CN118091328A - Abnormal fault prediction method and system for distribution cable - Google Patents
Abnormal fault prediction method and system for distribution cable Download PDFInfo
- Publication number
- CN118091328A CN118091328A CN202410478994.1A CN202410478994A CN118091328A CN 118091328 A CN118091328 A CN 118091328A CN 202410478994 A CN202410478994 A CN 202410478994A CN 118091328 A CN118091328 A CN 118091328A
- Authority
- CN
- China
- Prior art keywords
- temperature
- data
- power data
- prediction
- power
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 80
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000001514 detection method Methods 0.000 claims abstract description 56
- 238000010276 construction Methods 0.000 claims abstract description 12
- 238000012549 training Methods 0.000 claims description 129
- 238000012544 monitoring process Methods 0.000 claims description 57
- 238000012937 correction Methods 0.000 claims description 49
- 239000013598 vector Substances 0.000 claims description 48
- 238000012545 processing Methods 0.000 claims description 28
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 239000002245 particle Substances 0.000 claims description 3
- 238000004088 simulation Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007405 data analysis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
Landscapes
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention relates to the field of fault detection, in particular to an abnormal fault prediction method and system for a distribution cable. An abnormal fault prediction system for a power distribution cable, comprising: the system comprises a power data acquisition module, a temperature data acquisition module, a power data prediction set construction module, a temperature data time sequence set construction module, a predicted temperature data set construction module, a power data prediction module and a cable abnormal fault detection module. According to the method, the power data are predicted, the influence of temperature is considered in the prediction process, so that the prediction of the power data is more accurate, and the abnormal fault prediction of the cable is performed in advance based on the predicted power data, so that the abnormal fault of the cable is predicted in advance, and the normal use of the cable is ensured.
Description
Technical Field
The invention relates to the field of fault detection, in particular to an abnormal fault prediction method and system for a distribution cable.
Background
With the rapid development of power systems, higher requirements are put on the quality and reliability of power supply. In order to realize effective fault prevention and prediction, it is particularly important to monitor and analyze the running state of the distribution cable in real time by combining a big data analysis technology. However, the existing fault prediction method is mainly based on traditional power data such as current and voltage of a cable, and the influence of factors such as external environment temperature is not fully considered.
Disclosure of Invention
According to the method, the power data are predicted, the influence of temperature is considered in the prediction process, so that the prediction of the power data is more accurate, and the abnormal fault prediction of the cable is performed in advance based on the predicted power data, so that the abnormal fault of the cable is predicted in advance, and the normal use of the cable is ensured.
An abnormal fault prediction method for a power distribution cable, comprising:
Acquiring power data at a current monitoring time point; meanwhile, acquiring temperature data uploaded by the set temperature monitoring points, and forming a temperature data set from the temperature data uploaded by all the currently acquired temperature monitoring points;
combining the currently acquired power data and the power data acquired for the previous T-1 times into a power data prediction set; forming a temperature data time sequence set by the current acquired temperature data set and the temperature data set acquired for the previous T-1 times;
Aiming at each temperature monitoring point, the temperature data obtained at present and the temperature data obtained for the previous T-1 times form a temperature data prediction set, the temperature data prediction set is sent into a temperature prediction model for processing, and the predicted temperature data corresponding to each temperature monitoring point is output; then, the predicted temperature data corresponding to all the temperature monitoring points are formed into a predicted temperature data set;
Sending the power data prediction set, the temperature data time sequence set and the predicted temperature data set into a power data prediction model for processing, and outputting predicted power data corresponding to the next monitoring time point;
And sending the predicted power data into a cable abnormal fault detection model for processing, and outputting a cable abnormal fault detection result.
Preferably, the temperature prediction model is built based on an LSTM model and comprises T temperature prediction LSTM units, 1 first fully-connected layer and 1 first output layer, wherein the temperature prediction LSTM units are used for extracting time sequence characteristics in a temperature data prediction set, and temperature hiding characteristic vectors are transmitted between adjacent temperature prediction LSTM units; the first full-connection layer is used for extracting the characteristics of the outputs of all the temperature prediction LSTM units; the first output layer is used for outputting predicted temperature data;
Training for the temperature prediction model comprises the following steps:
Acquiring a plurality of temperature data training samples, wherein T+1 temperature data arranged according to time sequence are stored in the temperature data training samples, the first T temperature data in the temperature data training samples are used as input data of a temperature prediction model, and the last 1 temperature data in the temperature data training samples are used as output data of the temperature prediction model; all temperature data training samples form a temperature training set, the temperature training set is sent to a parameter initialized temperature prediction model for training, whether a first training condition is met or not is judged, and if the first training condition is met, a trained temperature prediction model is output; otherwise, continuing to train the temperature prediction model through the temperature training set.
Preferably, the power data prediction model is built based on an LSTM model, and comprises T power data prediction LSTM units, T-1 first temperature correction layers, 1 second full connection layer, 1 second temperature correction layer and 1 second output layer, wherein the power data prediction LSTM units are used for extracting time sequence characteristics in power data prediction sets, and power hiding characteristic vectors are transmitted between adjacent power data prediction LSTM units; the first temperature correction layer is used for correcting the power hiding characteristic vector through the temperature data time sequence set; the second full connection layer is used for extracting the characteristics of the output of all the power data prediction LSTM units so as to construct the predicted power data to be processed; the second temperature correction layer is used for correcting the predicted power data to be processed through the predicted temperature data set so as to construct predicted power data; the second output layer is used for outputting the predicted power data.
Preferably, the power data prediction set, the temperature data time sequence set and the predicted temperature data set are sent to a power data prediction model for processing, and predicted power data corresponding to the next monitoring time point is output, and the method specifically comprises the following steps:
the method comprises the steps of inputting power data in a power data prediction set into T power data prediction LSTM units one by one according to a time sequence for processing, correcting power hiding feature vectors transmitted between adjacent power data prediction LSTM units through a first temperature correction layer during processing, specifically operating to select a temperature data set acquired by the power data input to the power data prediction LSTM units of the adjacent power data prediction LSTM units from a temperature data time sequence set, performing point multiplication on the selected temperature data set and the first temperature correction weight vector, outputting a first temperature correction vector, performing point multiplication on the first temperature correction vector and the power hiding feature vector output by the power data prediction LSTM units of the former, outputting a power hiding feature intermediate vector, performing residual connection and standardization on the power hiding feature vector output by the power data prediction LSTM units of the former, and then, serving as a power hiding feature vector input by the next power data prediction LSTM unit;
the output of all the power data prediction LSTM units is sent to a second full-connection layer for processing, and the predicted power data to be processed is output;
In the second temperature correction layer, performing point multiplication on the predicted temperature data set and a second temperature correction weight vector, outputting a second temperature correction vector, performing point multiplication on the second temperature correction vector and predicted power data to be processed, outputting corrected predicted power data, and performing residual connection and standardization on the corrected predicted power data and the predicted power data to be processed to generate predicted power data;
and finally, outputting predicted power data corresponding to the next monitoring time point through the second output layer.
Preferably, the training of the power data prediction model comprises the following steps:
Acquiring a plurality of electric power data training samples, wherein T+1 electric power data arranged according to a time sequence are stored in the electric power data training samples, the first T electric power data in the electric power data training samples are used as input data of an electric power data prediction model, and the last 1 electric power data in the electric power data training samples are used as output data of the electric power data prediction model; for each power data training sample, acquiring a temperature data time sequence set training sample corresponding to the power data training sample, wherein T+1 temperature data sets arranged according to time sequence are stored in the temperature data time sequence set training sample, and the time corresponding to the temperature data set in the temperature data time sequence set training sample corresponds to the time corresponding to the power data in the power data training sample one by one; the method comprises the steps that a power data training sample and a corresponding temperature data time sequence set training sample form a power data prediction sample, all power data prediction samples form a power data training set, then the power data training set is sent into a parameter initialized power data prediction model to be trained, during the period, the first T-1 temperature data sets in the temperature data time sequence set training sample are used as input data of T-1 first temperature correction layers in the power data prediction model, the last temperature data set in the temperature data time sequence set training sample is used as input data of a second temperature correction layer in the power data prediction model, whether the second training condition is met is judged, and if the second training condition is met, the trained power data prediction model is output; otherwise, continuing to train the power data prediction model through the power data training set.
Preferably, the cable abnormal fault detection model is built based on the BP neural network, and super parameters of the cable abnormal fault detection model are adjusted through a particle swarm optimization algorithm.
Preferably, training for the cable abnormal fault detection model specifically comprises the following steps:
Acquiring a plurality of pieces of electric power data marked with the cable abnormal fault detection results, forming a cable abnormal detection training set by all pieces of electric power data marked with the cable abnormal fault detection results, sending the cable abnormal detection training set into a cable abnormal fault detection model initialized by parameters for training, judging whether a third training condition is met, and outputting the trained cable abnormal fault detection model if the third training condition is met; otherwise, continuing to train the cable abnormal fault detection model through the cable abnormal detection training set.
An abnormal fault prediction system for a power distribution cable, comprising:
the power data acquisition module is used for acquiring power data at the current monitoring time point;
the temperature data acquisition module is used for acquiring temperature data uploaded by the set temperature monitoring points and forming a temperature data set from the temperature data uploaded by all the currently acquired temperature monitoring points;
The power data prediction set construction module is used for forming a power data prediction set from currently acquired power data and power data acquired for the previous T-1 times;
The temperature data time sequence set construction module is used for forming a temperature data time sequence set by the current acquired temperature data set and the temperature data set acquired for the previous T-1 times;
The predicted temperature data set construction module is used for forming a temperature data predicted set by the currently acquired temperature data and the temperature data acquired for the previous T-1 time aiming at each temperature monitoring point, sending the temperature data predicted set into a temperature prediction model for processing, and outputting predicted temperature data corresponding to each temperature monitoring point; then, the predicted temperature data corresponding to all the temperature monitoring points are formed into a predicted temperature data set;
The power data prediction module is used for sending the power data prediction set, the temperature data time sequence set and the predicted temperature data set into the power data prediction model for processing and outputting predicted power data corresponding to the next monitoring time point;
And the cable abnormal fault detection module is used for sending the predicted power data into the cable abnormal fault detection model for processing and outputting a cable abnormal fault detection result.
The invention has the following advantages:
1. According to the method, the power data are predicted, the influence of temperature is considered in the prediction process, so that the prediction of the power data is more accurate, and the abnormal fault prediction of the cable is performed in advance based on the predicted power data, so that the abnormal fault of the cable is predicted in advance, and the normal use of the cable is ensured.
2. According to the method, the influence of the temperature condition on the cable power data is considered in the power data prediction process, the first temperature correction is performed based on the temperature data time sequence set corresponding to the input power data prediction set, and the second temperature correction is performed based on the prediction temperature data set, so that the effect of improving the power data prediction accuracy can be achieved, and the accuracy of detecting the abnormal faults of the follow-up cable is improved.
Drawings
Fig. 1 is a schematic structural diagram of an abnormal fault prediction system of a distribution cable according to an embodiment of the present invention.
Detailed Description
In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.
Examples
An abnormal fault prediction method for a power distribution cable, comprising:
Acquiring power data of the distribution cable in the using process at the current monitoring time point, wherein the power data comprise three-phase voltage effective values, three-phase current effective values and the like, and the power data can reflect the performance change of the distribution cable in the using process; meanwhile, temperature data uploaded by temperature monitoring points arranged at each position of the cable are acquired, and the temperature data uploaded by all the currently acquired temperature monitoring points are formed into a temperature data set, and the temperature monitoring points can be set at fixed intervals for one section of cable, and a temperature sensor is arranged at the temperature monitoring points to acquire the temperature data;
The current acquired power data and the power data acquired for the previous T-1 times form a power data prediction set, and the change trend of the power data can be learned through the power data prediction set; forming a temperature data time sequence set by the current acquired temperature data set and the temperature data set acquired for the previous T-1 times;
aiming at each temperature monitoring point, the temperature data obtained at present and the temperature data obtained for the previous T-1 times form a temperature data prediction set, the temperature change of the corresponding position of each temperature monitoring point can be analyzed through the temperature data prediction set of each temperature monitoring point, the temperature data prediction set is sent into a temperature prediction model for processing, and the predicted temperature data corresponding to each temperature monitoring point is output; the predicted temperature data corresponding to all the temperature monitoring points are formed into a predicted temperature data set, the predicted temperature data set can represent the next temperature change, and correction references are provided for the prediction of the power data;
The power data prediction set, the temperature data time sequence set and the predicted temperature data set are sent into a power data prediction model to be processed, predicted power data corresponding to the next monitoring time point is output, and as the power data of the cable can be affected by temperature, the power data prediction is corrected through the temperature data predicted at the next monitoring time point, so that a more accurate prediction effect can be achieved;
And sending the predicted power data into a cable abnormal fault detection model for processing, and outputting a cable abnormal fault detection result, wherein the cable abnormal fault detection result comprises faults and no faults.
According to the method, the power data are predicted, the influence of temperature is considered in the prediction process, so that the prediction of the power data is more accurate, and the abnormal fault prediction of the cable is performed in advance based on the predicted power data, so that the abnormal fault of the cable is predicted in advance, and the normal use of the cable is ensured.
The temperature prediction model is built based on an LSTM model and comprises T temperature prediction LSTM units, 1 first full-connection layer and 1 first output layer, wherein the temperature prediction LSTM units are used for parameter setting by referring to the LSTM model and used for extracting time sequence characteristics in temperature data prediction sets, temperature hiding characteristic vectors are transmitted between adjacent temperature prediction LSTM units, and the temperature hiding characteristic vectors are data stored by cell units of the temperature prediction LSTM units in the temperature prediction model and are used for realizing long-dependence time sequence characteristic learning; the first full-connection layer is used for extracting the characteristics of the outputs of all the temperature prediction LSTM units; the first output layer is used for outputting predicted temperature data;
Training for the temperature prediction model comprises the following steps:
Acquiring a plurality of temperature data training samples, wherein T+1 temperature data arranged according to time sequence are stored in the temperature data training samples, the first T temperature data in the temperature data training samples are used as input data of a temperature prediction model, the last 1 temperature data in the temperature data training samples are used as output data of the temperature prediction model, and the temperature data training samples are acquired according to field temperature monitoring and can also be acquired according to field environment modeling; all temperature data training samples form a temperature training set, the temperature training set is sent to a parameter initialized temperature prediction model for training, whether a first training condition is met or not is judged, the first training condition can be a certain training frequency or a target accuracy rate, and if the first training condition is met, the trained temperature prediction model is output; otherwise, continuing to train the temperature prediction model through the temperature training set.
The power data prediction model is built based on an LSTM model and comprises T power data prediction LSTM units, T-1 first temperature correction layers, 1 second full-connection layer, 1 second temperature correction layer and 1 second output layer, wherein the power data prediction LSTM units are used for extracting time sequence characteristics in power data prediction sets, power hiding characteristic vectors are transmitted between adjacent power data prediction LSTM units, and the power hiding characteristic vectors are data stored by cell units of the power data prediction LSTM units in the power data prediction model; the first temperature correction layer is used for correcting the power hiding characteristic vector through the temperature data time sequence set; the second full connection layer is used for extracting the characteristics of the output of all the power data prediction LSTM units so as to construct the predicted power data to be processed; the second temperature correction layer is used for correcting the predicted power data to be processed through the predicted temperature data set so as to construct predicted power data; the second output layer is used for outputting the predicted power data.
The power data prediction set, the temperature data time sequence set and the predicted temperature data set are sent into a power data prediction model to be processed, and predicted power data corresponding to the next monitoring time point is output, specifically comprising the following steps:
The method comprises the steps of inputting power data in a power data prediction set into T power data prediction LSTM units one by one according to a time sequence for processing, correcting power hiding feature vectors transmitted between adjacent power data prediction LSTM units through a first temperature correction layer during processing, specifically, selecting a temperature data set acquired at the same time as power data input to the power data prediction LSTM units of the adjacent two power data prediction LSTM units from a temperature data time sequence set, performing point multiplication on the selected temperature data set and the first temperature correction weight vector, outputting a first temperature correction vector, performing point multiplication on the first temperature correction vector and the power hiding feature vector output by the power data prediction LSTM units of the former, outputting a power hiding feature intermediate vector, performing residual connection and normalization on the power hiding feature vector output by the power data prediction LSTM units of the former, and enabling residual connection to avoid information loss when correcting the power hiding feature intermediate vector to be used as the power hiding feature vector input by the next power data prediction LSTM unit; the first temperature correction weight vector can reflect the influence of temperature on the power data in the cable, the power data prediction LSTM unit learns the time sequence characteristics which are long in dependence, and the power hiding feature vector can be corrected in a priori mode according to actual temperature data, so that the power data prediction process is more fit with the actual situation, and the power data prediction accuracy is improved;
the output of all the power data prediction LSTM units is sent to a second full-connection layer for processing, and the predicted power data to be processed is output;
In the second temperature correction layer, performing point multiplication on the predicted temperature data set and a second temperature correction weight vector, outputting a second temperature correction vector, performing point multiplication on the second temperature correction vector and predicted power data to be processed, outputting corrected predicted power data, and performing residual connection and standardization on the corrected predicted power data and the predicted power data to be processed to generate predicted power data; the second temperature correction vector can also reflect the influence of temperature on the power data in the cable, and the predicted power data is further corrected through the predicted temperature data set, so that the prediction accuracy of the power data can be further improved;
and finally, outputting predicted power data corresponding to the next monitoring time point through the second output layer.
According to the method, the influence of the temperature condition on the cable power data is considered in the power data prediction process, the first temperature correction is performed based on the temperature data time sequence set corresponding to the input power data prediction set, and the second temperature correction is performed based on the prediction temperature data set, so that the effect of improving the power data prediction accuracy can be achieved, and the accuracy of detecting the abnormal faults of the follow-up cable is improved.
Training for a power data prediction model, comprising the steps of:
Acquiring a plurality of electric power data training samples, wherein T+1 electric power data arranged according to time sequence are stored in the electric power data training samples, the first T electric power data in the electric power data training samples are used as input data of an electric power data prediction model, the last 1 electric power data in the electric power data training samples are used as output data of the electric power data prediction model, and the electric power data training samples are obtained through actual simulation according to a cable simulation model; for each power data training sample, acquiring a temperature data time sequence training sample corresponding to the power data training sample, wherein T+1 temperature data sets arranged according to time sequence are stored in the temperature data time sequence training sample, the time corresponding to the temperature data sets in the temperature data time sequence training sample corresponds to the time corresponding to the power data in the power data training sample one by one, namely the power data stored in the power data training sample has a temperature data set acquired in the same time at the corresponding position in the temperature data time sequence training sample, in the process of performing cable simulation, temperature setting can be directly performed at each monitoring point position of a cable, a temperature change mathematical model is set through an actual environment, temperature simulation change is realized, and the temperature data set is acquired; the method comprises the steps that a power data training sample and a corresponding temperature data time sequence set training sample are formed into a power data prediction sample, all power data prediction samples are formed into a power data training set, then the power data training set is sent into a parameter initialized power data prediction model to be trained, during the period, the first T-1 temperature data sets in the temperature data time sequence set training sample are used as input data of T-1 first temperature correction layers in the power data prediction model, the last temperature data set in the temperature data time sequence set training sample is used as input data of a second temperature correction layer in the power data prediction model, whether a second training condition is met is judged, the second training condition can be that a certain training frequency is achieved, the accuracy rate of a target is achieved, and if the second training condition is met, the trained power data prediction model is output; otherwise, continuing to train the power data prediction model through the power data training set.
The cable abnormal fault detection model is built based on the BP neural network, the super parameters of the cable abnormal fault detection model are adjusted through the particle swarm optimization algorithm, the cable abnormal fault detection model can be effectively prevented from sinking into a local optimal solution in the training process, and then the detection accuracy of the cable abnormal fault detection model is improved.
Training a cable abnormal fault detection model, specifically comprising the following steps:
Acquiring a plurality of pieces of electric data marked with the cable abnormal fault detection results, wherein the electric data can be acquired by a professional according to a cable simulation model, forming a cable abnormal detection training set by all pieces of electric data marked with the cable abnormal fault detection results, then sending the cable abnormal detection training set into a cable abnormal fault detection model with initialized parameters for training, judging whether a third training condition is met, wherein the third training condition is generally consistent with the first training condition and the second training condition, can reach a certain training frequency, can reach target accuracy, and outputs the trained cable abnormal fault detection model if the third training condition is met; otherwise, continuing to train the cable abnormal fault detection model through the cable abnormal detection training set.
Examples
An abnormal fault prediction system for a power distribution cable, as shown in fig. 1, comprising:
The power data acquisition module is used for acquiring power data of the distribution cable in the use process at the current monitoring time point, wherein the power data comprises three-phase voltage effective values, three-phase current effective values and the like;
the temperature data acquisition module is used for acquiring temperature data uploaded by temperature monitoring points arranged at each position of the cable and forming a temperature data set from the temperature data uploaded by all the currently acquired temperature monitoring points;
The power data prediction set construction module is used for forming a power data prediction set from currently acquired power data and power data acquired for the previous T-1 times;
The temperature data time sequence set construction module is used for forming a temperature data time sequence set by the current acquired temperature data set and the temperature data set acquired for the previous T-1 times;
The temperature data prediction set construction module is used for forming a temperature data prediction set by aiming at each temperature monitoring point, combining the currently acquired temperature data with the temperature data acquired for the previous T-1 times, analyzing the temperature change of the corresponding position of each temperature monitoring point through the temperature data prediction set of each temperature monitoring point, sending the temperature data prediction set into the temperature prediction model for processing, and outputting the predicted temperature data corresponding to each temperature monitoring point; then, the predicted temperature data corresponding to all the temperature monitoring points are formed into a predicted temperature data set;
The power data prediction module is used for sending the power data prediction set, the temperature data time sequence set and the predicted temperature data set into the power data prediction model for processing and outputting predicted power data corresponding to the next monitoring time point;
the cable abnormal fault detection module is used for sending the predicted power data into the cable abnormal fault detection model for processing, and outputting cable abnormal fault detection results, wherein the cable abnormal fault detection results comprise faults and no faults.
It will be understood that modifications and variations will be apparent to those skilled in the art from the foregoing description, and it is intended that all such modifications and variations be included within the scope of the following claims. Parts of the specification not described in detail belong to the prior art known to those skilled in the art.
Claims (8)
1. A method of predicting an abnormal fault in a power distribution cable, comprising:
Acquiring power data at a current monitoring time point; meanwhile, acquiring temperature data uploaded by the set temperature monitoring points, and forming a temperature data set from the temperature data uploaded by all the currently acquired temperature monitoring points;
combining the currently acquired power data and the power data acquired for the previous T-1 times into a power data prediction set; forming a temperature data time sequence set by the current acquired temperature data set and the temperature data set acquired for the previous T-1 times;
Aiming at each temperature monitoring point, the temperature data obtained at present and the temperature data obtained for the previous T-1 times form a temperature data prediction set, the temperature data prediction set is sent into a temperature prediction model for processing, and the predicted temperature data corresponding to each temperature monitoring point is output; then, the predicted temperature data corresponding to all the temperature monitoring points are formed into a predicted temperature data set;
Sending the power data prediction set, the temperature data time sequence set and the predicted temperature data set into a power data prediction model for processing, and outputting predicted power data corresponding to the next monitoring time point;
And sending the predicted power data into a cable abnormal fault detection model for processing, and outputting a cable abnormal fault detection result.
2. The method for predicting abnormal faults of a power distribution cable according to claim 1, wherein a temperature prediction model is built based on an LSTM model and comprises T temperature prediction LSTM units, 1 first full-connection layer and 1 first output layer, wherein the temperature prediction LSTM units are used for extracting time sequence characteristics in a temperature data prediction set, and temperature hiding characteristic vectors are transmitted between adjacent temperature prediction LSTM units; the first full-connection layer is used for extracting the characteristics of the outputs of all the temperature prediction LSTM units; the first output layer is used for outputting predicted temperature data;
Training for the temperature prediction model comprises the following steps:
Acquiring a plurality of temperature data training samples, wherein T+1 temperature data arranged according to time sequence are stored in the temperature data training samples, the first T temperature data in the temperature data training samples are used as input data of a temperature prediction model, and the last 1 temperature data in the temperature data training samples are used as output data of the temperature prediction model; all temperature data training samples form a temperature training set, the temperature training set is sent to a parameter initialized temperature prediction model for training, whether a first training condition is met or not is judged, and if the first training condition is met, a trained temperature prediction model is output; otherwise, continuing to train the temperature prediction model through the temperature training set.
3. The method for predicting abnormal faults of a power distribution cable according to claim 2, wherein the power data prediction model is built based on an LSTM model and comprises T power data prediction LSTM units, T-1 first temperature correction layers, 1 second full connection layer, 1 second temperature correction layer and 1 second output layer, wherein the power data prediction LSTM units are used for extracting time sequence characteristics in power data prediction sets, and power hiding characteristic vectors are transmitted between adjacent power data prediction LSTM units; the first temperature correction layer is used for correcting the power hiding characteristic vector through the temperature data time sequence set; the second full connection layer is used for extracting the characteristics of the output of all the power data prediction LSTM units so as to construct the predicted power data to be processed; the second temperature correction layer is used for correcting the predicted power data to be processed through the predicted temperature data set so as to construct predicted power data; the second output layer is used for outputting the predicted power data.
4. The method for predicting abnormal faults of a power distribution cable according to claim 3, wherein the power data prediction set, the temperature data time sequence set and the predicted temperature data set are sent to a power data prediction model for processing, and predicted power data corresponding to a next monitoring time point is output, and the method specifically comprises the following steps:
the method comprises the steps of inputting power data in a power data prediction set into T power data prediction LSTM units one by one according to a time sequence for processing, correcting power hiding feature vectors transmitted between adjacent power data prediction LSTM units through a first temperature correction layer during processing, specifically operating to select a temperature data set acquired by the power data input to the power data prediction LSTM units of the adjacent power data prediction LSTM units from a temperature data time sequence set, performing point multiplication on the selected temperature data set and the first temperature correction weight vector, outputting a first temperature correction vector, performing point multiplication on the first temperature correction vector and the power hiding feature vector output by the power data prediction LSTM units of the former, outputting a power hiding feature intermediate vector, performing residual connection and standardization on the power hiding feature vector output by the power data prediction LSTM units of the former, and then, serving as a power hiding feature vector input by the next power data prediction LSTM unit;
the output of all the power data prediction LSTM units is sent to a second full-connection layer for processing, and the predicted power data to be processed is output;
In the second temperature correction layer, performing point multiplication on the predicted temperature data set and a second temperature correction weight vector, outputting a second temperature correction vector, performing point multiplication on the second temperature correction vector and predicted power data to be processed, outputting corrected predicted power data, and performing residual connection and standardization on the corrected predicted power data and the predicted power data to be processed to generate predicted power data;
and finally, outputting predicted power data corresponding to the next monitoring time point through the second output layer.
5. The method for predicting abnormal faults in a power distribution cable of claim 4, including the steps of:
Acquiring a plurality of electric power data training samples, wherein T+1 electric power data arranged according to a time sequence are stored in the electric power data training samples, the first T electric power data in the electric power data training samples are used as input data of an electric power data prediction model, and the last 1 electric power data in the electric power data training samples are used as output data of the electric power data prediction model; for each power data training sample, acquiring a temperature data time sequence set training sample corresponding to the power data training sample, wherein T+1 temperature data sets arranged according to time sequence are stored in the temperature data time sequence set training sample, and the time corresponding to the temperature data set in the temperature data time sequence set training sample corresponds to the time corresponding to the power data in the power data training sample one by one; the method comprises the steps that a power data training sample and a corresponding temperature data time sequence set training sample form a power data prediction sample, all power data prediction samples form a power data training set, then the power data training set is sent into a parameter initialized power data prediction model to be trained, during the period, the first T-1 temperature data sets in the temperature data time sequence set training sample are used as input data of T-1 first temperature correction layers in the power data prediction model, the last temperature data set in the temperature data time sequence set training sample is used as input data of a second temperature correction layer in the power data prediction model, whether the second training condition is met is judged, and if the second training condition is met, the trained power data prediction model is output; otherwise, continuing to train the power data prediction model through the power data training set.
6. The abnormal fault prediction method for the power distribution cable according to claim 5, wherein a cable abnormal fault detection model is established based on a BP neural network, and super parameters of the cable abnormal fault detection model are adjusted through a particle swarm optimization algorithm.
7. The method for predicting abnormal faults in a power distribution cable as claimed in claim 6, wherein the training of the cable abnormal fault detection model specifically comprises the following steps:
Acquiring a plurality of pieces of electric power data marked with the cable abnormal fault detection results, forming a cable abnormal detection training set by all pieces of electric power data marked with the cable abnormal fault detection results, sending the cable abnormal detection training set into a cable abnormal fault detection model initialized by parameters for training, judging whether a third training condition is met, and outputting the trained cable abnormal fault detection model if the third training condition is met; otherwise, continuing to train the cable abnormal fault detection model through the cable abnormal detection training set.
8. An abnormal-fault prediction system for a power distribution cable, wherein the system is used for an abnormal-fault prediction method for a power distribution cable according to any one of claims 1 to 7, and the system comprises:
the power data acquisition module is used for acquiring power data at the current monitoring time point;
the temperature data acquisition module is used for acquiring temperature data uploaded by the set temperature monitoring points and forming a temperature data set from the temperature data uploaded by all the currently acquired temperature monitoring points;
The power data prediction set construction module is used for forming a power data prediction set from currently acquired power data and power data acquired for the previous T-1 times;
The temperature data time sequence set construction module is used for forming a temperature data time sequence set by the current acquired temperature data set and the temperature data set acquired for the previous T-1 times;
The predicted temperature data set construction module is used for forming a temperature data predicted set by the currently acquired temperature data and the temperature data acquired for the previous T-1 time aiming at each temperature monitoring point, sending the temperature data predicted set into a temperature prediction model for processing, and outputting predicted temperature data corresponding to each temperature monitoring point; then, the predicted temperature data corresponding to all the temperature monitoring points are formed into a predicted temperature data set;
The power data prediction module is used for sending the power data prediction set, the temperature data time sequence set and the predicted temperature data set into the power data prediction model for processing and outputting predicted power data corresponding to the next monitoring time point;
And the cable abnormal fault detection module is used for sending the predicted power data into the cable abnormal fault detection model for processing and outputting a cable abnormal fault detection result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410478994.1A CN118091328B (en) | 2024-04-20 | Abnormal fault prediction method and system for distribution cable |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410478994.1A CN118091328B (en) | 2024-04-20 | Abnormal fault prediction method and system for distribution cable |
Publications (2)
Publication Number | Publication Date |
---|---|
CN118091328A true CN118091328A (en) | 2024-05-28 |
CN118091328B CN118091328B (en) | 2024-07-05 |
Family
ID=
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0572248A (en) * | 1991-09-17 | 1993-03-23 | Fuji Electric Co Ltd | Method for locating fault point of three-phase cable |
WO2000014557A1 (en) * | 1998-09-08 | 2000-03-16 | Ea Technology Limited | Locating underground power cable faults |
CN108061843A (en) * | 2017-12-27 | 2018-05-22 | 温岭市创嘉信息科技有限公司 | A kind of equipment for Cable Fault Location |
WO2022048168A1 (en) * | 2020-09-03 | 2022-03-10 | 上海上讯信息技术股份有限公司 | Training method and device for failure prediction neural network model |
CN115526292A (en) * | 2022-09-27 | 2022-12-27 | 福州亿力电力工程有限公司配电工程分公司 | Cable joint fault prediction method based on temperature difference long-time memory network learning |
US20230003198A1 (en) * | 2019-11-25 | 2023-01-05 | Envision Digital International Pte, Ltd. | Method and apparatus for detecting fault, method and apparatus for training model, and device and storage medium |
CN116310782A (en) * | 2022-12-06 | 2023-06-23 | 广州番禺电缆集团有限公司 | Overhead cable fault prediction system |
CN117214774A (en) * | 2023-08-22 | 2023-12-12 | 广州番禺电缆集团有限公司 | Online fault detection device, method and equipment for XLPE cable |
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH0572248A (en) * | 1991-09-17 | 1993-03-23 | Fuji Electric Co Ltd | Method for locating fault point of three-phase cable |
WO2000014557A1 (en) * | 1998-09-08 | 2000-03-16 | Ea Technology Limited | Locating underground power cable faults |
CN108061843A (en) * | 2017-12-27 | 2018-05-22 | 温岭市创嘉信息科技有限公司 | A kind of equipment for Cable Fault Location |
US20230003198A1 (en) * | 2019-11-25 | 2023-01-05 | Envision Digital International Pte, Ltd. | Method and apparatus for detecting fault, method and apparatus for training model, and device and storage medium |
WO2022048168A1 (en) * | 2020-09-03 | 2022-03-10 | 上海上讯信息技术股份有限公司 | Training method and device for failure prediction neural network model |
CN115526292A (en) * | 2022-09-27 | 2022-12-27 | 福州亿力电力工程有限公司配电工程分公司 | Cable joint fault prediction method based on temperature difference long-time memory network learning |
CN116310782A (en) * | 2022-12-06 | 2023-06-23 | 广州番禺电缆集团有限公司 | Overhead cable fault prediction system |
CN117214774A (en) * | 2023-08-22 | 2023-12-12 | 广州番禺电缆集团有限公司 | Online fault detection device, method and equipment for XLPE cable |
Non-Patent Citations (1)
Title |
---|
汪梅: "基于神经网络的三相电缆故障预测定位系统", 西安科技学院学报, no. 02, 30 June 2004 (2004-06-30) * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhou et al. | Two-stage aging trajectory prediction of LFP lithium-ion battery based on transfer learning with the cycle life prediction | |
Li et al. | Big data driven lithium-ion battery modeling method based on SDAE-ELM algorithm and data pre-processing technology | |
CN111612050B (en) | Method for detecting remote measurement data abnormality | |
JP3718765B2 (en) | Plant diagnostic equipment | |
CN112763967A (en) | BiGRU-based intelligent electric meter metering module fault prediction and diagnosis method | |
CN117154263A (en) | Lithium battery cascade utilization charging and discharging system and control method | |
CN112816881A (en) | Battery differential pressure abnormality detection method, battery differential pressure abnormality detection device and computer storage medium | |
CN116418117A (en) | Equipment detection system for intelligent power grid | |
CN116467950A (en) | Unmanned aerial vehicle flight data anomaly detection method based on uncertain characterization | |
CN118091328B (en) | Abnormal fault prediction method and system for distribution cable | |
Chen et al. | State of health estimation of lithium-ion batteries based on equivalent circuit model and data-driven method | |
CN115878992A (en) | Monitoring method and monitoring system for comprehensive pipe rack power supply system | |
Wang et al. | Large-scale field data-based battery aging prediction driven by statistical features and machine learning | |
CN118091328A (en) | Abnormal fault prediction method and system for distribution cable | |
CN109697330B (en) | Fault detection system design method for information physical system | |
CN116488578A (en) | Photovoltaic equipment analysis method based on AI visual assistance | |
CN113570473B (en) | Equipment fault monitoring method, device, computer equipment and storage medium | |
CN116338545A (en) | Method, system, equipment and medium for identifying metering error state of current transformer | |
Lu et al. | Deep learning to predict battery voltage behavior after uncertain cycling-induced degradation | |
CN115184734A (en) | Power grid line fault detection method and system | |
Garmaev et al. | Deep Koopman Operator-based degradation modelling | |
CN112818870A (en) | Method for predicting residual life of bearing based on gated neural network framework | |
US20220382940A1 (en) | Simulation of a performance of an energy storage | |
CN111583594A (en) | Distribution box early warning system and early warning method | |
CN110866652A (en) | Online PMU data error correction method and system based on LSTM model |
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 |