Power distribution network active first-aid repair auxiliary decision-making method and auxiliary system

Abstract

The invention relates to an auxiliary decision-making method and an auxiliary system for active first-aid repair of a power distribution network. A thermal circuit model is obtained based on a 10kV three-core cable joint structure of the power distribution network; based on the thermal circuit model, decomposing the joint temperature sequence to obtain a plurality of IMF components and residual components, and constructing a GRNN neural network prediction model for each sub-sequence component; performing temperature anomaly detection according to all the GRNN neural network prediction models; and when it is determined that the temperature is abnormal, an alarm is given, so that a worker can carry out active first-aid repair on the cable in the power distribution network conveniently. An integrated and intelligent temperature monitoring and explosion-proof device is formed. The method is beneficial for an electric power company to manage the governed cable, and can be convenient for electric power personnel to improve the resistance near the hot spot of the joint, carry out repair in time, and reduce the operation risk of the cable.

Description

Technical Field

[0001] The present invention relates to the technical field of active emergency repair of distribution network, and in particular to an auxiliary decision-making method and auxiliary system for active emergency repair of distribution network. Background Art

[0002] In the distribution network, the core temperature of the power cable is often too high, which affects the safety of electricity use. The core temperature of the power cable is closely related to its insulation performance, the load condition of the distribution system, and the laying environment of the cable. Obtaining the core temperature and its changes can not only grasp the operating conditions of the cable, but also enable the power department to regulate the current carrying capacity of the cable line. Through detection technical means, the temperature of the surface of the cable joint is generally obtained. Under existing technical conditions, it is often difficult to obtain the temperature of the inside of the joint structure and the core conductor during operation. The outer surface of the cable is the boundary for heat transfer between it and the environment, and changes in ambient temperature will also cause the temperature of the outer surface of the cable to change, thereby affecting the temperature of the core conductor.

[0003] Therefore, in the distribution network, it is necessary to obtain the cable core temperature in real time to provide auxiliary decision-making for proactive repairs of the distribution network. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a distribution network active emergency repair auxiliary decision-making method and system.

[0005] To solve the above problems, the technical solution adopted by the present invention is:

[0006] On the one hand, a method for assisting decision making for active emergency repair of a distribution network is provided, the method comprising:

[0007] Step 1: Obtain a thermal circuit model based on the 10kV three-core cable joint structure of the distribution network;

[0008] Step 2: Based on the thermal circuit model, the joint temperature sequence is decomposed to obtain multiple IMF components and residual components, and a GRNN neural network prediction model is constructed for each subsequence component;

[0009] Step 3: Perform temperature anomaly detection based on all GRNN neural network prediction models;

[0010] Step 4: When it is determined that the temperature is abnormal, an alarm is issued to facilitate the staff to actively repair the cables in the distribution network.

[0011] As an implementation method of the invention, step one includes: dividing the 10kV three-core cable joint structure of the distribution network, considering thermal inertia, and establishing a transient thermal circuit basic model of the joint, wherein the model includes the thermal capacity of each part of the structure.

[0012] As an implementation mode of the invention, for any subsequence component, the GRNN neural network prediction model is constructed as follows:

[0013] Obtaining the subsequence component as external input;

[0014] The neurons in the input layer directly receive the input and pass it forward to the pattern layer;

[0015] The neurons in the pattern layer correspond to different learning samples, and the transfer function of neuron i in this layer is:

[0016]

[0017] In the formula, p i is the Gaussian kernel function; x is the network input variable; x i is the learning sample corresponding to neuron i; σ is the smoothing factor, and the output result of neuron i in the pattern layer is: the exponential form of the square of the Euclid distance between the input variable and its corresponding sample;

[0018] The summation layer performs arithmetic summation and weighted summation on the outputs of all mode layers;

[0019] The output layer divides the weighted sum output of the neurons in the sum layer by the arithmetic sum output, and the output of each neuron corresponds to the jth element in the output result:

[0020]

[0021] As an implementation mode of the invention, the summation layer uses the following method to perform weighted summation:

[0022] The connection coefficient between the i-th neuron in the pattern layer and the j-th molecular summation neuron in the summation layer is the j-th element y in the i-th output sample ij :

[0023]

[0024] As an implementation mode of the invention, in step three, anomaly detection is performed using a temperature feature sequence obtained using an isolation forest algorithm.

[0025] On the other hand, a distribution network active emergency repair auxiliary system is provided, the system comprising:

[0026] A thermal circuit model generation module is used to obtain a thermal circuit model based on a 10kV three-core cable connector structure of a distribution network;

[0027] A construction module is used to decompose the joint temperature sequence based on the thermal circuit model to obtain multiple IMF components and residual components, and to construct a GRNN neural network prediction model for each subsequence component;

[0028] A detection module, which is used to detect temperature anomalies based on all GRNN neural network prediction models;

[0029] The alarm module is used to issue an alarm when it is determined that the temperature is abnormal, so as to facilitate the staff to actively repair the cables in the distribution network.

[0030] As an embodiment of the invention, the thermal circuit model generation module is used to divide the distribution network 10kV three-core cable joint structure, consider thermal inertia, and establish a transient thermal circuit basic model of the joint, which includes the thermal capacity of each part of the structure.

[0031] As an embodiment of the invention, the building block is also used for:

[0032] Obtaining the subsequence component as external input;

[0033] The neurons in the input layer directly receive the input and pass it forward to the pattern layer;

[0034] The neurons in the pattern layer correspond to different learning samples, and the transfer function of neuron i in this layer is:

[0035]

[0036] In the formula, p i is the Gaussian kernel function; x is the network input variable; x i is the learning sample corresponding to neuron i; σ is the smoothing factor, and the output result of neuron i in the pattern layer is: the exponential form of the square of the Euclid distance between the input variable and its corresponding sample;

[0037] The summation layer performs arithmetic summation and weighted summation on the outputs of all mode layers;

[0038] The output layer divides the weighted sum output of the neurons in the sum layer by the arithmetic sum output, and the output of each neuron corresponds to the jth element in the output result:

[0039]

[0040] As an implementation mode of the invention, the summation layer uses the following method to perform weighted summation:

[0041] The connection coefficient between the i-th neuron in the pattern layer and the j-th molecular summation neuron in the summation layer is the j-th element y in the i-th output sample ij :

[0042]

[0043] As an implementation mode of the invention, the detection module uses the temperature feature sequence obtained by the isolation forest algorithm to perform anomaly detection.

[0044] The beneficial effects of adopting the above technical solution are:

[0045] The present invention provides a distribution network active repair auxiliary decision-making method and auxiliary system, which obtains a thermal circuit model based on the 10kV three-core cable joint structure of the distribution network; based on the thermal circuit model, the joint temperature sequence is decomposed and processed to obtain multiple IMF components and residual components, and a GRNN neural network prediction model is constructed for each subsequence component; temperature anomaly detection is performed according to all GRNN neural network prediction models; when it is determined that the temperature is abnormal, an alarm is issued to facilitate the staff to actively repair the cables in the distribution network. It forms an integrated, intelligent temperature monitoring and explosion-proof device. The proposed method is conducive to the power company to manage the cables under its jurisdiction, and can facilitate power personnel to improve the resistance near the joint hot spots, carry out repairs in time, and reduce the risk of cable operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a structural schematic diagram of a basic model of a transient thermal circuit of a joint provided in an embodiment of the present invention.

[0047] Figure 2 It is a schematic diagram of the anatomical structure of the cable intermediate joint provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is clearly and completely described below in conjunction with specific embodiments.

[0049] Example 1

[0050] An embodiment of the present invention provides a distribution network active emergency repair auxiliary decision-making method, the method comprising:

[0051] Step 1: Obtain a thermal circuit model based on the 10kV three-core cable joint structure of the distribution network;

[0052] This step includes: dividing the structure of the 10kV three-core cable joint of the distribution network, considering the thermal inertia, and establishing the transient thermal circuit basic model of the joint. The model includes the thermal capacity of each part of the structure, such as Figure 1 shown.

[0053] On the basis of the basic model of the mold, the thinner parts of the joint with small heat capacity and thermal resistance are ignored or merged with the adjacent structures, and finally a simplified thermal circuit model is obtained.

[0054] Step 2: Based on the thermal circuit model, the joint temperature sequence is decomposed and processed to obtain multiple IMF components and residual components, and a GRNN neural network prediction model is constructed for each subsequence component.

[0055] The actual temperature that a cable joint can reach is not only related to its own insulation level, but also to many factors such as ambient temperature, current, and air environment. The joint temperature series is nonlinear and non-stationary to a certain extent, and the conventional prediction method has very limited effect on improving the prediction accuracy. The GRNN model can give full play to the advantages of the GRNN neural network in predicting time series data.

[0056] For any subsequence component, the GRNN neural network prediction model is constructed as follows:

[0057] Step S201, obtaining the subsequence component and taking it as external input;

[0058] Step S202: The neurons in the input layer directly receive the input and pass it forward to the pattern layer;

[0059] Step S203: The neurons in the pattern layer correspond to different learning samples, and the transfer function of the neuron i in this layer is:

[0060]

[0061] In the formula, p i is the Gaussian kernel function; x is the network input variable; x i is the learning sample corresponding to neuron i; σ is the smoothing factor, and the output result of neuron i in the pattern layer is: the exponential form of the square of the Euclid distance between the input variable and its corresponding sample;

[0062] Step S204: The summing layer performs arithmetic summing and weighted summing on the outputs of all mode layers;

[0063] Specifically, the weighted summation method is as follows:

[0064] The connection coefficient between the i-th neuron in the pattern layer and the j-th molecular summation neuron in the summation layer is the j-th element y in the i-th output sample ij :

[0065]

[0066] Step S205: The output layer divides the weighted sum output of the neurons in the sum layer by the arithmetic sum output. The output of each neuron corresponds to the jth element in the output result:

[0067]

[0068] Step 3: Perform temperature anomaly detection based on all GRNN neural network prediction models;

[0069] The temperature feature sequence obtained by the isolation forest algorithm is used for anomaly detection.

[0070] In the process of collecting, transmitting, storing and processing cable joint temperature data, accidents may occur in various links such as sensors and transmission networks, affecting the authenticity and effectiveness of data collection. Excessively high joint temperature and extremely fast temperature rise are potential fault signals. These temperature data are numerically different from the data in the normal state and can be regarded as outliers in the time series. The occurrence of abnormal temperature values ​​is related to the operation of the joint and on-site problems. The abnormal temperature detection of the joint is completed by using anomaly detection, a machine learning technology that closely links the actual engineering. Combining time series and unsupervised learning algorithms can solve the problems of low detection accuracy, misjudgment and missed judgment in the traditional single threshold judgment method.

[0071] Step 4: When it is determined that the temperature is abnormal, an alarm is issued to facilitate the staff to actively repair the cables in the distribution network.

[0072] In the present invention, the score of each data sample is calculated. When the obtained score exceeds the range, it is defined as an abnormal score, so whether it is abnormal or not can be judged by its size.

[0073] The present invention provides a distribution network active repair auxiliary decision-making method, which obtains a thermal circuit model based on the 10kV three-core cable joint structure of the distribution network; based on the thermal circuit model, the joint temperature sequence is decomposed and processed to obtain multiple IMF components and residual components, and a GRNN neural network prediction model is constructed for each subsequence component; temperature anomaly detection is performed according to all GRNN neural network prediction models; when it is determined that the temperature is abnormal, an alarm is issued to facilitate the staff to actively repair the cables in the distribution network. It forms an integrated, intelligent temperature monitoring and explosion-proof device. The proposed method is conducive to the power company to manage the cables under its jurisdiction, and can facilitate power personnel to improve the resistance near the joint hot spots, carry out repairs in time, and reduce the risk of cable operation.

[0074] Example 2

[0075] The present invention provides a distribution network active emergency repair auxiliary system, the system comprising:

[0076] A thermal circuit model generation module is used to obtain a thermal circuit model based on a 10kV three-core cable connector structure of a distribution network;

[0077] A construction module is used to decompose the joint temperature sequence based on the thermal circuit model to obtain multiple IMF components and residual components, and to construct a GRNN neural network prediction model for each subsequence component;

[0078] A detection module, which is used to detect temperature anomalies based on all GRNN neural network prediction models;

[0079] The alarm module is used to issue an alarm when it is determined that the temperature is abnormal, so as to facilitate the staff to actively repair the cables in the distribution network.

[0080] Furthermore, the thermal circuit model generation module is used to divide the distribution network 10kV three-core cable joint structure, consider thermal inertia, and establish a transient thermal circuit basic model of the joint, which includes the thermal capacity of each part of the structure.

[0081] Furthermore, the building block is also used for:

[0082] Obtaining the subsequence component as external input;

[0083] The neurons in the input layer directly receive the input and pass it forward to the pattern layer;

[0084] The neurons in the pattern layer correspond to different learning samples, and the transfer function of neuron i in this layer is:

[0085]

[0086] In the formula, p i is the Gaussian kernel function; x is the network input variable; x i is the learning sample corresponding to neuron i; σ is the smoothing factor, and the output result of neuron i in the pattern layer is: the exponential form of the square of the Euclid distance between the input variable and its corresponding sample;

[0087] The summation layer performs arithmetic summation and weighted summation on the outputs of all mode layers;

[0088] The output layer divides the weighted sum output of the neurons in the sum layer by the arithmetic sum output, and the output of each neuron corresponds to the jth element in the output result:

[0089]

[0090] Furthermore, the summation layer performs weighted summation using the following method:

[0091] The connection coefficient between the i-th neuron in the pattern layer and the j-th molecular summation neuron in the summation layer is the j-th element y in the i-th output sample ij :

[0092]

[0093] Furthermore, the detection module uses the temperature feature sequence obtained by the isolation forest algorithm to perform anomaly detection.

[0094] The present invention provides a distribution network active repair auxiliary system, which obtains a thermal circuit model based on the 10kV three-core cable joint structure of the distribution network; based on the thermal circuit model, the joint temperature sequence is decomposed and processed to obtain multiple IMF components and residual components, and a GRNN neural network prediction model is constructed for each subsequence component; temperature anomaly detection is performed according to all GRNN neural network prediction models; when it is determined that the temperature is abnormal, an alarm is issued to facilitate the staff to actively repair the cables in the distribution network. It forms an integrated, intelligent temperature monitoring and explosion-proof device. The proposed method is conducive to the power company to manage the cables under its jurisdiction, and can facilitate power personnel to improve the resistance near the joint hot spots, carry out repairs in time, and reduce the risk of cable operation.

Claims

1. A distribution network active repair auxiliary decision-making method, characterized in that: The method comprises: Step 1: Obtain a thermal circuit model based on the 10kV three-core cable joint structure of the distribution network; Step 2: Based on the thermal circuit model, the joint temperature sequence is decomposed to obtain multiple IMF components and residual components, and a GRNN neural network prediction model is constructed for each subsequence component; Step 3: Perform temperature anomaly detection based on all GRNN neural network prediction models; Step 4: When it is determined that the temperature is abnormal, an alarm is issued to facilitate the staff to actively repair the cables in the distribution network.

2. A distribution network active repair auxiliary decision-making method according to claim 1, characterized in that: The step 1 includes: dividing the distribution network 10kV three-core cable joint structure, considering thermal inertia, and establishing a transient thermal circuit basic model of the joint, wherein the model includes the thermal capacity of each part of the structure.

3. A distribution network active repair auxiliary decision-making method according to claim 2, characterized in that: For any subsequence component, the GRNN neural network prediction model is constructed as follows: Obtaining the subsequence component as external input; The neurons in the input layer directly receive the input and pass it forward to the pattern layer; The neurons in the pattern layer correspond to different learning samples, and the transfer function of neuron i in this layer is: In the formula, p i is the Gaussian kernel function; x is the network input variable; x i is the learning sample corresponding to neuron i; σ is the smoothing factor, and the output result of neuron i in the pattern layer is: the exponential form of the square of the Euclid distance between the input variable and its corresponding sample; The summation layer performs arithmetic summation and weighted summation on the outputs of all mode layers; The output layer divides the weighted sum output of the neurons in the sum layer by the arithmetic sum output, and the output of each neuron corresponds to the jth element in the output result:

4. A distribution network active repair auxiliary decision-making method according to claim 3, characterized in that: The summation layer performs weighted summation using the following method: The connection coefficient between the i-th neuron in the pattern layer and the j-th molecular summation neuron in the summation layer is the j-th element y in the i-th output sample ij :

5. The method for assisting decision-making in active emergency repair of distribution network according to claim 1, characterized in that: In the step three, the temperature feature sequence obtained by the isolation forest algorithm is used to perform anomaly detection.

6. A distribution network active emergency repair auxiliary system, characterized in that: The system comprises: A thermal circuit model generation module is used to obtain a thermal circuit model based on a 10kV three-core cable connector structure of a distribution network; A construction module is used to decompose the joint temperature sequence based on the thermal circuit model to obtain multiple IMF components and residual components, and to construct a GRNN neural network prediction model for each subsequence component; A detection module, which is used to detect temperature anomalies based on all GRNN neural network prediction models; The alarm module is used to issue an alarm when it is determined that the temperature is abnormal, so as to facilitate the staff to actively repair the cables in the distribution network.

7. The active repair auxiliary system for distribution network according to claim 6, characterized in that: The thermal circuit model generation module is used to divide the distribution network 10kV three-core cable joint structure, consider thermal inertia, and establish a transient thermal circuit basic model of the joint, which includes the thermal capacity of each part of the structure.

8. The active repair auxiliary system for distribution network according to claim 7, characterized in that: The building blocks are also used to: Obtaining the subsequence component as external input; The neurons in the input layer directly receive the input and pass it forward to the pattern layer; The neurons in the pattern layer correspond to different learning samples, and the transfer function of neuron i in this layer is: In the formula, p i is the Gaussian kernel function; x is the network input variable; x i is the learning sample corresponding to neuron i; σ is the smoothing factor, and the output result of neuron i in the pattern layer is: the exponential form of the square of the Euclid distance between the input variable and its corresponding sample; The summation layer performs arithmetic summation and weighted summation on the outputs of all mode layers; The output layer divides the weighted sum output of the neurons in the sum layer by the arithmetic sum output, and the output of each neuron corresponds to the jth element in the output result:

9. The active repair auxiliary system for distribution network according to claim 8, characterized in that: The summation layer performs weighted summation using the following method: The connection coefficient between the i-th neuron in the pattern layer and the j-th molecular summation neuron in the summation layer is the j-th element y in the i-th output sample ij :

10. The distribution network active repair auxiliary system according to claim 1, characterized in that: The detection module uses the temperature feature sequence obtained by the isolation forest algorithm to perform anomaly detection.

Images