Dynamic capacity prediction method for overhead transmission line

A technology of overhead transmission line and prediction method, applied in the field of dynamic capacity prediction of transmission lines, can solve problems such as limited input, and achieve the effect of improving dimension and accuracy

Inactive Publication Date: 2016-05-11
ELECTRIC POWER RESEARCH INSTITUTE OF STATE GRID SHANDONG ELECTRIC POWER COMPANY +1
2 Cites 12 Cited by

AI-Extracted Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to solve the above problems, provide a dynamic capacity prediction method for overhead transmission lines, use the restricted Boltzmann machine to realize the prediction of wind speed and s...
View more

Abstract

The invention discloses a dynamic capacity prediction method for an overhead transmission line. The dynamic capacity prediction method comprises the following steps of: measuring a real-time wind speed near a conducting wire through a wind speed sensor mounted on an iron tower, measuring a real-time solar radiation temperature and a real-time ambient temperature through a temperature sensor, predicating the wind speeds and solar radiation temperatures within 1 hour, 2 hours and 4 hours in the future through a RBM depth learning mechanism, obtaining a real-time conducting wire temperature through a tensiometer and a tension-temperature fitting curve, substituting the wind speed, a predicted value of solar radiation temperature and the real-time conducting wire temperature into a conducting line capacity calculation model to obtain predicted capacity values of the conducting wire within 1 hour, 2 hours and 4 hours in the future. Compared with the existing neural network and other prediction methods, the prediction accuracy is effectively improved. The problem that the input of the existing neural network prediction method is extremely limited is solved, and the dimension of the input is greatly improved, and a theoretical basis is provided for the analysis of large grid data in the future.

Application Domain

Climate change adaptationForecasting +2

Technology Topic

Radiation temperaturePower grid +7

Image

  • Dynamic capacity prediction method for overhead transmission line
  • Dynamic capacity prediction method for overhead transmission line
  • Dynamic capacity prediction method for overhead transmission line

Examples

  • Experimental program(1)

Example Embodiment

[0041] The present invention will be further described below in conjunction with the drawings and embodiments.
[0042] A method for predicting the dynamic capacity of overhead transmission lines, including the following steps:
[0043] Step one, such as image 3 As shown, the wind speed data collected by the wind speed sensor 3 on the iron tower and the weather forecast wind speed data provided by the National Meteorological Information Center are used to train the wind speed RMB deep learning machine, and iteratively calculate the total wind speed RMB energy until the wind speed RMB total energy is for each learning sample When all reach the minimum value, the wind speed RMB deep learning machine training is completed;
[0044] Step 2: Input the current wind speed and the weather forecast wind speed collected by the wind speed sensor 3 on the iron tower into the wind speed RMB deep learning machine to predict the wind speed in the next 1 hour, 2 hours and 4 hours;
[0045] Step three, such as image 3 As shown, the insolation radiation temperature and ambient temperature collected by the temperature sensor 2 on the iron tower and the ambient temperature forecast value provided by the National Meteorological Information Center are used to train the insolation radiation temperature RMB deep learning machine, and the total energy of the insolation radiation temperature RMB is calculated by iteration until the sun shines. When the total energy of the radiation temperature RMB reaches the minimum value for each learning sample, the training of the deep learning machine of the solar radiation temperature RMB is completed;
[0046] Step 4: Input the current solar radiation temperature, ambient temperature and weather forecast ambient temperature collected by the temperature sensor 2 on the iron tower into the solar radiation temperature RMB deep learning machine to predict the solar radiation temperature in the next 1 hour, 2 hours and 4 hours ;
[0047] Step 5: Calculate the real-time temperature Tc of the wire by using the tension meter 3 installed at the wire connection fittings to measure the real-time tension of the wire and the experimentally measured tension-temperature fitting curve;
[0048] Step 6. Bring the wind speed value predicted in Step 2, the solar radiation temperature value predicted in Step 4, and the real-time temperature Tc of the wire obtained in Step 5 into the wire capacity formula I = Q f ( T c , T s , V ) + Q r ( T c , T s ) R ( T c ) , Where Q r Is the radiation heat dissipation power, Q f Is the convection heat dissipation power, R(T c ) Is the AC resistance of the wire, T S Is the solar radiation temperature, V is the wind speed in the vertical direction of the wire, and dynamically predicts the capacity of the wire in the next 1 hour, 2 hours and 4 hours.
[0049] The first step includes: the wind speed sensor 3 on the iron tower measures the wind speed v near the wire at a certain time t at the target location t , At the same time, it records the forecast data v’ t+1 , V’ t+2 , V’ t+4 , As the 4 neurons of the input layer of the wind speed RMB deep learning machine {v t , V’ t+1 , V’ t+2 , V’ t+4 }; The output layer is set to 3 neurons {v t+1 , V t+2 , V t+4 }, v t+1 , V t+2 , V t+4 Respectively represent the wind speed prediction values ​​at t+1, t+2 and t+4; use 48 hours of data to train the learning machine;
[0050] Such as Figure 4 As shown, the method for training the wind speed RMB deep learning machine is: set the initial structure of the RBM network, including visible variables and hidden variables; adjust the weights of the visible layer and hidden layer nodes, respectively; derivate the parameters of the RBM learning machine, and find the third Parameters W, c, b; iterate repeatedly until the wind speed RMB total energy reaches the minimum value for each learning sample, the wind speed RMB deep learning machine training is completed.
[0051] (1) Set the initial structure of the RBM network, including visible variables and hidden variables, and input samples; there are a total of m hidden layer nodes, and a total of n visible layer nodes;
[0052] (2) Initialize the partial derivatives of the three parameters W, b, and c (initialized to 0), where b is the bias of the visible layer node in the deep learning machine, and c is the bias of the hidden layer node in the deep learning machine , W is the weight connecting the visible layer node and the hidden layer node, corresponding to formula (10);
[0053] (3) The weights of visible layer and hidden layer nodes are initially set to 1;
[0054] (4) Derivation of the three parameters W, c, b of the wind speed RMB deep learning machine, and calculate the total RBM energy E(v, h);
[0055] (5) Re-adjust the weight (1 or 0) of each node in the visible layer and the hidden layer, and iterate repeatedly. When each learning sample reaches the minimum value, the wind speed RMB deep learning machine training is completed.
[0056] The step three includes: measuring the solar radiation temperature T at time t using the temperature sensor 2 installed on the iron tower st And ambient temperature T at , Read the forecast value T’ of the ambient temperature of the target area in the next 1 hour, 2 hours and 4 hours from the National Meteorological Information Center at+1 , T’ at+2 And T’ at+4 , And set the input layer of the RBM deep learning machine to 5 neurons {T st , T at , T’ at+1 , T’ at+2 , T’ at+4 }; Set the output layer of the RBM deep learning machine for solar radiation temperature to 3 neurons {T st+1 , T st+2 , T st+4 }, T st+1 , T st+2 , T st+4 It respectively represents the predicted value of solar radiation temperature 1 hour, 2 hours and 4 hours after t time; 48 hours of historical data are used to train the solar radiation temperature RBM deep learning machine.
[0057] The training method of the RMB deep learning machine for solar radiation temperature is:
[0058] 1) Set the initial structure of the RBM network, including visible variables and hidden variables, and input samples. There are a total of m hidden layer nodes and a total of n visible layer nodes;
[0059] 2) Initialize the partial derivatives of the three parameters W, b, and c (initialized to 0), where b is the bias of the visible layer node in the deep learning machine, and c is the bias of the hidden layer node in the deep learning machine, W is the weight connecting the visible layer node and the hidden layer node, corresponding to formula (10);
[0060] 3) The weights of nodes in the visible layer and hidden layer are initially set to 1;
[0061] 4) Derivation of the three parameters W, c, b of the deep learning machine of the solar radiation temperature RMB, and calculate the total RBM energy E(v, h);
[0062] 5) Re-adjust the weight (1 or 0) of each node in the visible layer and hidden layer, and iterate repeatedly. When each learning sample reaches the minimum value, the insolation radiation temperature RMB deep learning machine training is completed.
[0063] 1) Calculation of wire capacity
[0064] The current-carrying capacity of overhead transmission lines is mainly related to environmental factors, conductor temperature, conductor material size, and other conductor factors. The calculation is mainly based on the steady-state heat balance equation of the wire, as follows
[0065] Q r +Q f =Q s +I 2 R(T c )(1)
[0066] Where Q r Is the radiation heat dissipation power, Q f Is the convection cooling power, Q s Is the solar heat absorption power, I 2 R(T c ) Is the heating power of the AC resistance of the wire. After determining the wire type, that is, the parameters of the wire, the physical meaning and calculation of the four powers in the formula (1) are as follows.
[0067] Radiant heat dissipation power Q r It shows the thermal radiation power on the surface of the wire, which is related to the wire's own parameters and the wire temperature T c , Ambient temperature T a For related calculations, see the following formula.
[0068] Q r =πDE 1 S 1 [(T c +273) 4 -(T a +273) 4 ](2)
[0069] Among them, D is the outer diameter of the wire; E 1 Is the coefficient of heat radiation on the surface of the wire. In our national standard, it is generally 0.9; S 1 Is the Stefan-Boltzmann constant, which is 5.67×10 -8 , W/m 2; T a , T c Is the wire temperature and the ambient temperature, ℃.
[0070] Convection cooling power Q f It shows that the wire is convectively cooled by the nearby air. When the temperature of the wire is greater than the ambient air temperature, the air close to the surface of the wire is heated and the density becomes smaller, and finally gas heat convection is generated, thereby taking away part of the heat of the wire. The convection heat dissipation power can be calculated by the following formula.
[0071] Q c =0.57πλ f θRe 0.485 (3)
[0072] Where λ f Is the heat transfer coefficient of the air layer near the wire surface, W/m℃, see (4); Re is the Reynolds number, see (5).
[0073] λ f =2.42×10 -2 +7(T a +(T c -T a )/2)×10 -5 (4)
[0074] Re=VD/υ(5)
[0075] In the above two formulas, V is the wind speed in the vertical direction of the wire, m/s; υ is the kinematic viscosity of the air layer on the wire surface, m 2 /s, can be calculated by the following formula.
[0076] υ=1.32×10 -5 +9.6(T a +(T c -T a )/2)×10 -8 (6)
[0077] Sunshine heat absorption power Q s It indicates the power absorbed by the wire to the solar radiation, which is related to the size of the wire and the solar intensity.
[0078] Q s =α S J S D(7)
[0079] Where α S Is the heat absorption coefficient of the wire surface, which is generally equal to the radiation coefficient of the wire in Chinese standards; J s For the sunshine intensity, for the sake of conservativeness, in our country standard, it is generally taken as 1000W/m 2.
[0080] Heating power of wire AC resistance I 2 R(T c ) Is composed of the steady-state current carrying capacity I of the wire and the AC resistance R(T c ) Is calculated. Wire temperature is T c When the AC resistance R(T c ) And wire temperature T c related:
[0081] R(T c )=(1+k)R 20 [1+α 20 (T c -20)](8)
[0082] Among them, k is the skin effect coefficient of the wire, which is related to the cross-sectional area of ​​the wire, when the cross-sectional area is not greater than 400mm 2 When k = 0.0025 is greater than 400mm 2 , Take k = 0.01; R 20 It is the temperature coefficient of the wire at 20°C. For aluminum material, it can be 0.00403/°C.
[0083] Sunshine radiation temperature T S It is defined as the wire temperature when the wire load current is zero and only the sun radiation is input. The monitoring of solar radiation uses "net radiation sensor" to monitor the temperature of solar radiation instead of direct monitoring of solar radiation. The net radiation sensor is composed of a small piece of wire with the same new and old material, size and direction as the wire to be monitored, and a thermocouple temperature sensor 2, which is installed at the end of the wire at the tensile end to be tested. It has the same absorptivity and reflectivity as the monitored wire, which can be used to replace direct monitoring of solar radiation.
[0084] Using the wire climate model when the wire load current I=0 and I≠0, the ambient temperature can be eliminated, and the comprehensive influence of the ambient temperature and the intensity of the sunshine on the wire can be characterized by the solar radiation temperature. Finally, the wire capacity calculation formula is as follows:
[0085] I = Q f ( T c , T s , V ) + Q r ( T c , T s ) R ( T c ) - - - ( 9 )
[0086] 2) Restricted Boltzmann Machine (RBM)
[0087] The invention uses a restricted Boltzmann machine to predict wind speed and solar radiation temperature.
[0088] The restricted Boltzmann machine can be regarded as an undirected graph model. It consists of a visible layer and a hidden layer. The visible layer and the hidden layer are connected to each other, but there is no connection between the units in the two layers, such as figure 1 Shown.
[0089] There are a total of m hidden layer units and a total of n visible layer units. h i Indicates the state of the i-th hidden layer node; v j Represents the state of the j-th visible layer node; W ij Means h i And v j The weight value of the connection between two nodes; c i Indicates the bias of the i-th hidden layer node; b j Represents the bias of the j-th visible layer node. Inspired by statistical mechanics, the concept of energy function was introduced. A measure to describe the state of the system under study using an energy function. The energy of the system is related to the entropy of the system itself. The higher the order of the system or the concentrated probability distribution, the smaller the energy of the system as a whole. Therefore, in order to obtain the weight of the restricted Boltzmann machine and the bias of each unit, the minimum energy function is used to calculate the solution.
[0090] Restricted Boltzmann machine as a system, its energy is expressed as
[0091] E ( v , h ) = - X i = 1 n X j = 1 m v y W i j h j - X j = 1 m b j v j - X i = 1 n c i h i - - - ( 10 )
[0092] In the above formula, W ij , V j , H i Is the parameter of the restricted Boltzmann machine, v j ,h i ∈{0,1}.
[0093] According to the energy formula, the joint probability of the visible layer node and the hidden layer node is
[0094] p ( v , h ) = e - E ( v , h ) X v , h e - E ( v , h ) - - - ( 11 )
[0095] Given the model parameters of the restricted Boltzmann machine, the distribution of the visible layer unit of the observation data is calculated as follows:
[0096] p ( v ) = 1 X v , h e - E ( v , h ) X h p ( v , h ) - - - ( 12 )
[0097] From the fact that the cells in the constrained Boltzmann machine are not connected, and the cells between the layers are connected, it can be seen that for a given state of the visible layer, the state of the hidden layer is opposite to each other, and vice versa. Therefore, the probability that the i-th hidden layer unit takes a value of 1 is
[0098] p ( h i = 1 | v ) = σ ( X j = 1 m W i j v j + c i ) - - - ( 13 )
[0099] The probability that the j-th visible layer unit is 1 is
[0100] p ( v j = 1 | h ) = σ ( X i = 1 n W i j h i + b j ) - - - ( 14 )
[0101] among them, σ ( x ) = 1 1 + e - x .
[0102] Let θ={W ij ,c i ,b j } Are the parameters required by the restricted Boltzmann machine, which are solved by maximum likelihood according to the above formula.
[0103] ∂ ln P ( v ) ∂ W i j = P ( h j = 1 | v ) v i - X v P ( v ) ( P ( h j = 1 | v ) v i )
[0104] ∂ ln P ( v ) ∂ c j = P ( h j = 1 | v ) - X v P ( v ) ( P ( h j = 1 | v ) )
[0105] ∂ ln P ( v ) ∂ b i = v i - X v P ( v ) v i
[0106] Although the specific embodiments of the present invention are described above with reference to the accompanying drawings, they do not limit the protection scope of the present invention. Those skilled in the art should understand that on the basis of the technical solutions of the present invention, those skilled in the art do not need to pay creative work Various modifications or variations that can be made are still within the protection scope of the present invention.

PUM

no PUM

Description & Claims & Application Information

We can also present the details of the Description, Claims and Application information to help users get a comprehensive understanding of the technical details of the patent, such as background art, summary of invention, brief description of drawings, description of embodiments, and other original content. On the other hand, users can also determine the specific scope of protection of the technology through the list of claims; as well as understand the changes in the life cycle of the technology with the presentation of the patent timeline. Login to view more.

Similar technology patents

Measurement method of thickness of subsurface damaged layer of bucky optical material

InactiveCN101672625AHigh precisionStrong engineering applicability
Owner:XI AN JIAOTONG UNIV

Classification and recommendation of technical efficacy words

  • add dimension
  • High precision

System for EMI/RFI filtering and transient voltage suppression

InactiveUS20070159753A1add dimensionimprove device density capability
Owner:RANDALL MICHAEL STEVEN +1

Multi-antenna diversity method and device

InactiveCN101873201Aadd dimensiongood diversity gain
Owner:ZTE CORP

Transformer substation intelligent safety management and control method and system

ActiveCN112152129AImprove operation and maintenance specificationsadd dimension
Owner:HUZHOU ELECTRIC POWER SUPPLY CO OF STATE GRID ZHEJIANG ELECTRIC POWER CO LTD +1

Video pushing method and system and server

InactiveCN105898423Aadd dimension
Owner:LETV INFORMATION TECH BEIJING

Method for forecasting short-term power in wind power station

InactiveCN102102626AHigh precision
Owner:NORTH CHINA ELECTRIC POWER UNIV (BAODING) +1

Numerical control machine tool wear monitoring method

InactiveCN102091972AHigh precisionReal-time monitoring of tool wear
Owner:HUAZHONG UNIV OF SCI & TECH +1

Advertisement recommendation method and apparatus

ActiveCN104965890AHigh precisionPrecise screening
Owner:SHENZHEN TENCENT COMP SYST CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products