Power outage prediction method and system under extreme disasters in distribution network based on BP neural network
A BP neural network-based method integrates meteorological and network data to predict power outages, enhancing accuracy and enabling timely response to extreme disasters in distribution networks.
Patent Information
- Application Number
- CN202210261777.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-16
- Publication Date
- 2024-12-20
- Estimated Expiration
- 2042-03-16
AI Technical Summary
Existing meteorological predictions alone are insufficient for accurately predicting power outages in distribution networks during extreme disasters, necessitating a method that integrates meteorological data with distribution network data to enhance prediction accuracy.
A method using a BP neural network to integrate fault and non-fault sample libraries, incorporating meteorological and operating status parameters, to predict power outages in distribution networks by training a power outage prediction model.
Enables accurate prediction of power outage probability and duration under extreme disasters, supporting rapid response and mitigation strategies in distribution networks.
Smart Images
Figure CN114548601B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of power grid disaster prevention, and in particular relates to a method and system for predicting power outages under extreme disasters in a distribution network based on a BP neural network. Background Art
[0002] With global climate change, extreme disasters have occurred frequently in recent years, causing particularly serious disasters to the power grid. The safe operation of the distribution network is related to the reliability of power supply. The distribution network is extremely susceptible to extreme disasters and failures, resulting in power outages. Distribution networks are frequently affected by different extreme disasters, resulting in pole breakage, wire breakage, and hardware damage, and their severity is far greater than the impact of extreme disasters on the main grid's power transmission and transformation equipment. While affecting the safe and stable operation of the power grid, it also puts forward high requirements for emergency repairs and rapid power restoration. Therefore, under the premise of the stable demand for power load, the urgency of the accurate prediction of power outages in distribution networks under extreme disasters has significantly increased. It is urgent to propose a method for predicting power outages in distribution networks under extreme disasters.
[0003] At present, in terms of extreme disaster prediction, professional meteorological institutions have been able to make relatively accurate predictions of some key meteorological factors, such as freezing rain, through years of development. Other meteorological elements have also achieved a certain degree of accuracy. Combined with the continuously developing mesoscale numerical weather forecast system, the refined meteorological numerical forecast of 3km*3km has been studied. Combined with the downscaling technology, a downscaling prediction result of 1km*1km has been formed. Although the scale of meteorological prediction is getting smaller and smaller, meteorological prediction alone cannot meet the prediction and early warning needs of extreme meteorological disaster-related failures in power grids, especially distribution networks. It is very necessary to combine the existing meteorological prediction results, based on the historical extreme disaster failure disaster samples of distribution networks, use statistical methods, and combine the ledger data and operation status data of distribution networks to realize the prediction of power outages under extreme disasters in distribution networks. Summary of the invention
[0004] In order to overcome the above-mentioned defects, the object of the present invention is to provide a method and system for predicting power outages under extreme disasters in distribution networks based on BP neural network.
[0005] To achieve the above object, the present invention adopts the following technical solution:
[0006] A method for predicting power outages under extreme disasters in a distribution network based on a BP neural network comprises the following steps:
[0007] Step 1: Collect the fault set of power outages in distribution networks caused by extreme natural disasters;
[0008] Step 2: For each power outage fault in the fault set, match the entity parameters, meteorological parameters, extreme disaster type, operating status parameters, and power outage duration data when the fault occurs, and establish a fault sample library A under historical extreme disasters in the distribution network;
[0009] Step 3: For each type of extreme disaster in statistics, collect the meteorological parameters and operating status parameters of the distribution network under the same main body parameters during the meteorological disaster process when no power outage occurs, and form a non-fault sample library B under the historical extreme disasters of the distribution network. The power outage duration in the non-fault sample library B is marked as 0;
[0010] Step 4: Perform numerical preprocessing on the ontological parameters, meteorological parameters, extreme disaster types, operating status parameters and power outage duration after the fault occurs in the fault sample library A and the non-fault sample library B, and integrate the two preprocessed sample libraries into the training sample library X;
[0011] Step 5: Based on the training sample library X in step 4, the BP neural network method is used to train the power outage probability and power outage duration of the distribution network under extreme disasters, and a power outage prediction model for the distribution network under extreme disasters based on the BP neural network is established;
[0012] Step 6: In the prediction stage, based on the power outage prediction model trained in step 5, the extreme disaster type, meteorological parameters, distribution network parameters and operating status parameters at a certain time that needs to be predicted are input to obtain the possible power outage probability and duration, and finally realize the prediction of power outage probability and duration under extreme disasters in the distribution network.
[0013] Preferably, the main body parameters include tower information and line information of the distribution network.
[0014] Preferably, the tower information includes: tower type, longitude, latitude, altitude, geological environment information, horizontal span, vertical span, and tower insulator string model, number of strings, number of insulator pieces, and tension tower rotation angle.
[0015] Preferably, the line information includes: voltage level, line specification, line type, number of loops, transmission length, design wind speed, design ice thickness, split number and split gap.
[0016] Preferably, the extreme disaster types include: strong winds, heavy rains, snowfall, and sleet.
[0017] Preferably, the meteorological parameters include: temperature, humidity, instantaneous wind speed, average wind speed within 1 hour, instantaneous wind direction, angle between wind direction and line, whether there is freezing rain, whether there is precipitation, precipitation amount, and Doppler radar original parameters.
[0018] Preferably, the operating status parameters include current information and switch status information.
[0019] Preferably, the steps of establishing the power outage prediction model under extreme disasters of the distribution network based on the BP neural network in step 5 are as follows:
[0020] 5.1 Design the BP network into three layers, where the input layer is S, the output layer is O, and the hidden layer is H. The calculation formula for the number of neurons in the hidden layer is:
[0021]
[0022] Among them, a is the number of neurons in the input layer, that is, the number of input parameters; c is the number of neurons in the output layer, that is, the number of output parameters; b is the number of neurons in the hidden layer; Υ is the adjustment constant, which ranges from 1 to 10;
[0023] 5.2 Training sample set X of distribution network under extreme disasters:
[0024] X={(X1,y 11 ,y 12 ),(X2,y 21 ,y 22 ),...(X i ,y i1 ,y i2 ),(X N ,y N1 ,y N2 )}
[0025] Among them, X i =(x i1 ,x i2 ,...x iM ), X i is the input parameter of the ith sample, i.e., the distribution network parameters, meteorological parameters, extreme disaster type and operating status parameters when the ith extreme disaster sample occurs, x iM is the Mth parameter in the i-th sample. Each sample has M parameters, y i1 is the power outage state of the distribution network when the i-th extreme disaster sample occurs, y i2 is the power outage duration of the distribution network when the i-th extreme disaster sample occurs;
[0026] 5.3 After BP neural network training, we obtain an a×b prediction model weight matrix ω1 and a b×c prediction model weight matrix ω2.
[0027] Preferably, the steps for predicting the power outage probability and power outage duration under extreme disasters in the distribution network described in step 6 are as follows:
[0028] Based on the prediction model matrix ω1 obtained through training in step 5 and the input result of the input layer, we get:
[0029] H(i)=ω1[1,i].S(1)+ω1[2,i].S(2)+...+ω1[k,i].S(k)+...+ω1[a,i].S(a)
[0030] Among them, H(i) is the i-th output result of the hidden layer, ω1[k,i] represents the value of the k-th row and i-th column of the weight matrix ω1, that is, the connection weight from the k-th input neuron to the i-th hidden neuron, and S(k) is the k-th parameter value of the input layer;
[0031] Based on the prediction model matrix ω2 obtained through training in step 5 and the output results of the hidden layer, we obtain:
[0032] O(i)=ω2[1,i].H(1)+ω2[2,i].H(2)+...+ω2[k,i].H(k)+...+ω2[b,i].H(b)
[0033] O(i) is the i-th output result of the output layer, ω2[k,i] represents the value of the k-th row and i-th column of the weight matrix ω2, that is, the connection weight from the k-th hidden neuron to the i-th output neuron;
[0034] The output layer has two output results, namely, the power outage probability and power outage duration of the distribution network under extreme disasters.
[0035] A power outage prediction system for distribution network under extreme disasters based on BP neural network, comprising:
[0036] The information module collects data on meteorological parameters, extreme disaster types, operating status parameters, and power outage duration after the occurrence of power outages in distribution networks with the same physical parameters under extreme natural disasters, and forms a fault sample library under historical extreme disasters in distribution networks and a non-fault sample library under historical extreme disasters in distribution networks;
[0037] The data preprocessing module performs numerical preprocessing on the fault sample library and the non-fault sample library's body parameters, meteorological parameters, extreme disaster types, operating status parameters, and power outage duration after the fault occurs, and integrates them into a training sample library;
[0038] The training module uses the BP neural network method to train the power outage probability and duration of the distribution network under extreme disasters, and establishes a power outage prediction model for the distribution network under extreme disasters based on the BP neural network;
[0039] The prediction module, based on the trained power outage prediction model, inputs the extreme disaster type, meteorological parameters, distribution network parameters and operating status parameters at a certain moment to be predicted, and predicts the power outage probability and duration under extreme disasters in the distribution network.
[0040] The positive beneficial effects of the present invention are:
[0041] 1. The present invention matches the main body parameters, meteorological parameters, extreme disaster types and operating status parameters of the distribution network when the historical extreme disaster faults occur, and the power outage duration after the fault occurs through statistical analysis of the fault set under the historical extreme disasters of the distribution network. A fault sample library under the historical extreme disasters of the distribution network is established. At the same time, a non-fault sample library without faults under the same distribution network main body parameters when extreme disasters occur is incorporated. After numerical preprocessing of each parameter in the two sample libraries, they are integrated into a training sample library. With the power outage probability and the power outage duration as the output results, a BP neural network is used for training to obtain a power outage prediction model based on the BP neural network. Finally, according to the prediction model obtained by training, the power outage situation when extreme disasters occur in the future can be predicted, including the power outage probability and the power outage duration. The prediction model is suitable for disaster prevention and mitigation of distribution networks with complex situations. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 The present invention is a flow chart of a method for predicting power outages under extreme disasters in a distribution network based on a BP neural network.
[0043] Figure 2 It is the structural diagram of the BP neural network of the present invention. DETAILED DESCRIPTION
[0044] The present invention is further described below in conjunction with some specific embodiments.
[0045] Example 1
[0046] See also Figure 1 A method for predicting power outages under extreme disasters in distribution networks based on BP neural network includes the following steps:
[0047] Step 1: Collect the fault set of power outages in distribution networks caused by extreme natural disasters;
[0048] Step 2: For each power outage fault in the fault set, match the entity parameters, meteorological parameters, extreme disaster type, operating status parameters, and power outage duration data when the fault occurs, and establish a fault sample library A under historical extreme disasters in the distribution network;
[0049] Step 3: For each type of extreme disaster in statistics, collect the meteorological parameters and operating status parameters of the distribution network under the same main body parameters during the meteorological disaster process when no power outage occurs, and form a non-fault sample library B under the historical extreme disasters of the distribution network. The power outage duration in the non-fault sample library B is marked as 0;
[0050] Step 4: Perform numerical preprocessing on the ontological parameters, meteorological parameters, extreme disaster types, operating status parameters and power outage duration after the fault occurs in the fault sample library A and the non-fault sample library B, and integrate the two preprocessed sample libraries into the training sample library X;
[0051] Step 5: Based on the training sample library X in step 4, the BP neural network method is used to train the power outage probability and power outage duration of the distribution network under extreme disasters, and a power outage prediction model for the distribution network under extreme disasters based on the BP neural network is established;
[0052] Step 6: In the prediction stage, based on the power outage prediction model trained in step 5, the extreme disaster type, meteorological parameters, distribution network parameters and operating status parameters at a certain time that needs to be predicted are input to obtain the possible power outage probability and duration, and finally realize the prediction of power outage probability and duration under extreme disasters in the distribution network.
[0053] Among them, the power outage duration after the fault occurs refers to the power outage duration of the distribution network caused by the extreme disaster until the next commissioning. The power outage duration in the non-fault sample library B (that is, the sample library without power outage fault) is marked as 0.
[0054] Furthermore, the main body parameters include tower information and line information of the distribution network.
[0055] Furthermore, the tower information includes: tower type, longitude, latitude, altitude, geological environment information, horizontal span, vertical span, and the tower's insulator string model, number of strings, number of insulator pieces, and tension tower rotation angle.
[0056] Furthermore, the line information includes: voltage level, line specification, line type, number of loops, transmission length, design wind speed, design ice thickness, split number and split gap.
[0057] Furthermore, extreme disaster types include four types: strong winds, heavy rains, snowfall, and sleet.
[0058] Furthermore, meteorological parameters include: temperature, humidity, instantaneous wind speed, average wind speed within 1 hour, instantaneous wind direction, angle between wind direction and line, whether there is freezing rain, whether there is precipitation, precipitation amount, and Doppler radar original parameters.
[0059] Furthermore, the operating status parameters include current information and switch status information.
[0060] Furthermore, the steps for establishing the power outage prediction model under extreme disasters of the distribution network based on the BP neural network described in step 5 are as follows:
[0061] 5.1 Design the BP network into three layers, see Figure 2 , where the input layer is S, the output layer is O, the hidden layer is H, and the calculation formula for the number of neurons in the hidden layer is:
[0062]
[0063] Among them, a is the number of neurons in the input layer, that is, the number of input parameters; c is the number of neurons in the output layer, that is, the number of output parameters; b is the number of neurons in the hidden layer; Υ is the adjustment constant, which ranges from 1 to 10;
[0064] 5.2 Training sample set X of distribution network under extreme disasters:
[0065] X={(X1,y 11 ,y 12 ),(X2,y 21 ,y 22 ),...(X i ,y i1 ,y i2 ),(X N ,y N1 ,y N2 )}
[0066] Among them, X i =(x i1 ,x i2 ,...x iM ), X i is the input parameter of the ith sample, i.e., the distribution network parameters, meteorological parameters, extreme disaster type and operating status parameters when the ith extreme disaster sample occurs, x iM is the Mth parameter in the i-th sample. Each sample has M parameters, y i1 is the power outage state of the distribution network when the i-th extreme disaster sample occurs, including power outage and no power outage, y i2 is the power outage duration of the distribution network when the i-th extreme disaster sample occurs;
[0067] 5.3 After BP neural network training, we obtain an a×b prediction model weight matrix ω1 and a b×c prediction model weight matrix ω2.
[0068] Furthermore, the prediction steps of the power outage probability and power outage duration under extreme disasters in the distribution network described in step 6 are as follows:
[0069] Based on the prediction model matrix ω1 obtained through training in step 5 and the input result of the input layer, we get:
[0070] H(i)=ω1[1,i].S(1)+ω1[2,i].S(2)+...+ω1[k,i].S(k)+...+ω1[a,i].S(a)
[0071] Among them, H(i) is the i-th output result of the hidden layer, ω1[k,i] represents the value of the k-th row and i-th column of the weight matrix ω1, that is, the connection weight from the k-th input neuron to the i-th hidden neuron, and S(k) is the k-th parameter value of the input layer;
[0072] Based on the prediction model matrix ω2 obtained through training in step 5 and the output results of the hidden layer, we obtain:
[0073] O(i)=ω2[1,i].H(1)+ω2[2,i].H(2)+...+ω2[k,i].H(k)+...+ω2[b,i].H(b)
[0074] O(i) is the i-th output result of the output layer, ω2[k,i] represents the value of the k-th row and i-th column of the weight matrix ω2, that is, the connection weight from the k-th hidden neuron to the i-th output neuron;
[0075] The output layer has two output results, namely, the power outage probability and power outage duration of the distribution network under extreme disasters.
[0076] A power outage prediction system for distribution network under extreme disasters based on BP neural network, comprising:
[0077] The information module collects data on meteorological parameters, extreme disaster types, operating status parameters, and power outage duration after the occurrence of power outages in distribution networks with the same physical parameters under extreme natural disasters, and forms a fault sample library under historical extreme disasters in distribution networks and a non-fault sample library under historical extreme disasters in distribution networks;
[0078] The data preprocessing module performs numerical preprocessing on the fault sample library and the non-fault sample library's body parameters, meteorological parameters, extreme disaster types, operating status parameters, and power outage duration after the fault occurs, and integrates them into a training sample library;
[0079] The training module uses the BP neural network method to train the power outage probability and duration of the distribution network under extreme disasters, and establishes a power outage prediction model for the distribution network under extreme disasters based on the BP neural network;
[0080] The prediction module, based on the trained power outage prediction model, inputs the extreme disaster type, meteorological parameters, distribution network parameters and operating status parameters at a certain moment to be predicted, and predicts the power outage probability and duration under extreme disasters in the distribution network.
[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Other modifications or equivalent substitutions made to the technical solution of the present invention by ordinary technicians in the field should be included in the scope of the claims of the present invention as long as they do not depart from the spirit and scope of the technical solution of the present invention.
Claims
1. A method for predicting power outages under extreme disasters in distribution networks based on BP neural network, characterized in that: The following steps are involved: Step 1: Collect the fault set of power outages in distribution networks caused by extreme natural disasters; Step 2: For each power outage fault in the fault set, match the entity parameters, meteorological parameters, extreme disaster type, operating status parameters, and power outage duration data when the fault occurs, and establish a fault sample library A under historical extreme disasters in the distribution network; Step 3: For each type of extreme disaster in statistics, collect the meteorological parameters and operating status parameters of the distribution network under the same main body parameters during the meteorological disaster process when no power outage occurs, and form a non-fault sample library B under the historical extreme disasters of the distribution network. The power outage duration in the non-fault sample library B is marked as 0; Step 4: Perform numerical preprocessing on the ontological parameters, meteorological parameters, extreme disaster types, operating status parameters and power outage duration after the fault occurs in the fault sample library A and the non-fault sample library B, and integrate the two preprocessed sample libraries into the training sample library X; Step 5: Based on the training sample library X in step 4, the BP neural network method is used to train the power outage probability and power outage duration of the distribution network under extreme disasters, and a power outage prediction model for the distribution network under extreme disasters based on the BP neural network is established; Step 6: In the prediction stage, based on the power outage prediction model trained in step 5, the extreme disaster type, meteorological parameters, distribution network parameters and operating status parameters at a certain time to be predicted are input to obtain the possible power outage probability and power outage duration, and finally realize the prediction of power outage probability and power outage duration under extreme disasters in the distribution network; The steps for establishing the power outage prediction model under extreme disasters of the distribution network based on the BP neural network described in step 5 are as follows: 5.1 Design the BP network into three layers, where the input layer is S, the output layer is O, and the hidden layer is H. The calculation formula for the number of neurons in the hidden layer is: Among them, a is the number of neurons in the input layer, that is, the number of input parameters; c is the number of neurons in the output layer, that is, the number of output parameters; b is the number of neurons in the hidden layer; Υ is the adjustment constant, which ranges from 1 to 10; 5.2 Training sample set X of distribution network under extreme disasters: X={(X1,y 11 ,and 12 ),(X2,y 21 ,and 22 ),...(X i ,and i1 ,and i2 ),(X N ,and N1 ,and N2 )} Among them, X i =(x i1 ,x i2 ,...x iM ), X i is the input parameter of the ith sample, i.e., the distribution network parameters, meteorological parameters, extreme disaster type and operation status parameters when the ith extreme disaster sample occurs, x iM is the Mth parameter in the i-th sample. Each sample has M parameters, y i1 is the power outage state of the distribution network when the i-th extreme disaster sample occurs, y i2 is the power outage duration of the distribution network when the i-th extreme disaster sample occurs; 5.3 After BP neural network training, an a×b prediction model weight matrix ω1 and a b×c prediction model weight matrix ω2 are obtained; The steps for predicting the power outage probability and power outage duration under extreme disasters in the distribution network described in step 6 are as follows: Based on the prediction model matrix ω1 obtained through training in step 5 and the input result of the input layer, we get: H(i)=ω1[1,i].S(1)+ω1[2,i].S(2)+...+ω1[k,i].S(k)+...+ω1[a,i].S(a) Among them, H(i) is the i-th output result of the hidden layer, ω1[k,i] represents the value of the k-th row and i-th column of the weight matrix ω1, that is, the connection weight from the k-th input neuron to the i-th hidden neuron, and S(k) is the k-th parameter value of the input layer; Based on the prediction model matrix ω2 obtained through training in step 5 and the output results of the hidden layer, we obtain: O(i)=ω2[1,i].H(1)+ω2[2,i].H(2)+...+ω2[k,i].H(k)+...+ω2[b,i].H(b) O(i) is the i-th output result of the output layer, ω2[k,i] represents the value of the k-th row and i-th column of the weight matrix ω2, that is, the connection weight from the k-th hidden neuron to the i-th output neuron; The output layer has two output results, namely, the power outage probability and power outage duration of the distribution network under extreme disasters.
2. The method for predicting power outages in distribution network under extreme disasters based on BP neural network according to claim 1 is characterized in that: The main body parameters include the tower information and line information of the distribution network.
3. The method for predicting power outages in distribution network under extreme disasters based on BP neural network according to claim 2 is characterized in that: The tower information includes: tower type, longitude, latitude, altitude, geological environment information, horizontal span, vertical span, tower insulator string model, number of strings, number of insulator pieces, and tension tower rotation angle.
4. The method for predicting power outages in distribution network under extreme disasters based on BP neural network according to claim 2 is characterized in that: The line information includes: voltage level, line specification, line type, number of loops, transmission length, design wind speed, design ice thickness, split number and split gap.
5. The method for predicting power outages in distribution network under extreme disasters based on BP neural network according to claim 1 is characterized in that: The extreme disaster types include: strong winds, heavy rains, snowfall, and sleet.
6. The method for predicting power outages under extreme disasters in distribution networks based on BP neural network according to claim 1 is characterized in that: The meteorological parameters include: temperature, humidity, instantaneous wind speed, average wind speed within 1 hour, instantaneous wind direction, angle between wind direction and line, whether there is freezing rain, whether there is precipitation, precipitation amount, and Doppler radar original parameters.
7. The method for predicting power outages in distribution network under extreme disasters based on BP neural network according to claim 1 is characterized in that: The operating status parameters include current information and switch status information.
8. A power outage prediction system for distribution network under extreme disasters based on BP neural network, characterized in that: include: The information module collects data on meteorological parameters, extreme disaster types, operating status parameters, and power outage duration after the occurrence of power outages in distribution networks with the same physical parameters under extreme natural disasters, and forms a fault sample library under historical extreme disasters in distribution networks and a non-fault sample library under historical extreme disasters in distribution networks; The data preprocessing module performs numerical preprocessing on the fault sample library and the non-fault sample library's body parameters, meteorological parameters, extreme disaster types, operating status parameters, and power outage duration after the fault occurs, and integrates them into a training sample library; The training module uses the BP neural network method to train the power outage probability and duration of the distribution network under extreme disasters, and establishes a power outage prediction model for the distribution network under extreme disasters based on the BP neural network; The prediction module, based on the trained power outage prediction model, inputs the extreme disaster type, meteorological parameters, distribution network parameters and operating status parameters at a certain time to be predicted, and predicts the power outage probability and duration under extreme disasters in the distribution network; The steps of establishing a power outage prediction model under extreme disasters of the distribution network based on BP neural network are as follows: 5.1 Design the BP network into three layers, where the input layer is S, the output layer is O, and the hidden layer is H. The calculation formula for the number of neurons in the hidden layer is: Among them, a is the number of neurons in the input layer, that is, the number of input parameters; c is the number of neurons in the output layer, that is, the number of output parameters; b is the number of neurons in the hidden layer; Υ is the adjustment constant, which ranges from 1 to 10; 5.2 Training sample set X of distribution network under extreme disasters: X={(X1,y 11 ,and 12 ),(X2,y 21 ,and 22 ),...(X i ,and i1 ,and i2 ),(X N ,and N1 ,and N2 )} Among them, X i =(x i1 ,x i2 ,...x iM ), X i is the input parameter of the ith sample, i.e., the distribution network parameters, meteorological parameters, extreme disaster type and operation status parameters when the ith extreme disaster sample occurs, x iM is the Mth parameter in the i-th sample. Each sample has M parameters, y i1 is the power outage state of the distribution network when the i-th extreme disaster sample occurs, y i2 is the power outage duration of the distribution network when the i-th extreme disaster sample occurs; 5.3 After BP neural network training, an a×b prediction model weight matrix ω1 and a b×c prediction model weight matrix ω2 are obtained; The steps of predicting the power outage probability and power outage duration under extreme disasters in the distribution network are as follows: Based on the trained prediction model matrix ω1 and the input result of the input layer, we get: H(i)=ω1[1,i].S(1)+ω1[2,i].S(2)+...+ω1[k,i].S(k)+...+ω1[a,i].S(a) Among them, H(i) is the i-th output result of the hidden layer, ω1[k,i] represents the value of the k-th row and i-th column of the weight matrix ω1, that is, the connection weight from the k-th input neuron to the i-th hidden neuron, and S(k) is the k-th parameter value of the input layer; Based on the trained prediction model matrix ω2 and the output results of the hidden layer, we get: O(i)=ω2[1,i].H(1)+ω2[2,i].H(2)+...+ω2[k,i].H(k)+...+ω2[b,i].H(b) O(i) is the i-th output result of the output layer, ω2[k,i] represents the value of the k-th row and i-th column of the weight matrix ω2, that is, the connection weight from the k-th hidden neuron to the i-th output neuron; The output layer has two output results, namely, the power outage probability and power outage duration of the distribution network under extreme disasters.