CN111723839B - Method for predicting line loss rate of transformer area based on edge calculation - Google Patents
Method for predicting line loss rate of transformer area based on edge calculation Download PDFInfo
- Publication number
- CN111723839B CN111723839B CN202010377652.2A CN202010377652A CN111723839B CN 111723839 B CN111723839 B CN 111723839B CN 202010377652 A CN202010377652 A CN 202010377652A CN 111723839 B CN111723839 B CN 111723839B
- Authority
- CN
- China
- Prior art keywords
- line loss
- loss rate
- electrical characteristic
- characteristic parameters
- area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
- G06F9/5061—Partitioning or combining of resources
- G06F9/5072—Grid computing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Supply And Distribution Of Alternating Current (AREA)
Abstract
The invention provides a transformer area line loss rate prediction method based on edge calculation. It comprises the following steps: 1, selecting historical electrical characteristic parameters and line loss rates of other transformer areas and historical electrical characteristic parameters and line loss rates of a transformer area to be tested to form a data set, and carrying out standardization processing on the data set; 2, constructing an AP clustering model on the cloud platform, clustering the data set, and screening out a historical data set of the same cluster as the data of the to-be-tested distribution area; 3, constructing a BP neural network model on the cloud platform, taking a historical data set as a training sample, and training the BP neural network model; and 4, transplanting the trained prediction model into an edge computing device, outputting the predicted distribution area line loss rate by the edge computing device, and judging whether to send alarm information to operation and maintenance personnel. The data processing pressure of the cloud center and the master station is reduced, the training samples are extracted by using the clustering algorithm with high clustering precision, and the prediction precision is improved.
Description
Technical Field
The invention belongs to the technical field of prediction of line loss rate of a transformer area, and particularly relates to a prediction method of the line loss rate of the transformer area based on edge calculation.
Background
Under the large background of the interconnection of everything, the number of power grid terminal devices and the data volume are greatly increased, the number of the terminal devices accessed to the current power grid system exceeds 5 hundred million (4.7 hundred million of electric meters and tens of millions of various protection, acquisition and control devices), the data structure is complex, the variety is various, and besides traditional structured data, the data structure also comprises a large amount of semi-structured and unstructured data.
As the number of low-voltage users in the transformer area is increased, the problem of line loss of the low-voltage transformer area is more and more prominent. The number of low-voltage transformer areas is large, management conditions are uneven, it is difficult to directly distinguish the transformer areas with abnormal line loss from the large number of transformer areas, the line loss rate of the low-voltage transformer areas is obtained through a manual method, the labor and the effort are wasted, the real-time performance is difficult to guarantee, the data processing pressure of a cloud platform and a main station is increased by using a traditional cloud computing method, and the defects of overlarge communication delay, complex safety authentication mechanism and poor expandability exist.
Disclosure of Invention
The invention aims to provide a method for predicting the line loss rate of a distribution room based on edge computing, which has the advantages of reducing the burden of a cloud center, having low time delay, supporting physical distribution computing and being suitable for real-time analysis and optimization decision.
The technical scheme for solving the technical problems of the invention is as follows: a method for predicting the line loss rate of a distribution area based on edge calculation is characterized by comprising the following steps:
s1: installing edge computing terminal equipment in the transformer area, and acquiring electrical characteristic parameters and line loss rates of other transformer areas and the transformer area to be tested at different moments in a certain historical time period by using an acquisition module in the edge computing terminal equipment as historical data;
s2: carrying out standardization processing on the electrical characteristic parameters of the transformer area in the historical data;
s3: clustering the electrical characteristic parameter data in the historical data of the distribution room by using an AP clustering algorithm, calculating the similarity between the electrical characteristic parameter data and setting a damping coefficient to obtain a plurality of clusters;
s4: selecting historical data of the same cluster as the electrical characteristic parameters at the moment to be tested as a training sample data set of the cloud platform prediction model, and training the data set in the cloud platform through a BP (back propagation) neural network to obtain a platform area line loss rate prediction model;
s5: transplanting the trained prediction model to edge computing terminal equipment, collecting electrical characteristic parameters of a to-be-tested distribution room at a to-be-tested moment through an acquisition module in the edge computing terminal equipment, and predicting through a distribution room line loss rate prediction model to obtain a real-time predicted line loss rate;
s6: comparing the predicted real-time line loss rate with the theoretical line loss rate of the transformer area, calculating an error rate, judging that the transformer area line loss rate is abnormal when the error rate is more than 5%, and then sending alarm information to operation and maintenance personnel through edge computing terminal equipment;
s7: deep training is carried out on the prediction model on the cloud platform through the collected station area historical data and the line loss rate predicted in real time, the prediction model of the edge equipment is updated regularly, and the accuracy rate of the prediction model of the edge computing terminal equipment is guaranteed.
The standardized processing method in the step S2 is as follows: set up the platform district number and be N, the electric characteristic parameter of every platform district is M, and platform district electric characteristic vector X is constituteed to the platform district electric characteristic parameter of the platform district's sample of N, has:
wherein x is ij The element of the ith row and the jth column of the electric characteristic vector X of the platform area is i =1,2, \8230, N, j =1,2, \8230, M;
the method for standardizing the electric characteristic parameters of the transformer area comprises the following steps:
wherein Z is ij Is x ij The amount after the treatment is standardized,is x ij Average value of (1), S ij Is x ij The variance of (c).
The clustering method in the step S3 is to use the attraction degree matrix R and the attribution degree matrix A to exchange information between data points of historical data, continuously iterate and update the two information matrixes until the iteration is finished, and the formula is as follows:
in the formula, r (i, j) and a (i, j) are respectively an attraction degree matrix and an attribution degree matrix between the point i and the point j; s (i, j) is the similarity between the i point and the j point,
each iteration adds a damping coefficient λ, λ ∈ (0, 1), with:
wherein the content of the first and second substances,andr calculated without considering damping coefficient when representing iteration i+1 (i, j) and a i+1 (i,j)。
Said step S4 comprises the following steps,
s4.1: building BP neural network model
The BP neural network model comprises an input layer, a hidden layer and an output layer, wherein a transfer function f (a) between the layers adopts a logsig function, and the model comprises the following steps:
wherein a is an argument of a transfer function f (a) between layers, 0<f (a) <1;
s4.2, utilizing BP neural network model to Z ij And d i Performing learning training, d i And inputting the electrical characteristic parameters of the transformer area into the BP neural network model for the transformer area line loss rate of the ith transformer area, and calculating the transformer area line loss rate d.
In the step S4.2, for any one of the distribution areas, the electrical characteristic parameters of the distribution area are set to be N, so that the input layer of the neural network has N BP neurons, and the input vector of the input layer is set to be Z r =(Z 1 ,Z 2 ,…,Z n ,…,Z N ) T The output vector of the hidden layer is Y r =(Y 1 ,Y 2 ,…,Y P ,…,Y P ) T Of the output layerThe output vector is O r =(O 1 ,O 2 ,…,O l ,…,O L ) T The desired output vector is d r =(d 1 ,d 2 ,…,d l ,…,d L ) T . Wherein T represents transpose, Z n Is the nth BP neuron of the input layer, Y P P-th BP neuron being a hidden layer, O l The first BP neuron of the output layer, d l For the ith expected output value, P is the number of BP neurons in the hidden layer, and L is the number of BP neurons in the output layer.
Z pair by BP neural network model ij And d i The forward propagation process for learning training is as follows:
the output error e is:
in the formula, w mp As weights of the input layer to the hidden layer, b mp For the threshold of input layer to hidden layer, w pl And b pl Respectively, the weight and the threshold from the hidden layer to the output layer.
In the (n + 1) th iteration process, e is expanded according to the Taylor formula, and the formula is obtained as follows:
e(w(n+1))=e(w(n))+g T (n)Δw(n)+0.5Δw T (n)A(n)Δw(n)
wherein, w (n) is the weight in the nth iteration process, and e (w (n)) is the output error in the nth iteration process; w (n + 1) is a weight value in the (n + 1) th iteration process, and e (w (n + 1)) is an output error in the (n + 1) th iteration process; g T (n) is a gradient vector, T represents transpose; Δ w (n) is the n +1 th iterationThe amount of change in the weights in the equation, i.e., Δ w (n) = w (n + 1) -w (n), when Δ w (n) = -A -1 E (w (n + 1)) is the minimum value when (n) g (n); a (n) is a Hessian matrix;
and (3) optimizing the BP neural network by adopting an LM algorithm, and expressing a Hessian matrix A (n) as follows:
A(n)=J T J
wherein J is Jacobian matrix;
the gradient vector g (n) is expressed as:
g(n)=J T e
w (n + 1) is corrected using the following equation:
w(n+1)=w(n)-[J T J+μI] -1 J T e
wherein, I is a unit vector, mu is a constant;
similarly, the threshold b (n + 1) in the (n + 1) th iteration is modified by the following equation:
b(n+1)=b(n)-[J T J+μI] -1 J T e
wherein b (n) is a threshold value in the nth iteration process.
The calculation formula of the theoretical line loss rate and the error rate in the step 6 is
The beneficial effects of the invention are as follows: the method has the advantages that the calculation and analysis function is expanded to equipment with certain calculation capacity at the network edge through an edge calculation technology, the trained model is loaded into edge calculation terminal equipment, and the trained model is updated regularly, so that the burden of a data center is shared.
Drawings
FIG. 1 is a flow chart of the present invention.
Fig. 2 is a schematic flow diagram of a cloud center training model in the embodiment of the present invention.
Fig. 3 is a flowchart of an AP clustering algorithm in the embodiment of the present invention.
Fig. 4 is a schematic structural diagram of a BP neural network model in the embodiment of the present invention.
FIG. 5 is a diagram illustrating predicted line loss rate results according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the present invention comprises the steps of:
s1: installing edge computing terminal equipment in the transformer area, and acquiring electrical characteristic parameters and line loss rates of other transformer areas and the transformer area to be tested at different moments in a certain historical time period by using an acquisition module in the edge computing terminal equipment as historical data;
before the station area electrical characteristic parameters in the historical data are subjected to standardization processing, determining the station area electrical characteristic parameters; the distribution room electrical characteristic parameters comprise parameters reflecting the grid structure and parameters related to load; parameters reflecting the grid structure comprise power supply radius and total length of a low-voltage line; the load-related parameter includes a load rate.
Selecting a group of district data in a certain area, wherein the electric characteristic parameters of 10 districts are shown in a table 1:
table 1:
the electrical characteristic index is used as the input of a neural network algorithm, each parameter has different units and magnitude, and the neural network algorithm can only distinguish the magnitude of a data value and cannot reflect the unit of the data. In order to better apply the above algorithm, it is necessary to eliminate the influence of different units and magnitudes between the parameters on the numerical value. The normalization of data is to scale the data to fall into a small specific interval, remove the unit limitation of the data, and convert it into a dimensionless pure numerical value.
S2: carrying out standardization processing on the electrical characteristic parameters of the transformer area in the historical data;
the standardized processing method in the step S2 is as follows: the number of the transformer areas is N, the electrical characteristic parameters of each transformer area are M, and the transformer area electrical characteristic parameters of N transformer area samples form a transformer area electrical characteristic vector X, which comprises the following steps:
wherein x is ij Is the ith row and jth column element of the platform region electrical characteristic vector X, i =1,2, \ 8230, N, j =1,2, \ 8230, M;
the method for standardizing the electric characteristic parameters of the transformer area comprises the following steps:
wherein Z is ij Is x ij The amount after the treatment is standardized,is x ij Average value of (1), S ij Is x ij The variance of (c).
The electrical characteristic parameters of the distribution area are normalized, and the results are shown in table 2:
TABLE 2
Platform area | Radius of power supply R/m | Total length of low voltage line D/m | Load rate L/%) |
1 | -0.78757 | -0.703 | 0.989197 |
2 | -0.75876 | -0.68834 | 0.662709 |
3 | -0.83943 | -0.68101 | 0.288615 |
4 | -0.80486 | -0.22361 | 0.193892 |
5 | -0.72996 | -0.1569 | -0.71134 |
6 | -0.88552 | -0.29544 | -0.66568 |
7 | -0.93738 | -0.34162 | -0.6729 |
8 | -0.86248 | -0.27492 | -0.59901 |
9 | -0.78181 | -0.20308 | -0.42171 |
10 | -0.73572 | -0.16203 | -0.56324 |
S3: clustering the electrical characteristic parameter data in the historical data of the distribution room through an AP clustering algorithm, calculating the similarity between the electrical characteristic parameter data and setting a damping coefficient to obtain a plurality of clusters;
the clustering method in the step S3 is to use the attraction degree matrix R and the attribution degree matrix A to exchange information between data points of historical data, continuously iterate and update the two information matrixes until the iteration is finished, and the formula is as follows:
in the formula, r (i, j) and a (i, j) are respectively an attraction degree matrix and an attribution degree matrix between the point i and the point j; s (i, j) is the similarity between the i point and the j point,
each iteration adds a damping coefficient λ, λ ∈ (0, 1), with:
wherein the content of the first and second substances,andr calculated without considering damping coefficient when representing iteration i+1 (i, j) and a i+1 (i,j)。
Clustering the obtained historical electrical characteristic parameters of the distribution room through an AP clustering algorithm in the cloud center, wherein b is set in the embodiment i And b j For data clustered in different station areas in the historical data, data point b i And b j The similarity between them is S (i, j) = - (b) i -b j ) 2 Setting a damping coefficient λ =0.55 and a maximum number of iterations 585, as shown in fig. 3And (3) obtaining a clustering number of 2 by using an AP clustering algorithm flow schematic diagram, namely obtaining 2 high-quality clustering centers, selecting a data set of the same type of clusters with the region to be tested as a neural network training set, wherein the original data set of the same type of clusters is shown in a table 3.
TABLE 3
Platform area | Radius of power supply R/m | Total length of low voltage line D/m | Load rate L/%) |
1 | 216 | 467 | 42.69 |
2 | 256 | 496 | 59.37 |
3 | 189 | 453 | 65.31 |
4 | 201 | 498 | 43.56 |
5 | 236 | 560 | 40.28 |
6 | 209 | 598 | 39.65 |
7 | 266 | 689 | 55.23 |
S4: selecting historical data of the same cluster as the electrical characteristic parameters at the moment to be tested as a training sample data set of the cloud platform prediction model, and training the data set in the cloud platform through a BP (back propagation) neural network to obtain a platform area line loss rate prediction model;
said step S4 comprises the following steps,
s4.1: construction of BP neural network model
The BP neural network model comprises an input layer, a hidden layer and an output layer, wherein a transfer function f (a) between the layers adopts a logsig function, and the model comprises the following steps:
wherein a is an argument of a transfer function f (a) between layers, 0<f (a) <1;
s4.2 utilizing BP neural network model to Z ij And d i Performing learning training, d i And inputting the electrical characteristic parameters of the transformer area into the BP neural network model for the transformer area line loss rate of the ith transformer area, and calculating the transformer area line loss rate d.
In step S4.2, for any distribution area, the number of electrical characteristic parameters of the distribution area is set to N, so that the nerveThe input layer of the network has N BP neurons, and the input vector of the input layer is set as Z r =(Z 1 ,Z 2 ,…,Z n ,…,Z N ) T The output vector of the hidden layer is Y r =(Y 1 ,Y 2 ,…,Y P ,…,Y P ) T The output vector of the output layer is O r =(O 1 ,O 2 ,…,O l ,…,O L ) T The desired output vector is d r =(d 1 ,d 2 ,…,d l ,…,d L ) T . Wherein T represents transpose, Z n Is the nth BP neuron of the input layer, Y P P-th BP neuron being a hidden layer, O l The first BP neuron of the output layer, d l For the ith expected output value, P is the number of BP neurons in the hidden layer, and L is the number of BP neurons in the output layer.
Using BP neural network model to Z ij And d i The forward propagation process for learning training is as follows:
the output error e is:
in the formula, w mp As weights of the input layer to the hidden layer, b mp For the threshold of input layer to hidden layer, w pl And b pl Respectively, the weight and the threshold from the hidden layer to the output layer.
In the (n + 1) th iteration process, e is expanded according to the Taylor formula, and the formula is obtained as follows:
e(w(n+1))=e(w(n))+g T (n)Δw(n)+0.5Δw T (n)A(n)Δw(n)
wherein, w (n) is the weight in the nth iteration process, and e (w (n)) is the output error in the nth iteration process; w (n + 1) is a weight value in the (n + 1) th iteration process, and e (w (n + 1)) is an output error in the (n + 1) th iteration process; g is a radical of formula T (n) is a gradient vector, T represents transpose; Δ w (n) is the amount of change in the weights in the n +1 th and n-th iterations, i.e., Δ w (n) = w (n + 1) -w (n), when Δ w (n) = -A -1 E (w (n + 1)) is the minimum value when (n) g (n); a (n) is a Hessian matrix;
and (3) optimizing the BP neural network by adopting an LM algorithm, and expressing a Hessian matrix A (n) as follows:
A(n)=J T J
wherein J is Jacobian matrix;
the gradient vector g (n) is expressed as:
g(n)=J T e
w (n + 1) is corrected using the following equation:
w(n+1)=w(n)-[J T J+μI] -1 J T e
wherein, I is a unit vector, mu is a constant;
similarly, the threshold b (n + 1) in the (n + 1) th iteration is modified by the following equation:
b(n+1)=b(n)-[J T J+μI] -1 J T e
wherein b (n) is a threshold value in the nth iteration process.
S5: transplanting the trained prediction model to edge computing terminal equipment, collecting electrical characteristic parameters of a to-be-tested distribution room at a to-be-tested moment through an acquisition module in the edge computing terminal equipment, and predicting through a distribution room line loss rate prediction model to obtain a real-time predicted line loss rate;
the 10 transformer areas are calculated, the actual values, the estimated values and the errors of the line loss rate of the transformer areas are shown in a table 4,
TABLE 4
Platform area | Amount of power supplied/kWh | Sales electricity/kWh | Actual line loss rate/%) | Line loss rate estimate/%) | Error rate/%) |
1 | 136795 | 128954 | 5.7319346 | 5.669752 | 1.0848 |
2 | 145987 | 136541 | 6.4704391 | 6.284695 | 2.8707 |
3 | 168975 | 158562 | 6.1624501 | 5.997823 | 2.6715 |
4 | 123689 | 115659 | 6.492089 | 6.371189 | 1.8623 |
5 | 289623 | 278964 | 3.6803016 | 3.589452 | 2.4685 |
6 | 375698 | 365987 | 2.5847888 | 2.468523 | 4.4981 |
7 | 445698 | 432139 | 3.0421945 | 2.991548 | 1.6648 |
8 | 597856 | 572364 | 4.263903 | 4.095689 | 3.9451 |
9 | 372698 | 362326 | 2.7829503 | 2.531536 | 9.0341 |
10 | 678954 | 639627 | 5.7922923 | 5.632512 | 2.7585 |
S6: comparing the predicted real-time line loss rate with the theoretical line loss rate of the transformer area, calculating an error rate, judging that the transformer area line loss rate is abnormal when the error rate is more than 5%, and then sending alarm information to operation and maintenance personnel through edge computing terminal equipment;
the calculation formula of the theoretical line loss rate and the error rate in the step 6 is
Compared with the traditional neural network prediction model, the prediction method shown in table 5 and fig. 5 has higher accuracy of the prediction result of the line loss rate of the transformer area, and the prediction method regularly optimizes and updates the prediction model in the edge computing terminal equipment, so that the prediction accuracy is higher, and the method has practical application value.
TABLE 5
Platform area | Actual line loss rate/%) | Conventional line loss rate estimate/%) | Error rate/%) |
1 | 5.7319346 | 5.577921 | 2.686939 |
2 | 6.4704391 | 6.215689 | 3.937138 |
3 | 6.1624501 | 5.936824 | 3.661305 |
4 | 6.492089 | 6.325463 | 2.566601 |
5 | 3.6803016 | 3.436541 | 6.623387 |
6 | 2.5847888 | 2.410035 | 6.760854 |
7 | 3.0421945 | 2.796523 | 8.075470 |
8 | 4.263903 | 4.036875 | 5.324418 |
9 | 2.7829503 | 2.436985 | 12.431602 |
10 | 5.7922923 | 5.578965 | 3.682951 |
S7: deep training is carried out on the prediction model on the cloud platform through the collected station area historical data and the line loss rate predicted in real time, the prediction model of the edge equipment is updated regularly, and the accuracy of the prediction model of the edge computing terminal equipment is guaranteed.
The method has the advantages that the calculation and analysis function is expanded to equipment with certain calculation capacity at the network edge through an edge calculation technology, the trained model is loaded into edge calculation terminal equipment, and the trained model is updated regularly, so that the burden of a data center is shared.
Claims (5)
1. A method for predicting the line loss rate of a distribution area based on edge calculation is characterized by comprising the following steps:
s1: installing edge computing terminal equipment in the distribution room, and acquiring electrical characteristic parameters and line loss rates of other distribution rooms and a distribution room to be tested at different moments in a certain historical time period by using an acquisition module in the edge computing terminal equipment as historical data;
s2: carrying out standardization processing on the electrical characteristic parameters of the transformer area in the historical data;
s3: clustering the electrical characteristic parameter data in the historical data of the distribution room through an AP clustering algorithm, calculating the similarity between the electrical characteristic parameter data and setting a damping coefficient to obtain a plurality of clusters;
s4: selecting historical data of the same cluster as the electrical characteristic parameters at the moment to be tested as a training sample data set of the cloud platform prediction model, and training the data set in the cloud platform through a BP (back propagation) neural network to obtain a platform area line loss rate prediction model;
s5: transplanting the trained prediction model to edge computing terminal equipment, collecting electrical characteristic parameters of a to-be-tested distribution room at a to-be-tested moment through an acquisition module in the edge computing terminal equipment, and predicting through a distribution room line loss rate prediction model to obtain a real-time predicted line loss rate;
s6: comparing the predicted real-time line loss rate with the theoretical line loss rate of the distribution room, calculating an error rate, judging that the distribution room line loss rate is abnormal when the error rate is more than 5%, and then sending alarm information to operation and maintenance personnel through edge computing terminal equipment;
s7: deep training is carried out on the prediction model on the cloud platform through the collected station area historical data and the line loss rate predicted in real time, the prediction model of the edge equipment is updated regularly, and the accuracy of the prediction model of the edge computing terminal equipment is guaranteed.
2. The method of claim 1, wherein the method for predicting the line loss rate of the cell based on the edge calculation comprises: the standardized processing method in the step S2 is as follows: the number of the transformer areas is N, the electrical characteristic parameters of each transformer area are M, and the transformer area electrical characteristic parameters of N transformer area samples form a transformer area electrical characteristic vector X, which comprises the following steps:
wherein x is ij Is the ith row and jth column element of the platform region electrical characteristic vector X, i =1,2, \ 8230, N, j =1,2, \ 8230, M;
the method for standardizing the electric characteristic parameters of the transformer area comprises the following steps:
3. The method of claim 1, wherein the method for predicting the line loss rate of the cell based on the edge calculation comprises: the clustering method in the step S3 is to use the attraction degree matrix R and the attribution degree matrix A to exchange information between data points of historical data, continuously iterate and update the two information matrixes until the iteration is finished, and the formula is as follows:
in the formula, r (i, j) and a (i, j) are respectively an attraction degree matrix and an attribution degree matrix between the point i and the point j; s (i, j) is the similarity between the i point and the j point,
each iteration adds a damping coefficient λ, λ ∈ (0, 1), with:
4. The method of claim 1, wherein the method for predicting the line loss rate of the cell based on the edge calculation comprises: said step S4 comprises the following steps,
s4.1: construction of BP neural network model
The BP neural network model comprises an input layer, a hidden layer and an output layer, wherein a transfer function f (a) between the layers adopts a logsig function, and comprises the following steps:
wherein a is an argument of a transfer function f (a) between layers, 0<f (a) <1;
s4.2, utilizing BP neural network model to Z ij And d i Performing learning training, d i And inputting the electrical characteristic parameters of the station area into a BP neural network model for the line loss rate of the station area of the ith station area in the historical data, and calculating the line loss rate d of the station area.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010377652.2A CN111723839B (en) | 2020-05-07 | 2020-05-07 | Method for predicting line loss rate of transformer area based on edge calculation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010377652.2A CN111723839B (en) | 2020-05-07 | 2020-05-07 | Method for predicting line loss rate of transformer area based on edge calculation |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111723839A CN111723839A (en) | 2020-09-29 |
CN111723839B true CN111723839B (en) | 2022-11-25 |
Family
ID=72564252
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010377652.2A Active CN111723839B (en) | 2020-05-07 | 2020-05-07 | Method for predicting line loss rate of transformer area based on edge calculation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111723839B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113177652A (en) * | 2020-10-16 | 2021-07-27 | 上海交通大学 | Photovoltaic output prediction method |
CN112649642A (en) * | 2020-12-14 | 2021-04-13 | 广东电网有限责任公司广州供电局 | Electricity stealing position judging method, device, equipment and storage medium |
CN113269462A (en) * | 2021-06-08 | 2021-08-17 | 内蒙古电力(集团)有限责任公司内蒙古电力科学研究院分公司 | Transformer area line loss calculation method based on edge calculation |
CN114638290B (en) * | 2022-03-07 | 2024-04-30 | 廖彤 | Environment monitoring instrument fault prediction method based on edge calculation and BP neural network |
CN116930667A (en) * | 2023-09-15 | 2023-10-24 | 广东电网有限责任公司 | Method, device, equipment and storage medium for testing edge of power grid of transformer area |
CN117706282A (en) * | 2024-02-06 | 2024-03-15 | 国网安徽省电力有限公司营销服务中心 | Method and system for monitoring phase line loss in station division |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106127387A (en) * | 2016-06-24 | 2016-11-16 | 中国电力科学研究院 | A kind of platform district based on BP neutral net line loss per unit appraisal procedure |
CN110348713A (en) * | 2019-06-28 | 2019-10-18 | 广东电网有限责任公司 | A kind of platform area line loss calculation method based on association analysis and data mining |
CN110782181A (en) * | 2019-11-05 | 2020-02-11 | 国网重庆市电力公司电力科学研究院 | Low-voltage transformer area line loss rate calculation method and readable storage medium |
-
2020
- 2020-05-07 CN CN202010377652.2A patent/CN111723839B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN111723839A (en) | 2020-09-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111723839B (en) | Method for predicting line loss rate of transformer area based on edge calculation | |
WO2021238505A1 (en) | Federated learning-based regional photovoltaic power probability prediction method, and cooperative regulation and control system | |
CN108694467B (en) | Method and system for predicting line loss rate of power distribution network | |
US20190265768A1 (en) | Method, system and storage medium for predicting power load probability density based on deep learning | |
CN108197744B (en) | Method and system for determining photovoltaic power generation power | |
CN107169628B (en) | Power distribution network reliability assessment method based on big data mutual information attribute reduction | |
CN108280552B (en) | Power load prediction method and system based on deep learning and storage medium | |
CN103105246A (en) | Greenhouse environment forecasting feedback method of back propagation (BP) neural network based on improvement of genetic algorithm | |
CN109002781B (en) | Fault prediction method for energy storage converter | |
CN111736084A (en) | Valve-regulated lead-acid storage battery health state prediction method based on improved LSTM neural network | |
CN107133684A (en) | A kind of random matrix construction method towards GA for reactive power optimization | |
CN108182490A (en) | A kind of short-term load forecasting method under big data environment | |
CN112990500A (en) | Transformer area line loss analysis method and system based on improved weighted gray correlation analysis | |
CN114777192B (en) | Secondary network heat supply autonomous optimization regulation and control method based on data association and deep learning | |
CN111525587A (en) | Reactive load situation-based power grid reactive voltage control method and system | |
WO2020228568A1 (en) | Method for training power generation amount prediction model of photovoltaic power station, power generation amount prediction method and device of photovoltaic power station, training system, prediction system and storage medium | |
CN112036067A (en) | Method for predicting steady-state temperature rise of groove cable | |
CN115879602A (en) | Ultra-short-term photovoltaic output prediction method based on transient weather | |
CN114757330A (en) | Urban instantaneous water consumption prediction method based on LSTM | |
CN113139698A (en) | Load prediction method, device and equipment | |
CN112363012A (en) | Power grid fault early warning device and method | |
CN110489893B (en) | Variable weight-based bus load prediction method and system | |
CN111612227A (en) | Load prediction method based on K-means clustering and bat optimization neural network | |
CN117076887A (en) | Pump station unit running state prediction and health assessment method and system | |
CN116415732A (en) | User side power load data processing method based on improved ARNN |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |