early warning training data, the first machine learning model and the second machine learning model and judging whether to detect the grounding electrode or not based on the initial detection period;
the method for updating the initial detection period comprises the following steps:
based on early warning training data and a first machine learning model, obtaining predicted abnormal probabilities of the grounding electrode, and judging whether early warning is needed or not based on the predicted abnormal probabilities of the grounding electrode;
if the early warning is needed, the early warning is carried out, and the initial detection period is updated for the first time;
if the early warning is not needed, based on early warning training data and a second machine learning model, the predicted detection cost is obtained, based on the initial detection period, the predicted abnormal probability of the grounding electrode and the predicted detection cost, an Actor network output in the Actor-Critic network model is used for carrying out a decision result of second updating on the initial detection period, and the Actor-Critic network model is trained.
Compared with the prior art, the invention has the beneficial effects that:
according to the method, historical early warning training data in a periodic detection history are collected in advance, a first machine learning model for predicting the abnormal probability of the grounding electrode is trained based on the historical early warning training data, a second machine learning model for predicting the detection cost is used for acquiring a detection period in the periodic detection history as an initial detection period, early warning training data is collected in an unscheduled detection process, the early warning is obtained based on the early warning training data and the first machine learning model, whether early warning is needed or not is judged based on the predicted abnormal probability of the grounding electrode, when early warning is not needed, the predicted detection cost is obtained based on the early warning training data and the second machine learning model, a decision result for second updating the initial detection period is output based on an Actor-Critic network model, training is performed on the Actor-Critic network model, and whether the grounding electrode is detected or not is judged based on the initial detection period; therefore, the probability that the grounding electrode needs to be overhauled is automatically analyzed according to the physical characteristic quantities of the grounding electrode, and the detection period duration is dynamically adjusted according to the overhauling probability, so that the detection cost is reduced and the detection efficiency is improved on the basis of guaranteeing the detection safety of the grounding electrode.
Drawings
Fig. 1 is a flowchart of a method for detecting and early warning a hvth power transmission grounding electrode in embodiment 1 of the invention;
fig. 2 is a module connection relation diagram of a detection and early warning system for a hvth grounding electrode in embodiment 2 of the invention;
fig. 3 is a schematic structural diagram of an electronic device in embodiment 3 of the present invention;
fig. 4 is a schematic diagram of a computer-readable storage medium according to embodiment 4 of the present invention.
Detailed Description
The technical solutions of the present invention will be clearly and completely described in connection with the embodiments, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
As shown in fig. 1, the method for detecting and early warning the grounding electrode of the high-voltage direct-current transmission comprises the following steps:
step one: the method comprises the steps of collecting historical early warning training data in a periodic detection history in advance, and training a first machine learning model for predicting the abnormal probability of the grounding electrode and a second machine learning model for predicting the detection cost based on the historical early warning training data;
step two: acquiring a detection period in the periodic detection history as an initial detection period;
in the non-periodic detection process, early warning training data are collected, the predicted abnormal probability of the grounding electrode is obtained based on the early warning training data and the first machine learning model, and whether early warning is needed or not is judged based on the predicted abnormal probability of the grounding electrode;
if the early warning is needed, the early warning is carried out, and the initial detection period is updated for the first time;
if early warning is not needed, based on early warning training data and a second machine learning model, obtaining predicted detection cost, based on an initial detection period, predicted abnormal grounding electrode probability and predicted detection cost, using an Actor network output in an Actor-Critic network model to carry out a decision result of second updating on the initial detection period, and training the Actor-Critic network model;
step three: judging whether to detect the grounding electrode or not based on the initial detection period;
the periodic detection history is a period of time for periodic detection of the grounding electrode in the past high-voltage direct-current transmission process; it can be understood that, for each hvth grounding electrode, after the establishment is completed, the historical early warning training data required by the embodiment can be collected by performing periodic detection for several times;
specifically, the historical early warning training data comprises ground electrode abnormal training data and detection cost training data;
the grounding electrode abnormal training data comprises collected grounding electrode training feature vectors and corresponding detection result labels determined by detection personnel when periodic detection is carried out each time;
wherein, the elements in the grounding electrode training feature vector comprise physical feature quantities of the grounding electrode, which influence the detection result; the physical characteristic quantity includes, but is not limited to, the ground resistance of the ground electrode, the soil humidity, the ground potential, the soil temperature, the current magnitude, the ground electrode displacement, the degree of corrosion, and the like;
it should be noted that, the grounding resistance may be obtained in real time using a resistance tester;
the soil humidity can be obtained in real time by using a humidity sensor;
the ground potential can be obtained in real time by using a ground potential tester;
the soil temperature can be obtained in real time by using a temperature sensor;
the current magnitude can be obtained in real time using a current sensor;
the earth electrode displacement can be obtained in real time by using a displacement sensor;
the degree of corrosion may be obtained in real time by using a metal corrosion sensor;
the detection result label is the result of each periodic detection of a detection person, and is one of 0 or 1; specifically, when the detection result is that maintenance is needed, the detection result label is 1; when the detection result is that maintenance is not needed, the detection result label is 0;
the detection cost training data are collected grounding electrode training feature vectors and corresponding detection cost labels determined by detection personnel;
the detection cost label is the comprehensive cost required to be spent for each detection;
specifically, the calculation mode of the comprehensive cost is as follows:
marking the time length of the detection personnel going to the grounding electrode as t1, and marking the time length of the detection personnel detecting the grounding electrode as t2;
the total cost spent in going to the ground and in the detection process is marked asThe method comprises the steps of carrying out a first treatment on the surface of the It will be appreciated that the total cost spent may include loss of the detection equipment or road tolls for the detection personnel to go to the ground, etc.;
the comprehensive cost is marked as H, and the calculation formula of the comprehensive cost H isThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>、/>Respectively preset proportional coefficients;
the first machine learning model for predicting the abnormal probability of the grounding electrode is trained in the following way:
taking each group of grounding electrode training feature vectors in the grounding electrode abnormal training data as input of a first machine learning model, wherein the first machine learning model takes the predicted grounding electrode abnormal probability of each group of grounding electrode training feature vectors as output, takes a detection result label corresponding to the grounding electrode training feature vectors in the grounding electrode abnormal training data as a prediction target, and takes the sum of the prediction errors between the minimum grounding electrode abnormal probability and the detection result label as a training target; training the first machine learning model until the sum of the prediction errors reaches convergence, and stopping training to obtain a first machine learning model which outputs predicted abnormal grounding electrode probability according to the grounding electrode training feature vector; the first machine learning model is any one of a polynomial regression model or an SVR model;
the second machine learning model that predicts the cost of detection is trained in the following manner:
taking each group of grounding electrode training feature vectors in the detection cost training data as input of a second machine learning model, wherein the second machine learning model takes the predicted detection cost labels of each group of grounding electrode training feature vectors as output, takes the detection cost labels corresponding to the grounding electrode training feature vectors in the detection cost training data as prediction targets, and takes the sum of all prediction errors of the detection cost labels as a training target; training the second machine learning model until the sum of the prediction errors reaches convergence, and stopping training to obtain the second machine learning model which outputs the predicted detection cost label according to the grounding electrode training feature vector; the second machine learning model is any one of a polynomial regression model or an SVR model;
it should be noted that, the calculation formula of the prediction error is:wherein c is the number of the feature data, uc is the prediction error, wc is the predicted state value corresponding to the feature data of the c-th group, and yc is the actual state value corresponding to the feature data of the c-th group; for example, in the first machine learning model, the feature data corresponds to a ground electrode training feature vector, and the state value corresponds to a detection result label; the feature data in the second machine learning model corresponds to a ground electrode training feature vector, and the state value corresponds to a detection cost label;
the non-periodic detection process is a time period for detecting the grounding electrode after training a first machine learning model and a second machine learning model based on historical early warning training data collected by periodic detection histories;
further, the mode of collecting the early warning training data is as follows:
presetting a sample collection period, collecting each physical characteristic quantity corresponding to each element in a grounding electrode training characteristic vector of a grounding electrode every sample collection period, and forming each physical characteristic quantity into early warning training data;
the predicted ground electrode anomaly probability is obtained by:
taking the early warning training data as the input of a first machine learning model, and obtaining the predicted abnormal probability of the grounding electrode output by the first machine learning model;
based on the predicted abnormal probability of the grounding electrode, judging whether the early warning is needed or not is as follows:
presetting an abnormal probability threshold, and judging that early warning is needed if the predicted abnormal probability of the grounding electrode is larger than the abnormal probability threshold; if the predicted abnormal probability of the grounding electrode is smaller than or equal to the abnormal probability threshold value, judging that early warning is not needed;
the first update of the initial detection period is performed in the following manner:
updating the initial detection period to be: starting from the last detection end to the time when the early warning is judged to be needed;
the method for obtaining the predicted detection cost comprises the following steps:
taking the early warning training data as the input of a second machine learning model, and obtaining a predicted detection cost label output by the second machine learning model;
the method for outputting a decision result for carrying out second updating on the initial detection period by using an Actor network in the Actor-Critic network model and training the Actor-Critic network model comprises the following steps:
initializing parameters of an Actor network and a Critic network; parameters include, but are not limited to, the dimensions of the state input layers of the Actor network, the number and size of hidden layers, the dimensions of the action output layers, the dimensions of the state input layers of the Critic network, the number and size of hidden layers and the dimensions of the prize value function output layers, the learning rate, discount factors, and network optimization algorithms (gradient descent method or Adam optimization algorithm, etc.);
when the early warning is judged not to be needed, the following steps are executed:
step 11: taking early warning training data corresponding to the current sample collection period as a current state;
step 12: the Actor network outputs a decision result of a second update of the initial detection period;
taking early warning training data of a next round of sample collection period as a next state;
step 13: calculating an actual rewarding value Q; the actual rewarding value Q is obtained after making a second updating decision on the initial detection period each time;
specifically, the calculation mode of the reward value Q is as follows:
marking the value of the initial detection period before the second update as R0, and marking the value of the initial detection period after the second update as R;
marking the predicted abnormal probability of the grounding electrode as P, and marking the predicted detection cost label as D;
the calculation formula of the reward value Q isThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>、/>And->Respectively preset proportional coefficients; it can be understood that when +.>Larger and->In the course of the time period of the smaller,will be greater than 0, thus->The bigger the->The larger the detection cost is, the lower the abnormal probability of the grounding electrode is, the detection period tends to be properly improved; conversely, the detection period tends to be reduced to perform detection as soon as possible; whileIs for avoiding that the initial detection period of the second update deviates too much from the past detection period;
step 14: updating the value of the prize value function using an update formula of the Critic network to adjust the estimate of the actual prize value Q for the decision result; it should be noted that the update formula may be a conventional update formula for those skilled in the art, for example:wherein->Is the current state->Is a bonus value function estimate of (2); />Is learning rate, controls the updated step length; />Is a discount factor for measuring the importance of future rewards; />Is the next state;
step 15: updating parameters of the Actor network by using an updating formula of the Actor network so as to improve the probability of selecting a high rewarding decision result under a given current state;
based on the initial detection period, the mode of judging whether to detect the grounding electrode is as follows:
marking the time length from the last detection end to the current time as T;
if the duration T is smaller than the initial detection period, not processing;
if the duration T is greater than or equal to the initial detection period, notifying a detector to go to the grounding electrode for detection.
Example 2
As shown in fig. 2, the detection and early warning system for the hvth grounding electrode comprises a training data collection module, a model training module, an early warning training data collection module and a detection period decision module; wherein, each module is electrically connected with each other;
the training data collection module is mainly used for collecting historical early warning training data in a periodic detection history in advance and sending the historical early warning training data to the model training module;
the model training module is mainly used for training a first machine learning model for predicting the abnormal probability of the grounding electrode based on historical early warning training data and a second machine learning model for predicting the detection cost, and sending the first machine learning model and the second machine learning model to the detection period decision module;
the early warning training data collection module is mainly used for obtaining a detection period in a periodic detection history as an initial detection period, collecting early warning training data in an unscheduled detection process, and sending the initial detection period and the early warning training data to the detection period decision module;
the detection period decision module is mainly used for updating an initial detection period based on early warning training data, a first machine learning model and a second machine learning model, and judging whether to detect a grounding electrode based on the initial detection period;
the method for updating the initial detection period comprises the following steps:
based on early warning training data and a first machine learning model, obtaining predicted abnormal probabilities of the grounding electrode, and judging whether early warning is needed or not based on the predicted abnormal probabilities of the grounding electrode;
if the early warning is needed, the early warning is carried out, and the initial detection period is updated for the first time;
if the early warning is not needed, based on early warning training data and a second machine learning model, the predicted detection cost is obtained, based on the initial detection period, the predicted abnormal probability of the grounding electrode and the predicted detection cost, an Actor network output in the Actor-Critic network model is used for carrying out a decision result of second updating on the initial detection period, and the Actor-Critic network model is trained.
Example 3
Fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present application. As shown in fig. 3, an electronic device 100 is also provided according to yet another aspect of the present application. The electronic device 100 may include one or more processors and one or more memories. Wherein the memory has stored therein computer readable code which, when executed by the one or more processors, performs the hvdc transmission grounding electrode detection and early warning method as described above.
The method or system according to embodiments of the present application may also be implemented by means of the architecture of the electronic device shown in fig. 3. As shown in fig. 3, the electronic device 100 may include a bus 101, one or more CPUs 102, a Read Only Memory (ROM) 103, a Random Access Memory (RAM) 104, a communication port 105 connected to a network, an input/output component 106, a hard disk 107, and the like. A storage device in the electronic device 100, such as the ROM103 or the hard disk 107, may store the hvth grounding electrode detection and early warning method provided in the present application. The method for detecting and early warning the high-voltage direct-current transmission grounding electrode can comprise the following steps: step one: the method comprises the steps of collecting historical early warning training data in a periodic detection history in advance, and training a first machine learning model for predicting the abnormal probability of the grounding electrode and a second machine learning model for predicting the detection cost based on the historical early warning training data; step two: acquiring a detection period in the periodic detection history as an initial detection period; in the non-periodic detection process, early warning training data are collected, the predicted abnormal probability of the grounding electrode is obtained based on the early warning training data and the first machine learning model, and whether early warning is needed or not is judged based on the predicted abnormal probability of the grounding electrode; if the early warning is needed, the early warning is carried out, and the initial detection period is updated for the first time; if early warning is not needed, based on early warning training data and a second machine learning model, obtaining predicted detection cost, based on an initial detection period, predicted abnormal grounding electrode probability and predicted detection cost, using an Actor network output in an Actor-Critic network model to carry out a decision result of second updating on the initial detection period, and training the Actor-Critic network model; step three: judging whether to detect the grounding electrode or not based on the initial detection period;
further, the electronic device 100 may also include a user interface 108. Of course, the architecture shown in fig. 3 is merely exemplary, and one or more components of the electronic device shown in fig. 3 may be omitted as may be practical in implementing different devices.
Example 4
Fig. 4 is a schematic structural diagram of a computer readable storage medium according to an embodiment of the present application. As shown in fig. 4, is a computer-readable storage medium 200 according to one embodiment of the present application. The computer-readable storage medium 200 has stored thereon computer-readable instructions. The method for detecting and early warning the hvth grounding electrode according to the embodiments of the application described with reference to the above figures may be performed when the computer readable instructions are executed by a processor. Computer-readable storage medium 200 includes, but is not limited to, for example, volatile memory and/or nonvolatile memory. Volatile memory can include, for example, random Access Memory (RAM), cache memory (cache), and the like. The non-volatile memory may include, for example, read Only Memory (ROM), hard disk, flash memory, and the like.
In addition, according to embodiments of the present application, the processes described above with reference to flowcharts may be implemented as computer software programs. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions executable by a processor to perform instructions corresponding to the method steps provided herein, which when executed by a Central Processing Unit (CPU), perform the functions defined above in the methods of the present application.
The methods and apparatus, devices, and apparatus of the present application may be implemented in numerous ways. For example, the methods and apparatus, devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, firmware. The above-described sequence of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the sequence specifically described above unless specifically stated otherwise. Furthermore, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, the programs including machine-readable instructions for implementing the methods according to the present application. Thus, the present application also covers a recording medium storing a program for executing the method according to the present application.
In addition, in the foregoing technical solutions provided in the embodiments of the present application, parts consistent with implementation principles of corresponding technical solutions in the prior art are not described in detail, so that redundant descriptions are avoided.
The purpose, technical scheme and beneficial effects of the invention are further described in detail in the detailed description. It is to be understood that the above description is only of specific embodiments of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
The above preset parameters or preset thresholds are set by those skilled in the art according to actual conditions or are obtained by mass data simulation.
The above embodiments are only for illustrating the technical method of the present invention and not for limiting the same, and it should be understood by those skilled in the art that the technical method of the present invention may be modified or substituted without departing from the spirit and scope of the technical method of the present invention.
Claims (8)
1. The high-voltage direct-current transmission grounding electrode detection and early warning method is characterized by comprising the following steps of:
step one: the method comprises the steps of collecting historical early warning training data in a periodic detection history in advance, and training a first machine learning model for predicting the abnormal probability of the grounding electrode and a second machine learning model for predicting the detection cost based on the historical early warning training data;
step two: acquiring a detection period in the periodic detection history as an initial detection period;
in the non-periodic detection process, early warning training data are collected, the predicted abnormal probability of the grounding electrode is obtained based on the early warning training data and the first machine learning model, and whether early warning is needed or not is judged based on the predicted abnormal probability of the grounding electrode;
if the early warning is needed, the early warning is carried out, and the initial detection period is updated for the first time;
if early warning is not needed, based on early warning training data and a second machine learning model, obtaining predicted detection cost, based on an initial detection period, predicted abnormal grounding electrode probability and predicted detection cost, using an Actor network output in an Actor-Critic network model to carry out a decision result of second updating on the initial detection period, and training the Actor-Critic network model;
step three: judging whether to detect the grounding electrode or not based on the initial detection period;
the mode of collecting early warning training data is as follows:
presetting a sample collection period, collecting each physical characteristic quantity corresponding to each element in a grounding electrode training characteristic vector of a grounding electrode every sample collection period, and forming each physical characteristic quantity into early warning training data;
based on the predicted abnormal probability of the grounding electrode, judging whether the early warning is needed or not is as follows:
presetting an abnormal probability threshold, and judging that early warning is needed if the predicted abnormal probability of the grounding electrode is larger than the abnormal probability threshold;
the first update of the initial detection period is performed in the following manner:
updating the initial detection period to be: starting from the last detection end to the time when the early warning is judged to be needed;
the method for outputting a decision result for carrying out second updating on the initial detection period by using an Actor network in the Actor-Critic network model and training the Actor-Critic network model comprises the following steps:
initializing parameters of an Actor network and a Critic network;
when the early warning is judged not to be needed, the following steps are executed:
step 11: taking early warning training data corresponding to the current sample collection period as a current state;
step 12: the Actor network outputs a decision result of a second update of the initial detection period;
taking early warning training data of a next round of sample collection period as a next state;
step 13: calculating an actual rewarding value Q; the actual rewarding value Q is obtained after making a second updating decision on the initial detection period each time;
the calculation mode of the reward value Q is as follows:
marking the value of the initial detection period before the second update as R0, and marking the value of the initial detection period after the second update as R;
marking the predicted abnormal probability of the grounding electrode as P, and marking the predicted detection cost label as D;
the calculation formula of the reward value Q isThe method comprises the steps of carrying out a first treatment on the surface of the Wherein,、/>and->Respectively preset proportional coefficients;
step 14: updating the value of the prize value function using an update formula of the Critic network to adjust the estimate of the actual prize value Q for the decision result;
step 15: and updating the parameters of the Actor network by using an updating formula of the Actor network.
2. The method for detecting and warning a grounding electrode for high voltage direct current transmission according to claim 1, wherein the periodic detection history is a period of time during which the grounding electrode is periodically detected in the past high voltage direct current transmission process;
the historical early warning training data comprises ground electrode abnormal training data and detection cost training data.
3. The method for detecting and early warning a high-voltage direct-current transmission grounding electrode according to claim 2, wherein the grounding electrode abnormal training data comprises a collected grounding electrode training feature vector and a corresponding detection result label determined by a detector during each periodic detection;
the elements in the grounding electrode training feature vector comprise physical feature quantities of the grounding electrode, which influence detection results;
the detection result label is the result of each periodic detection of a detection person, and is one of 0 or 1;
the detection cost training data are collected grounding electrode training feature vectors and corresponding detection cost labels determined by detection personnel;
the detection cost label is the comprehensive cost required by each detection.
4. The method for detecting and pre-warning a high voltage direct current transmission grounding electrode according to claim 3, wherein the comprehensive cost is calculated by the following steps:
marking the time length of the detection personnel going to the grounding electrode as t1, and marking the time length of the detection personnel detecting the grounding electrode as t2;
the total cost spent in going to the ground and in the detection process is marked as；
The comprehensive cost is marked as H, and the calculation formula of the comprehensive cost H isThe method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>、/>Respectively, are preset proportional coefficients.
5. The method for detecting and warning a grounding electrode for high voltage direct current transmission according to claim 4, wherein the first machine learning model for predicting the probability of the grounding electrode abnormality is trained by:
taking each group of grounding electrode training feature vectors in the grounding electrode abnormal training data as input of a first machine learning model, wherein the first machine learning model takes the predicted grounding electrode abnormal probability of each group of grounding electrode training feature vectors as output, takes a detection result label corresponding to the grounding electrode training feature vectors in the grounding electrode abnormal training data as a prediction target, and takes the sum of the prediction errors between the minimum grounding electrode abnormal probability and the detection result label as a training target; and training the first machine learning model until the sum of the prediction errors reaches convergence, and stopping training to obtain the first machine learning model which outputs the predicted abnormal probability of the grounding electrode according to the grounding electrode training feature vector.
6. The method for detecting and warning a hvth power transmission earthing pole according to claim 5, wherein the second machine learning model for training the predicted detection cost is:
taking each group of grounding electrode training feature vectors in the detection cost training data as input of a second machine learning model, wherein the second machine learning model takes the predicted detection cost labels of each group of grounding electrode training feature vectors as output, takes the detection cost labels corresponding to the grounding electrode training feature vectors in the detection cost training data as prediction targets, and takes the sum of all prediction errors of the detection cost labels as a training target; and training the second machine learning model until the sum of the prediction errors reaches convergence, and stopping training to obtain the second machine learning model for outputting the predicted detection cost label according to the grounding electrode training feature vector.
7. The method for detecting and warning a grounding electrode for high voltage direct current transmission according to claim 6, wherein the determining whether to detect the grounding electrode based on the initial detection period is:
marking the time length from the last detection end to the current time as T;
if the duration T is smaller than the initial detection period, not processing;
if the duration T is greater than or equal to the initial detection period, notifying a detector to go to the grounding electrode for detection.
8. The high-voltage direct-current transmission grounding electrode detection and early warning system is realized based on the high-voltage direct-current transmission grounding electrode detection and early warning method according to any one of claims 1-7, and is characterized by comprising a training data collection module, a model training module, an early warning training data collection module and a detection period decision module; wherein, each module is electrically connected with each other;
the training data collection module is used for pre-collecting historical early warning training data in the periodic detection history and sending the historical early warning training data to the model training module;
the model training module is used for training a first machine learning model for predicting the abnormal probability of the grounding electrode and a second machine learning model for predicting the detection cost based on the historical early warning training data, and sending the first machine learning model and the second machine learning model to the detection period decision module;
the early warning training data collection module is used for obtaining a detection period in the periodic detection history as an initial detection period, collecting early warning training data in the non-periodic detection process and sending the initial detection period and the early warning training data to the detection period decision module;
the detection period decision module is used for updating the initial detection period based on the early warning training data, the first machine learning model and the second machine learning model and judging whether to detect the grounding electrode or not based on the initial detection period;
the method for updating the initial detection period comprises the following steps:
based on early warning training data and a first machine learning model, obtaining predicted abnormal probabilities of the grounding electrode, and judging whether early warning is needed or not based on the predicted abnormal probabilities of the grounding electrode;
if the early warning is needed, the early warning is carried out, and the initial detection period is updated for the first time;
if the early warning is not needed, based on early warning training data and a second machine learning model, the predicted detection cost is obtained, based on the initial detection period, the predicted abnormal probability of the grounding electrode and the predicted detection cost, an Actor network output in the Actor-Critic network model is used for carrying out a decision result of second updating on the initial detection period, and the Actor-Critic network model is trained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311330062.4A CN117077065B (en) | 2023-10-16 | 2023-10-16 | High-voltage direct-current transmission grounding electrode detection and early warning method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311330062.4A CN117077065B (en) | 2023-10-16 | 2023-10-16 | High-voltage direct-current transmission grounding electrode detection and early warning method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117077065A CN117077065A (en) | 2023-11-17 |
| CN117077065B true CN117077065B (en) | 2024-01-26 |
Family
ID=88706395
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311330062.4A Active CN117077065B (en) | 2023-10-16 | 2023-10-16 | High-voltage direct-current transmission grounding electrode detection and early warning method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117077065B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2018189620A (en) * | 2017-05-08 | 2018-11-29 | 有限会社 山本エンジニアリング | Insulator remaining life measurement device by discharge amount detection of partial discharging, corona discharging, and creeping discharging (hereinafter, referred to as corona discharging) |
| CN111445010A (en) * | 2020-03-26 | 2020-07-24 | 南京工程学院 | Distribution network voltage trend early warning method based on evidence theory fusion quantum network |
| CN113239625A (en) * | 2021-05-24 | 2021-08-10 | 中国电力科学研究院有限公司 | Power grid stability change trend prediction method and system based on deep reinforcement learning |
| CN114520743A (en) * | 2022-02-24 | 2022-05-20 | 周口师范学院 | Method and system for detecting network abnormal flow and storable medium |
-
2023
- 2023-10-16 CN CN202311330062.4A patent/CN117077065B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2018189620A (en) * | 2017-05-08 | 2018-11-29 | 有限会社 山本エンジニアリング | Insulator remaining life measurement device by discharge amount detection of partial discharging, corona discharging, and creeping discharging (hereinafter, referred to as corona discharging) |
| CN111445010A (en) * | 2020-03-26 | 2020-07-24 | 南京工程学院 | Distribution network voltage trend early warning method based on evidence theory fusion quantum network |
| CN113239625A (en) * | 2021-05-24 | 2021-08-10 | 中国电力科学研究院有限公司 | Power grid stability change trend prediction method and system based on deep reinforcement learning |
| CN114520743A (en) * | 2022-02-24 | 2022-05-20 | 周口师范学院 | Method and system for detecting network abnormal flow and storable medium |
Non-Patent Citations (1)
| Title |
|---|
| 特高压直流输电接地极在线监测系统设计;王玮;倪平浩;唐琪;刘平竹;;北京交通大学学报(第02期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117077065A (en) | 2023-11-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10955456B2 (en) | Method and apparatus for automatic localization of a fault | |
| EP2989705B1 (en) | Method and apparatus for defect pre-warning of power device | |
| JP6815480B2 (en) | Methods and systems for discovering precursor substrings in time series | |
| CN112766342A (en) | Abnormity detection method for electrical equipment | |
| US11442444B2 (en) | System and method for forecasting industrial machine failures | |
| JP2004524609A (en) | Adaptive modeling of changing states in predictive state monitoring | |
| CN108892014A (en) | A kind of elevator internal contracting brake fault early warning method based on Elman neural network | |
| KR20210017651A (en) | Method for Fault Detection and Fault Diagnosis in Semiconductor Manufacturing Process | |
| CN115236469B (en) | Online monitoring method, system, medium and equipment for unattended transformer substation | |
| CN117407679A (en) | Data acquisition method and system of intelligent end screen sensor | |
| CN117077065B (en) | High-voltage direct-current transmission grounding electrode detection and early warning method and system | |
| CN116777088B (en) | Power supply emergency repair environment monitoring method and system for guaranteeing life safety | |
| KR102108975B1 (en) | Apparatus and method for condition based maintenance support of naval ship equipment | |
| CN115795999B (en) | Early warning method for abnormal performance of long-term service pumped storage unit | |
| US20230297095A1 (en) | Monitoring device and method for detecting anomalies | |
| CN116430272A (en) | Electrical disturbance-based coal mine equipment abnormality diagnosis method and system and electronic equipment | |
| CN113986700A (en) | Data acquisition frequency optimization method, system, device and storage medium | |
| CN113887640A (en) | Method and device for detecting abnormal time sequence of Internet of things card | |
| Herp et al. | State transition of wind turbines based on empirical inference on model residuals | |
| Razmi et al. | Data-Driven Intermittent Earth Fault Detection in Compensated and Isolated MV Networks | |
| CN116627759B (en) | Financial payment equipment circuit safety detection device | |
| US20240193025A1 (en) | Remaining useful life determination for power electronic devices | |
| CN110135763B (en) | Method for monitoring power consumption, readable storage medium and electronic device | |
| CN117743961A (en) | Cable joint fault detection method and system based on long-short-term memory LSTM | |
| CN117556238A (en) | Power equipment inspection method and system based on fault prediction |
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 |