CN111325315A - Distribution transformer power failure and power loss prediction method based on deep learning - Google Patents

Distribution transformer power failure and power loss prediction method based on deep learning Download PDF

Info

Publication number
CN111325315A
CN111325315A CN201911169829.3A CN201911169829A CN111325315A CN 111325315 A CN111325315 A CN 111325315A CN 201911169829 A CN201911169829 A CN 201911169829A CN 111325315 A CN111325315 A CN 111325315A
Authority
CN
China
Prior art keywords
load
power failure
formula
distribution transformer
gate
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.)
Pending
Application number
CN201911169829.3A
Other languages
Chinese (zh)
Inventor
罗晨
山宪武
张冬冬
孙羽森
俞海猛
张良
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Kashgar Power Supply Co Of State Grid Xinjiang Electric Power Co ltd
NARI Nanjing Control System Co Ltd
Electric Power Research Institute of State Grid Xinjiang Electric Power Co Ltd
Original Assignee
Kashgar Power Supply Co Of State Grid Xinjiang Electric Power Co ltd
NARI Nanjing Control System Co Ltd
Electric Power Research Institute of State Grid Xinjiang Electric Power Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Kashgar Power Supply Co Of State Grid Xinjiang Electric Power Co ltd, NARI Nanjing Control System Co Ltd, Electric Power Research Institute of State Grid Xinjiang Electric Power Co Ltd filed Critical Kashgar Power Supply Co Of State Grid Xinjiang Electric Power Co ltd
Priority to CN201911169829.3A priority Critical patent/CN111325315A/en
Publication of CN111325315A publication Critical patent/CN111325315A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • 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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • 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
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Economics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Computational Linguistics (AREA)
  • Marketing (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Primary Health Care (AREA)
  • Water Supply & Treatment (AREA)
  • Public Health (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses a distribution transformer power failure electric quantity loss prediction method based on deep learning, which is characterized in that classification of distribution transformer load curves is realized based on a fuzzy C-means clustering algorithm to obtain various clustering center load curves, the amplification of various distribution transformer loads and corresponding various center loads in various time periods is calculated, and the time period with the amplification rate smaller than a threshold value is selected as a time period to be predicted; predicting a load value of a time period to be predicted by using the trained neural network of the gated circulation unit as power failure loss electric quantity; an objective function is solved, and the electrical duration is obtained when the objective function is optimized. The invention realizes the calculation of the power failure loss electric quantity, provides data support for improving the reliability of power supply, and realizes the optimized management of planned power failure based on user load curve clustering.

Description

Distribution transformer power failure and power loss prediction method based on deep learning
Technical Field
The invention relates to a distribution transformer power failure electric quantity loss prediction method based on deep learning, and belongs to the technical field of electric power system load prediction.
Background
The loss caused by the power failure event to the national economy far exceeds the loss of the power system. The distribution network of the first-class city in the world requires that the annual average power failure time of users in the core region of the city is not more than 5 minutes, and the power supply reliability reaches about 99 percent. The average power failure time of a user in 2017 years in China is 16.27 hours per household, the optimization and the promotion of a power distribution network frame still need to be accelerated in many domestic urban power grids, high-end equipment is applied, the lean operation and maintenance and the intelligent interaction level are promoted, the fault processing time is reduced from an hour level to a minute level, the power failure time is reduced, and the average power supply reliability is improved.
The power failure in the actual operation of the power system is generally divided into planned power failure and fault power failure, wherein the fault power failure is determined by the real-time operation condition of the power system, is influenced by various factors and is generally difficult to perform manual intervention and control.
In order to evaluate the loss electric quantity in the power failure time period, the load in the power failure time period needs to be predicted. The prediction of the power failure loss electric quantity can provide scientific guidance basis for reasonably dealing with destructive impact on the power grid caused by large power load change, controlling the power grid to be stable, operating economically and efficiently and the like. The power utilization behavior of users in the distribution transformer area is influenced by factors such as weather and production plans, so that the distribution transformer load curve shows strong randomness and volatility, the cycle and regularity of the load curve are poor, and the load prediction difficulty of the distribution transformer area is increased. In addition, the distribution transformer load curve has the characteristics of various modes, rich types, non-stable sequences and the like, so that personalized treatment is needed when a prediction model is established, the model adaptability is poor, and the load prediction difficulty of a distribution transformer area is increased to a certain extent. The effective prediction of the load of the distribution transformer area can provide data support for planned power failure management, electrical equipment maintenance, optimized scheduling and the like.
Machine learning and deep learning: vapnik et al, 1995, proposed a machine learning technique with better generalization performance, namely, Support Vector Machines (SVM) algorithm. When the number of training samples is small, the SVM can still well solve the problems of classification and regression of nonlinear and high-dimensional data with high precision. However, the SVM has the defects of long data training time, requirement of setting error parameters and the like. In addition, when choosing the kernel function, the parameters are not easy to determine and the kernel function must satisfy the Mercer condition. Under the technical background of large electric power data, the SVM is used for processing massive load data samples, so that the severe challenge is faced, and the model training time and space complexity are greatly improved along with the increase of the number of the samples. Therefore, related documents provide improved methods such as a related vector machine and a core vector machine, and short plates with low SVM operation efficiency are avoided. The SVM parameter selection has obvious influence on the model prediction performance, and the parameters are mostly selected by adopting a grid search method, but the method selects an initial value by depending on manual experience, and the error is large. Therefore, the model hyper-parameters are determined by adopting an intelligent search algorithm, such as a particle swarm algorithm, a genetic algorithm and the like, so that the parameter optimization solution can reduce the model prediction error.
Most of the machine learning is established by a shallow perception machine prediction model, and dimension disasters often occur on high-dimensional feature data. Therefore, effective measures are required to be taken to select input variables before model training, deep learning is used as extension and extension of a machine learning neural network algorithm, a deep network structure with multiple hidden layers is established, and high-dimensional data scenes can be effectively processed. The Deep Belief Network (DBN) can be used for classifying and regressing mass data, the DBN is formed by stacking a plurality of layers of limited Boltzmann machines from bottom to top according to a certain rule, the parameters of each layer of model are adjusted by adopting a contrast divergence training method, and the parameters are finely adjusted based on error back propagation, so that the optimal solution is finally obtained. The restricted Boltzmann machine can efficiently solve massive complex nonlinear classification and regression problems and has autonomous feature processing capability.
In order to effectively reduce the load prediction error and enhance the reliability of the prediction model, a learner proposes an idea of utilizing advantage complementation between different models and integrates two or more different prediction methods, so as to obtain a more reliable prediction result. For example, two or more model prediction results are combined and processed through an optimization weight and a fusion strategy, different combination weights are set, different model prediction advantages can be exerted, and the reliability of the overall prediction result is enhanced; or preprocessing the original time sequence of the load by utilizing signal processing, predicting the decomposed subsequence, and adding the prediction results of all parts to obtain a final load prediction value.
Disclosure of Invention
The purpose is as follows: in order to solve the problems of low processing speed, large memory occupation, difficulty in processing high-dimensional characteristics and the like in the prior art when massive data are processed, the invention provides a distribution transformer power failure electric quantity loss prediction method based on deep learning.
The technical scheme is as follows: in order to solve the technical problems, the technical scheme adopted by the invention is as follows:
a method for predicting power loss of distribution transformer power failure based on deep learning comprises the following steps:
classifying distribution transformer load curves based on a fuzzy C-means clustering algorithm to obtain various clustering center load curves, calculating the amplification of various distribution transformer loads and corresponding various center loads in various time periods, and selecting the time period with the amplification rate smaller than a threshold value as a time period to be predicted;
predicting a load value of a time period to be predicted by using the trained neural network of the gated circulation unit as power failure loss electric quantity;
an objective function is solved, and the electrical duration is obtained when the objective function is optimized.
As a preferred scheme, the fuzzy C-means clustering algorithm comprises the following steps:
step 1: determining a generic number c, a load number n and an initial membership matrix
Figure RE-GDA0002487237830000036
Wherein u isikRepresenting the kth load xkMembership belonging to class i, 0 represents step 0 iteration. Let iteration variable be l, l ═ 1 denote step 1 iteration;
step 2: calculating a membership matrix U using the formula(l)：
Figure RE-GDA0002487237830000031
In the formula:
Figure RE-GDA0002487237830000032
wherein x iskIs the k-th load,
Figure RE-GDA0002487237830000033
For the ith step to iterate the cluster centers of the ith class,
Figure RE-GDA0002487237830000034
iterating the clustering centers of the jth class i, j ∈ 1, 2.. c in the ith step;
and step 3: correcting the cluster center v of step l +1 by(l+1)：
Figure RE-GDA0002487237830000035
In the formula: x is an input value of the model and represents a load sequence of historical 96 points acquired by the adoption system; v is a clustering center of the model and represents a characteristic load sequence obtained by clustering analysis; the other variables are intermediate variables and training parameters of the model;
and 4, step 4: given a membership termination tolerance ε >0 when
Figure RE-GDA0002487237830000041
And stopping, otherwise, turning to the step 2, wherein l is l + 1.
As a preferred scheme, the cluster center load curves are divided into 4 classes, including: day type load, all day type load, night type load 1, and night type load 2.
Preferably, the training step of the gated cyclic unit neural network is as follows:
acquiring historical load values of time periods of each day;
acquiring a historical load value of a time period to be measured on a date to be measured;
calculating the Pearson correlation coefficient of the historical load value of each time period on each day and the historical load value of the time period to be detected on the date to be detected, and sorting the historical load values of the time periods on the first N days from large to small according to the coefficient, wherein the historical load values of each time period on each day are taken as input characteristic quantities;
taking a historical load value of a time period to be measured on a date to be measured as an output characteristic quantity;
and learning the neural network of the gate control cycle unit by using the input characteristic quantity and the output characteristic quantity to obtain intermediate variables and training parameters of the neural network of the gate control cycle unit.
Preferably, the gated cyclic unit neural network:
for an input vector x, a weight vector W, an offset term b, the output value of the gate is expressed as:
g(x)＝σ(Wx+b) (5)
in the formula: sigma is sigmoid activation function, and the calculation formula is
Figure RE-GDA0002487237830000042
1) At time step t, the calculation formula of the update gate is as follows:
zt＝σ(Wzhht-1+Wzxxt+bz) (6)
in the formula: wzhTo update the gate cycle weight, WzxTo update the gate input weights, bzTo forget the gate bias term, ht-1Load prediction value, x, output by neural network of gated cyclic unit at last timetInputting characteristics;
2) the reset gate calculation formula is:
rt＝σ(Wrhht-1+Wrxxt+br) (7)
in the formula: wrhTo reset the gate cycle weight, WrxTo reset the gate input weights, brA reset gate bias term;
3) the current cell configuration will use a reset gate to store the past related information, which is calculated as:
Figure RE-GDA0002487237830000051
in the formula: whhAs cyclic weight, WhxAs input weights, bhFor the bias term, tanh is the activation function, the formula is calculated
Figure RE-GDA0002487237830000052
4) Predicted load value h finally output by GRUtThe updating gate is used as a weight vector, the current memory unit vector and the load predicted value at the previous moment are obtained by weighted average, and the calculation formula is as follows:
Figure RE-GDA0002487237830000053
in the formula: h istFor model output, representPredicting a load value at the moment t, namely a power failure loss load; the remaining variables are the intermediate variables and the training parameters of the model.
Preferably, the objective function is expressed by the following formula:
F＝(f1,f2) (10)
in the formula: f is an objective function; f. of1A reliability target; f. of2Is an economic objective;
the weights z (x) and the objective function formula are:
Figure RE-GDA0002487237830000054
Figure RE-GDA0002487237830000055
in the formula αiThe weight coefficient of the ith target; i is the target number, fi(x) For i objective functions, k is the number of objective functions, and the weight z (x) is the relative importance or value between the objectives;
1) reliability target
The reliability target calculation formula is as follows:
Figure RE-GDA0002487237830000056
in the formula: fiThe power failure frequency of the equipment i; n is a radical ofiThe number of users is influenced by power failure of equipment i; n is a radical ofsThe total number of the users;
Figure RE-GDA0002487237830000057
a distribution transformer set for power failure;
2) economic objective
The economic objective is to minimize the power failure loss and the maintenance cost of the power failure equipment, and the calculation formula is as follows:
Figure RE-GDA0002487237830000058
in the formula: ciThe unit time electricity charge loss caused by the power failure of the equipment i; t isiThe duration of the power failure of the device i; riWhich is the maintenance cost of the equipment i.
As a preferred scheme, the objective function is optimized and solved by a particle swarm optimization method.
Has the advantages that: the distribution transformer power failure electric quantity loss prediction method based on deep learning provided by the invention realizes calculation of power failure electric quantity, provides data support for improving power supply reliability, and realizes optimized management of planned power failure based on user load curve clustering. The big data technology is used for analyzing and mining the rule of distribution transformer load data, and the main realization method comprises the following steps:
(1) and realizing classification processing and fine analysis of the distribution transformation load curve based on a fuzzy C-means clustering algorithm. The method has the advantages that the scientific and reasonable power failure time period is determined, the power failure work is smoothly completed, the final purpose of reducing power failure loss is achieved, and the service quality of a power grid enterprise is improved. Meanwhile, the problem of electric quantity loss in the power failure time period of the distribution transformer area is converted into a short-term prediction problem of the load in the power failure time period;
(2) sorting the load influence factors at the moment to be predicted by using a Pearson correlation coefficient algorithm, extracting features with high correlation as input quantity of a neural network model of a gated loop unit, and constructing a training sample;
(3) the training samples are used for training the neural network of the gated circulation unit to obtain a load predicted value in the power failure time period, and a real load curve can be well fitted.
Therefore, the method can effectively and accurately evaluate the loss of the power failure electric quantity, and can reasonably provide scientific guidance basis for the work of destructive impact on the power grid caused by large power load change, stable control of the power grid, economic and efficient operation and the like.
Drawings
FIG. 1 is a schematic diagram of an RNN structural model and its development in accordance with an embodiment of the present invention;
FIG. 2 is a diagram of a neural network of gated cyclic units in an embodiment of the method of the present invention;
FIG. 3 is a clustering result of distribution transformer loads in the embodiment of the method of the present invention;
FIG. 4 is a graph of center load of different clusters according to an embodiment of the present invention;
FIG. 5 shows the weight of each type of load in the method of the present invention;
FIG. 6 is a graph illustrating the increase and percentage change of the daytime load cluster center according to an embodiment of the present invention;
FIG. 7 is a load prediction curve for a 99233 distribution transition from 0 to 2 in an embodiment of the method of the present invention;
FIG. 8 is a 12 hour load curve for a 99233 distribution transformer in accordance with an embodiment of the present invention;
fig. 9 is a load prediction curve of 99233 distribution different blackout time periods in the method embodiment of the invention.
Detailed Description
The present invention will be further described with reference to the accompanying drawings.
The invention discloses a distribution transformer power failure electric quantity loss prediction method based on deep learning, which comprises the following specific contents in one embodiment:
1 distribution transformer load characteristic clustering analysis
Based on fuzzy C-means clustering (FCM) algorithm, classification processing and fine analysis of distribution transformer load curves are realized. The FCM algorithm is implemented as follows:
step 1: determining a generic number c, a load number n and an initial membership matrix
Figure RE-GDA0002487237830000076
Wherein u isikRepresenting the kth load xkMembership belonging to class i, 0 represents step 0 iteration. Let the iteration variable be l, l ═ 1, denote step 1 iteration.
Step 2: calculating a membership matrix U using the formula(l)：
Figure RE-GDA0002487237830000071
In the formula:
Figure RE-GDA0002487237830000072
wherein x iskIs the k-th load,
Figure RE-GDA0002487237830000073
For the ith step to iterate the cluster centers of the ith class,
Figure RE-GDA0002487237830000074
for the jth class of cluster centers, i, j ∈ 1, 2.. c, is iterated in the l step.
And step 3: correcting the cluster center v of step l +1 by(l+1)：
Figure RE-GDA0002487237830000075
In the formula: x is an input value of the model and represents a load sequence of historical 96 points acquired by the adoption system; v is a clustering center of the model and represents a characteristic load sequence obtained by clustering analysis; the remaining variables are the intermediate variables and the training parameters of the model.
And 4, step 4: given a membership termination tolerance ε >0 when
Figure RE-GDA0002487237830000081
And stopping, otherwise, turning to the step 2, wherein l is l + 1.
2 power failure loss electric quantity prediction based on gate control circulation unit neural network
A Recurrent Neural Network (RNN) is typically characterized by not only internal feedback connections, but also feedforward connections between neurons. The RNN retains the previous information during calculation, and uses this information as part of the current input, and then calculates the current time output. Therefore, nodes between hidden layers of the RNN network structure have weight relation, and the relation is embodied in that RNN hidden layer input is obtained by the combined action of the output of the input layer at the moment and the output of the hidden layer at the previous moment, so that the dynamic characteristic is highlighted in the training step. Compared with a feedforward neural network, the RNN has stronger dynamic behavior and greatly improves the calculation capability. However, the RNN is prone to gradient disappearance and gradient explosion during training, so that the RNN cannot capture the influence of the remote output on the current output.
In fig. 1, x is an input vector, namely, the maximum 30 historical load values with the correlation value close to 1 with the load value to be predicted; u is a weight matrix of the input layer and the hidden layer; h is hidden layer output, and the load value o is predictedtThe hidden layer output is obtained through the action of the weight matrix and the activation function; v is the weight matrix of the hidden layer and the output layer, and the predicted load can be obtained as follows:
ot＝g(Vht) (3)
in the formula: g (-) is the output layer activation function, t represents the t-th time.
As can be seen from fig. 1, the RNN hidden layer input value at the current time t comprises two parts: 1) input x at the present momenttThe value after U action; 2) the previous time t-1, the output h of the hidden layert-1And the values acted on by the weight matrix W. The weight matrix W is the connection weight between the previous hidden layer and the current hidden layer. Therefore, the hidden layer output at the current time t is ht：
ht＝f(Uxt+Wht-1) (4)
In the formula: f (-) is the hidden layer activation function.
Gated cyclic unit neural network architecture, as shown in FIG. 2, x in FIG. 2tRepresenting characteristic inputs, namely the maximum 30 historical load values with the correlation value close to 1 with the load value to be predicted; z is a radical oftRepresenting an update gate structure; r istRepresenting a reset gate structure; h istRepresenting a characteristic output, i.e. a predicted load value;
Figure RE-GDA0002487237830000095
showing the memory cell structure.
For an input vector x, a weight vector W, an offset term b, the output value of the gate can be expressed as:
g(x)＝σ(Wx+b) (5)
in the formula: sigma is sigmoid activation function, and the calculation formula is
Figure RE-GDA0002487237830000091
The calculation process of each part of the GRU structure is described in conjunction with fig. 2 as follows:
1) at time step t, the calculation formula of the update gate is as follows:
zt＝σ(Wzhht-1+Wzxxt+bz) (6)
in the formula: wzhTo update the gate cycle weight, WzxTo update the gate input weights, bzTo forget the gate bias term, ht-1Load prediction value, x, output by neural network of gated cyclic unit at last timetIs a feature input.
2) The reset gate calculation formula is:
rt＝σ(Wrhht-1+Wrxxt+br) (7)
in the formula: wrhTo reset the gate cycle weight, WrxTo reset the gate input weights, brThe gate bias term is reset.
3) The current cell configuration will use a reset gate to store the past related information, which is calculated as:
Figure RE-GDA0002487237830000092
in the formula: whhAs cyclic weight, WhxAs input weights, bhFor the bias term, tanh is the activation function, the formula is calculated
Figure RE-GDA0002487237830000093
4) Predicted load value h finally output by GRU (gated cyclic Unit)tThe updating gate is used as a weight vector, the current memory unit vector and the load predicted value at the previous moment are obtained by weighted average, and the calculation formula is as follows:
Figure RE-GDA0002487237830000094
in the formula: h istThe load value of the moment t to be predicted, namely the power failure loss load, is output by the model; the remaining variables are the intermediate variables and the training parameters of the model.
Power outage optimization based on power utilization behavior of transformer area
3.1 Power failure management working principle
The power grid power failure management is a multi-objective optimization problem, and has a plurality of optimization objectives, and the invention considers two types of objectives: reliability targets (the influence of power outage on users with different electricity consumption behavior characteristics) and economic targets (users reduce electricity usage by power outage and cause economic loss to power companies). Let the influence factor of the power failure of the power supply system of the power grid be { x1,x2,…,xmM represents the number of influencing factors, such as the power failure times and duration of historical loads, user load types and the like; the power outage management model is described as R ═ F (x)1,x2,…,xm) In the formula, F represents an optimization function, and R represents the final specific power failure time arrangement obtained after optimization.
The multi-objective optimization problem is a problem which is composed of a plurality of targets which conflict with each other and influence each other, so that the plurality of targets obtain the optimal condition in a given area, and has a plurality of solving methods. The optimal reliability and economy is selected as an objective function, and the formula is as follows:
F＝(f1,f2) (10)
in the formula: f is an objective function; f. of1A reliability target; f. of2Is an economic objective. By analyzing both objectives, each optimization objective is optimized as much as possible.
The weights z (x) and the objective function set herein are formulated as:
Figure RE-GDA0002487237830000101
Figure RE-GDA0002487237830000102
in the formula αiThe weight coefficient of the ith target; i is the target number, fi(x) For i objective functions, k is the number of objective functions, and the weight z (x) is the relative importance or value between the objectives, i.e. the preference degree of the decision maker for the objective functions.
1) Reliability target
The reliability target calculation formula is as follows:
Figure RE-GDA0002487237830000103
in the formula: fiThe frequency of power failure for equipment i (year by year); n is a radical ofiThe number of users is influenced by power failure of equipment i; n is a radical ofsThe total number of the users;
Figure RE-GDA0002487237830000104
and distributing the transformer set for power failure.
2) Economic objective
The economic objective is to minimize the power failure loss and the maintenance cost of the power failure equipment, and the calculation formula is as follows:
Figure RE-GDA0002487237830000105
in the formula: ciThe unit time electricity charge loss caused by the power failure of the equipment i; t isiThe duration of the power failure of the device i; riWhich is the maintenance cost of the equipment i. The power loss of various users is shown in table 1.
TABLE 1 categorizing subscriber outage losses/min
Figure RE-GDA0002487237830000111
The constraint conditions of the power flow constraint of the power S of the ith balance node, the constraint of the power failure time t of the load, the constraint of the node voltage V and the constraint of the workload balance X are respectively as follows:
|Sl|≤Slmax(15)
timin≤ti≤timax(16)
Vimin≤Vi≤Vimax(17)
Xi≤Ximax(18)
Slmaxis the maximum power of the l-th balanced node, timin、timaxMinimum and maximum time, V, for which device i can be powered downimin、 VimaxMinimum and maximum voltage of device i node, XimaxDevice i workload balance max.
3.2 Power failure optimization objective function F solution based on particle swarm optimization
Setting the size of the population as m, deciding n dimensions of space, randomly generating initial speed and position of particles, and setting the speed of particles i at time t as
Figure RE-GDA0002487237830000112
Is positioned as
Figure RE-GDA0002487237830000113
Then the update formula of the velocity and position of the flight of the particle i in the dimensional subspace is as follows at j (j ═ 1,2, …, n) at time t + 1:
Figure RE-GDA0002487237830000114
in the formula: omega is an inertial weight function; c. C1、c2Is a learning factor; r is1、r2Is a random number between (0,1), and the velocity of the particles is limited to [ -V ]max,Vmax]In the meantime.
4 example test verification
4.1 distribution transform load characteristic fuzzy C-means clustering test analysis
Based on distribution transformer load data acquired from Yangzhou city in a certain province, 2000 pieces of load data of different distribution transformers in 8 months and 1 day in 2017 are randomly selected. Multiple times of simulation experiment comparison shows that the best clustering effect is obtained when the clustering number is 4.
Fig. 3 is a clustering result, in which a black line is an actual distribution load curve and a blue line is a calculated clustering center. As can be seen from the figure, different power utilization modes have different clustering center curve characteristics and show different load laws; and the distribution transformer in the same mode has similar load laws.
Fig. 4 shows different cluster center load curves, and 2000 distribution transformation load curves are divided into 4 types, i.e., daytime type load, all-day type load, night type load 1, and night type load 2, according to the cluster center characteristics. The proportion of different user type loads is shown in figure 5.
Day load: the load is higher in the daytime and lower in the nighttime, and 36 curves are provided. The load characteristic is good in regularity, and load reduction with a small amplitude occurs at about 12: 00.
All-day load: the load difference is small at 24 hours all day, and the load is basically not different between day and night. There are 1677 such curves.
Night load 1: the load level was higher at night and lower during the day, and there were 247 curves. The load uses a smaller amount of electricity as a whole.
Night load 2: such loads are also higher at night and lower during the day, but the overall load level is higher than the night type load 1. There were 40 such curves.
After all the load curves are clustered, different types of load clustering results and clustering center curves can be obtained. The clustering center curve reflects the load curve rule of the type of distribution transformer. The characteristics of the clustering center curve are analyzed, the loss electric quantity in different time periods is calculated and compared, and a scientific and reasonable time window for planned power failure optimization management is conveniently obtained. Table 2 shows data, amplification and percentage of the load per 15min of the day-long load cluster center, where the percentage in table 2 represents the ratio of the amplification of the current point to the load at the previous time, and is used to reflect the increase (decrease) of the load level in the time period. FIG. 6 is a graph of daytime load cluster center increase and percentage change. The clustering results of the whole day type load, night type load 1 and night type load 2 are not repeated for the sake of brevity
TABLE 2 daytime type load clustering center all-day load and its amplification
Figure RE-GDA0002487237830000131
4.2 Power failure loss electric quantity prediction test analysis based on gate control circulation unit neural network
Load data of each distribution transformer in Yangzhou city from 2017, 1 month and 1 day to 2018, 8 months and 31 days are adopted as data of the load data, the original load data is that the data of all the distribution transformers are arranged according to dates, in order to extract the input characteristics of a prediction model of each distribution transformer, the load data is firstly classified according to distribution transformer names, 5557 distribution transformers are provided, and each distribution transformer has 96-point load data of 608 days.
Since the blackout time is usually within 2 hours, it is assumed that the blackout time is 2 hours (8-point load) at maximum, and the power at each point in the blackout time is predicted. Because the input characteristics associated with each missing point are not necessarily the same, the load prediction model is established for the 8-point load respectively.
(1) Input feature selection
Because the load characteristics of the prediction time periods are different, 8-point loads need to be respectively modeled for prediction, and therefore, the input characteristics of each model are also different. Data of a distribution with ID of 99233 and a power failure period of 2017, 1, 30, 0 to 2 are taken as examples. Selecting the first 96 load points in the interval of power failure time (2h), wherein the characteristic quantity is named as x1-x96(ii) a The last 96 load points in the power failure time (2h) interval have the characteristic quantity name of x97-x192(ii) a The feature quantity name is x at 14 load points at the same time within 7 days before and after the power failure time point193-x206There are 206 dimensions of input data.
And calculating the correlation between each dimension and each power failure time point according to a Pearson correlation coefficient algorithm. The Pearson correlation coefficient calculation formula is as follows:
Figure RE-GDA0002487237830000141
in the formula: n is the number of training samples; x is corresponding input characteristic quantity and represents historical 206-dimensional load input needing analysis; and y is the actual load of the predicted power failure point.
The input characteristic quantity and the correlation coefficient thereof are shown in table 3 (the result of 60-120min is not repeated), and the characteristic quantity with the correlation coefficient >0.6 and the first 30 bits of 8 prediction points is selected as the input characteristic quantity according to calculation.
Table 315-60 min input characteristic quantity and related coefficient thereof
Figure RE-GDA0002487237830000142
Figure RE-GDA0002487237830000151
(2) Power loss prediction during power outage
Firstly, sorting load influence factors at a moment to be predicted by adopting a Pearson correlation coefficient algorithm, extracting 30-dimensional characteristics with the maximum correlation as input quantity, taking the load of a power failure point to be predicted as output quantity, respectively establishing load prediction models 15min-2h ahead of each point in power failure time (2h), preliminarily testing 1500 training samples of each model, and selecting the following two error indexes to measure the performance of a deterministic point prediction model, wherein an error statistical formula is as follows:
mean Absolute Percentage Error (MAPE):
Figure RE-GDA0002487237830000161
root Mean Square Error (RMSE):
Figure RE-GDA0002487237830000162
in the formula: n is the total number of the test points; y isi、
Figure RE-GDA0002487237830000163
Are respectively asAnd the real value and the predicted value of the user load of the ith test point.
The model prediction results are as follows, and table 4 shows that the distribution transformer with the ID of 99233 uses the GRU model to predict the load value of the power failure time interval from 1 month, 30 days 0 to 2 months in 2017; FIG. 7 is an actual load curve and a predicted load curve for a blackout period; fig. 8 is a load curve of the day of power failure at 12 hours, in which the model prediction period and results are shown in the dotted line frame.
Table 499233 predicted load values for 0-to-2 distribution
Figure RE-GDA0002487237830000164
In order to verify the prediction effect of the model on the lost electric quantity of the same distribution transformer in different power failure periods, the power failure time period from 1 month, 30 days and 20 hours to 22 hours in 2017 of the distribution transformer is continuously selected. Table 5 shows the load prediction results in the blackout period. Fig. 9 is a comparison of the 20 to 22 blackout periods with the 0 to 2 blackout period load prediction curves. The actual electric quantity can be obtained by integrating the actual load curve in the power failure time period, the lost electric quantity in the power failure time period can be obtained by integrating the predicted load curve, and the table 5 shows the load prediction error, the lost electric quantity prediction value and the lost electric quantity error in two time periods.
Table 599233 predicted load values for 20-22 distribution
Figure RE-GDA0002487237830000171
Load prediction error and lost electric quantity error of different time periods of distribution transformer of table 699233
Figure RE-GDA0002487237830000172
It can be seen from table 6 that there is a large difference between the prediction errors of the two blackout periods, and the actual load curves of the two blackout periods are analyzed. The load curve at 0-2 is relatively gentle and has small fluctuation and obvious descending trend; the load curve at 20-22 generally tends to rise, but at 21, there is a large fluctuation and a peak appears, so that the model cannot obtain a more accurate prediction result. Therefore, the prediction error is larger at the position where the load curve changes faster and the peak or the trough appears at a certain time. However, the predicted result can reflect the change trend of the curve more accurately relative to the load in the whole power failure time period.
And then analyzing the influence of the urban network or the rural network in the distribution and transformation attributes on the prediction result. 4 distribution transformers which belong to the urban network or the rural network respectively are selected to predict the loss electric quantity. Table 7, table 8 and table 9 are the load prediction results with IDs 37391, 38082 and 96273, respectively. Table 10 shows the information of 4 distribution transformers and the statistics of load prediction errors. Table 11 shows 4 predictions of the distribution transformer loss power and the error thereof.
Table 737391 predicted load value of distribution transformer
Figure RE-GDA0002487237830000173
Figure RE-GDA0002487237830000181
Table 838082 predicted load value of distribution transformer
Figure RE-GDA0002487237830000182
Table 996273 predicted load value of distribution transformer
Figure RE-GDA0002487237830000183
TABLE 10 distribution transform information and load prediction error statistics
Figure RE-GDA0002487237830000184
Figure RE-GDA0002487237830000191
(rural power network is 1 and city network is 2)
Table 11 distribution transformation information and loss electric quantity predicted value and error
Figure RE-GDA0002487237830000192
It can be seen from tables 8, 9, 10 and 11 that the load prediction MAPE indexes are basically the same, and the errors are basically the same, wherein the distribution transformers belong to the same county and have the same rated capacity, and the MAPE indexes of the load prediction are all between 7.5% and 7.7%; distribution transformers with the same rated capacity belong to different counties, the maximum difference of MAPE indexes of load prediction is about 1%, and the error difference is small, so that the distribution transformers belong to rural or urban networks and have small influence on prediction results.
4.3 Power failure optimization test analysis based on power utilization behavior of transformer area
The result of clustering analysis on the distribution transformation data by using the fuzzy C-means clustering algorithm is shown in section 4.1. Based on the power failure optimization model and the model solution provided by the text, the weights of the reliability target and the economic target are 0.5 and 0.5 respectively, and the average global fitness value is 0.932. The algorithm has achieved convergence in the previous 50 iterations, and the calculation results of each index of the power failure scheme are shown in table 12, wherein scheme 1 is the final optimal scheme, and other schemes 2 to 5 are optimal solutions formed in the partial iteration process.
Table 12 power-off scheme index calculation results
Figure RE-GDA0002487237830000193
Specific contents of the global optimal blackout scenario 1 are shown in table 13 (excerpt), where blackout start and end times are expressed as yyyymmddhmm, where yyyy represents year, MM represents month, dd represents date, hh represents hour, and MM represents minute.
Table 13 power cut plan table
Figure RE-GDA0002487237830000201
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.

Claims (7)

1. A method for predicting power loss of distribution transformer power failure based on deep learning is characterized by comprising the following steps: the method comprises the following steps:
classifying distribution transformer load curves based on a fuzzy C-means clustering algorithm to obtain various clustering center load curves, calculating the amplification of various distribution transformer loads and corresponding various center loads in various time periods, and selecting the time period with the amplification rate smaller than a threshold value as a time period to be predicted;
predicting a load value of a time period to be predicted by using the trained neural network of the gated circulation unit as power failure loss electric quantity;
an objective function is solved, and the electrical duration is obtained when the objective function is optimized.
2. The method for predicting the power loss of the distribution transformer based on the deep learning of claim 1, wherein: the fuzzy C-means clustering algorithm comprises the following steps:
step 1: determining a generic number c, a load number n and an initial membership matrix
Figure RE-FDA0002487237820000011
Wherein u isikRepresenting the kth load xkMembership belonging to class i, 0 represents step 0 iteration. Let iteration variable be l, l ═ 1 denote step 1 iteration;
step 2: calculating a membership matrix U using the formula(l)：
Figure RE-FDA0002487237820000012
In the formula:
Figure RE-FDA0002487237820000013
wherein x iskIs the k-th load,
Figure RE-FDA0002487237820000014
For the ith step to iterate the cluster centers of the ith class,
Figure RE-FDA0002487237820000015
iterating the clustering centers of the jth class i, j ∈ 1, 2.. c in the ith step;
and step 3: correcting the cluster center v of step l +1 by(l+1)：
Figure RE-FDA0002487237820000016
In the formula: x is an input value of the model and represents a load sequence of historical 96 points acquired by the adoption system; v is a clustering center of the model and represents a characteristic load sequence obtained by clustering analysis; the other variables are intermediate variables and training parameters of the model;
and 4, step 4: given a membership termination tolerance ε >0 when
Figure RE-FDA0002487237820000021
And stopping, otherwise, turning to the step 2, wherein l is l + 1.
3. The method for predicting the power loss of the distribution transformer based on the deep learning of claim 1, wherein: the cluster center load curves are divided into 4 classes, including: day type load, all day type load, night type load 1, and night type load 2.
4. The method for predicting the power loss of the distribution transformer based on the deep learning of claim 1, wherein: the training steps of the gated loop unit neural network are as follows:
acquiring historical load values of time periods of each day;
acquiring a historical load value of a time period to be measured on a date to be measured;
calculating the Pearson correlation coefficient of the historical load value of each time period on each day and the historical load value of the time period to be detected on the date to be detected, and sorting the historical load values of the time periods on the first N days from large to small according to the coefficient, wherein the historical load values of each time period on each day are taken as input characteristic quantities;
taking a historical load value of a time period to be measured on a date to be measured as an output characteristic quantity;
and learning the neural network of the gate control cycle unit by using the input characteristic quantity and the output characteristic quantity to obtain intermediate variables and training parameters of the neural network of the gate control cycle unit.
5. The method for predicting the power loss of the distribution transformer based on the deep learning of claim 1, wherein: the gated cyclic unit neural network:
for an input vector x, a weight vector W, an offset term b, the output value of the gate is expressed as:
g(x)＝σ(Wx+b) (5)
in the formula: sigma is sigmoid activation function, and the calculation formula is
Figure RE-FDA0002487237820000022
1) At time step t, the calculation formula of the update gate is as follows:
zt＝σ(Wzhht-1+Wzxxt+bz) (6)
in the formula: wzhTo update the gate cycle weight, WzxTo update the gate input weights, bzTo forget the gate bias term, ht-1Load prediction value, x, output by neural network of gated cyclic unit at last timetInputting characteristics;
2) the reset gate calculation formula is:
rt＝σ(Wrhht-1+Wrxxt+br) (7)
in the formula: wrhTo reset the gate cycle weight, WrxTo reset the gate input weights, brA reset gate bias term;
3) the current cell configuration will use a reset gate to store the past related information, which is calculated as:
Figure RE-FDA0002487237820000031
in the formula: whhAs cyclic weight, WhxAs input weights, bhFor the bias term, tanh is the activation function, the formula is calculated
Figure RE-FDA0002487237820000032
4) Predicted load value h finally output by GRUtThe updating gate is used as a weight vector, the current memory unit vector and the load predicted value at the previous moment are obtained by weighted average, and the calculation formula is as follows:
Figure RE-FDA0002487237820000033
in the formula: h istThe load value of the moment t to be predicted, namely the power failure loss load, is output by the model; the remaining variables are the intermediate variables and the training parameters of the model.
6. The method for predicting the power loss of the distribution transformer based on the deep learning of claim 1, wherein: the target function is expressed as follows:
F＝(f1,f2) (10)
in the formula: f is an objective function; f. of1A reliability target; f. of2Is an economic objective;
the weights z (x) and the objective function formula are:
Figure RE-FDA0002487237820000041
Figure RE-FDA0002487237820000042
in the formula αiThe weight coefficient of the ith target; i is the target number, fi(x) For i objective functions, k is the number of objective functions, and the weight z (x) is the relative importance or value between the objectives;
1) reliability target
The reliability target calculation formula is as follows:
Figure RE-FDA0002487237820000043
in the formula: fiThe power failure frequency of the equipment i; n is a radical ofiThe number of users is influenced by power failure of equipment i; n is a radical ofsThe total number of the users;
Figure RE-FDA0002487237820000044
a distribution transformer set for power failure;
2) economic objective
The economic objective is to minimize the power failure loss and the maintenance cost of the power failure equipment, and the calculation formula is as follows:
Figure RE-FDA0002487237820000045
in the formula: ciThe unit time electricity charge loss caused by the power failure of the equipment i; t isiThe duration of the power failure of the device i; riWhich is the maintenance cost of the equipment i.
7. The method for predicting the power loss of the distribution transformer based on the deep learning of claim 1, wherein: and the objective function is optimized and solved by a particle swarm method.
CN201911169829.3A 2019-11-22 2019-11-22 Distribution transformer power failure and power loss prediction method based on deep learning Pending CN111325315A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911169829.3A CN111325315A (en) 2019-11-22 2019-11-22 Distribution transformer power failure and power loss prediction method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911169829.3A CN111325315A (en) 2019-11-22 2019-11-22 Distribution transformer power failure and power loss prediction method based on deep learning

Publications (1)

Publication Number Publication Date
CN111325315A true CN111325315A (en) 2020-06-23

Family

ID=71170903

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911169829.3A Pending CN111325315A (en) 2019-11-22 2019-11-22 Distribution transformer power failure and power loss prediction method based on deep learning

Country Status (1)

Country Link
CN (1) CN111325315A (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112446614A (en) * 2020-11-26 2021-03-05 深圳供电局有限公司 Uninterrupted power operation priority evaluation method based on user power failure loss
CN113657682A (en) * 2021-08-24 2021-11-16 国网安徽省电力有限公司 Power failure data classification prediction method and system, storage medium and computing device
CN113872186A (en) * 2021-09-06 2021-12-31 广西电网有限责任公司电力科学研究院 Simulation learning-based micro-grid control strategy recommendation model training system
CN116526496A (en) * 2023-06-16 2023-08-01 国网山西省电力公司晋城供电公司 Novel auxiliary decision-making method for power system load control
CN116667326A (en) * 2023-05-30 2023-08-29 淮阴工学院 Electric automobile charging load prediction method
CN117408394A (en) * 2023-12-14 2024-01-16 国网天津市电力公司电力科学研究院 Carbon emission factor prediction method and device for electric power system and electronic equipment

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104331628A (en) * 2014-11-17 2015-02-04 国网河南省电力公司 Power grid safety aggregative indicator evaluation method
CN105574611A (en) * 2015-12-14 2016-05-11 国家电网公司 Overhaul plan optimization method containing distributed power distribution network
CN105631483A (en) * 2016-03-08 2016-06-01 国家电网公司 Method and device for predicting short-term power load
CN105762793A (en) * 2016-03-16 2016-07-13 国网江苏省电力公司电力科学研究院 Distribution network fault power failure loss assessment method considering distributed new energy
CN108960596A (en) * 2018-06-19 2018-12-07 国网湖南省电力有限公司 A kind of power consumption management method based on duration of load application elasticity
CN109063911A (en) * 2018-08-03 2018-12-21 天津相和电气科技有限公司 A kind of Load aggregation body regrouping prediction method based on gating cycle unit networks
CN109886498A (en) * 2019-03-01 2019-06-14 北京邮电大学 A kind of EMD-GRU Short-Term Load Forecasting Method based on feature selecting
CN109995094A (en) * 2019-01-31 2019-07-09 中国电力科学研究院有限公司 A kind of method and system for planning of alternating current-direct current mixing micro-capacitance sensor

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104331628A (en) * 2014-11-17 2015-02-04 国网河南省电力公司 Power grid safety aggregative indicator evaluation method
CN105574611A (en) * 2015-12-14 2016-05-11 国家电网公司 Overhaul plan optimization method containing distributed power distribution network
CN105631483A (en) * 2016-03-08 2016-06-01 国家电网公司 Method and device for predicting short-term power load
CN105762793A (en) * 2016-03-16 2016-07-13 国网江苏省电力公司电力科学研究院 Distribution network fault power failure loss assessment method considering distributed new energy
CN108960596A (en) * 2018-06-19 2018-12-07 国网湖南省电力有限公司 A kind of power consumption management method based on duration of load application elasticity
CN109063911A (en) * 2018-08-03 2018-12-21 天津相和电气科技有限公司 A kind of Load aggregation body regrouping prediction method based on gating cycle unit networks
CN109995094A (en) * 2019-01-31 2019-07-09 中国电力科学研究院有限公司 A kind of method and system for planning of alternating current-direct current mixing micro-capacitance sensor
CN109886498A (en) * 2019-03-01 2019-06-14 北京邮电大学 A kind of EMD-GRU Short-Term Load Forecasting Method based on feature selecting

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
L. KUAN 等: ""Short-term electricity load forecasting method based on multilayered self-normalizing GRU network"", 《IEEE》 *
郭松荣 等: ""电网供电系统多目标停电优化管理仿真"", 《计算机仿真》 *

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112446614A (en) * 2020-11-26 2021-03-05 深圳供电局有限公司 Uninterrupted power operation priority evaluation method based on user power failure loss
CN113657682A (en) * 2021-08-24 2021-11-16 国网安徽省电力有限公司 Power failure data classification prediction method and system, storage medium and computing device
CN113657682B (en) * 2021-08-24 2023-12-29 国网安徽省电力有限公司 Power failure data classification prediction method, system, storage medium and computing device
CN113872186A (en) * 2021-09-06 2021-12-31 广西电网有限责任公司电力科学研究院 Simulation learning-based micro-grid control strategy recommendation model training system
CN116667326A (en) * 2023-05-30 2023-08-29 淮阴工学院 Electric automobile charging load prediction method
CN116667326B (en) * 2023-05-30 2024-02-23 淮阴工学院 Electric automobile charging load prediction method
CN116526496A (en) * 2023-06-16 2023-08-01 国网山西省电力公司晋城供电公司 Novel auxiliary decision-making method for power system load control
CN116526496B (en) * 2023-06-16 2023-09-08 国网山西省电力公司晋城供电公司 Novel auxiliary decision-making method for power system load control
CN117408394A (en) * 2023-12-14 2024-01-16 国网天津市电力公司电力科学研究院 Carbon emission factor prediction method and device for electric power system and electronic equipment

Similar Documents

Publication Publication Date Title
CN111325315A (en) Distribution transformer power failure and power loss prediction method based on deep learning
CN109711620B (en) Short-term power load prediction method based on GRU neural network and transfer learning
Ke et al. Short-term electrical load forecasting method based on stacked auto-encoding and GRU neural network
CN113053115B (en) Traffic prediction method based on multi-scale graph convolution network model
CN110766212B (en) Ultra-short-term photovoltaic power prediction method for historical data missing electric field
Khotanzad et al. A neuro-fuzzy approach to short-term load forecasting in a price-sensitive environment
Wu et al. Combined model with secondary decomposition-model selection and sample selection for multi-step wind power forecasting
CN109359786A (en) A kind of power station area short-term load forecasting method
CN110674999A (en) Cell load prediction method based on improved clustering and long-short term memory deep learning
CN113554466B (en) Short-term electricity consumption prediction model construction method, prediction method and device
CN112232561A (en) Power load probability prediction method based on constrained parallel LSTM quantile regression
Lv et al. An improved long short-term memory neural network for stock forecast
CN115688993A (en) Short-term power load prediction method suitable for power distribution station area
Wu et al. Automatic fuzzy model identification for short-term load forecast
CN111784019A (en) Power load processing method and device
Yang et al. A novel short-term load forecasting approach for data-poor areas based on K-MIFS-XGBoost and transfer-learning
CN112418485A (en) Household load prediction method and system based on load characteristics and power consumption behavior mode
CN113537607B (en) Power failure prediction method
CN113762591B (en) Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning
CN108615091A (en) Electric power meteorology load data prediction technique based on cluster screening and neural network
Bin et al. A short-term power load forecasting method based on eemd-abgru
Wang et al. Optimization of Convolutional Long Short-Term Memory Hybrid Neural Network Model Based on Genetic Algorithm for Weather Prediction
Di et al. Online gaussian process regression for short-term probabilistic interval load prediction
Sala et al. Enhanced load forecasting methodology by means of probabilistic prediction intervals estimation
Dewangan et al. Load Forecasting Models in Smart Grid Using Smart Meter Information: A Review. Energies 2023, 16, 1404

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20200623

RJ01 Rejection of invention patent application after publication