CN113379005B - Intelligent energy management system and method for power grid power equipment - Google Patents

Intelligent energy management system and method for power grid power equipment Download PDF

Info

Publication number
CN113379005B
CN113379005B CN202110921796.4A CN202110921796A CN113379005B CN 113379005 B CN113379005 B CN 113379005B CN 202110921796 A CN202110921796 A CN 202110921796A CN 113379005 B CN113379005 B CN 113379005B
Authority
CN
China
Prior art keywords
accuracy
network
equipment
training
power
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
Application number
CN202110921796.4A
Other languages
Chinese (zh)
Other versions
CN113379005A (en
Inventor
陈早军
陈敏
李志刚
高栋
陈天雁
张元吉
苏宝聚
刘丽敏
姬脉胜
阮敬稳
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.)
Windsun Science and Technology Co Ltd
Original Assignee
Windsun Science and Technology 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 Windsun Science and Technology Co Ltd filed Critical Windsun Science and Technology Co Ltd
Priority to CN202110921796.4A priority Critical patent/CN113379005B/en
Publication of CN113379005A publication Critical patent/CN113379005A/en
Application granted granted Critical
Publication of CN113379005B publication Critical patent/CN113379005B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • 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
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply

Abstract

The invention relates to a system and a method for intelligently managing power grid power equipment energy, wherein the system comprises: the data acquisition module is used for acquiring basic information of the power equipment; the device comprises an electric energy output quantity prediction module, a device fluctuation evaluation module and a power generation device, wherein the electric energy output quantity prediction module is used for training an output power prediction network and determining a corresponding training sample; other power equipment is used as secondary stable power generation equipment; and the equipment control module is used for selecting the primary stable power generation equipment to carry out grid-connected input control according to the demand of power consumption, and the secondary stable power generation equipment is used as standby equipment. The method can reasonably classify the power equipment, and specifically screen out the power equipment which is beneficial to the safety and the stability of the power grid, and preferentially adopts the primary power generation equipment to be put into use under the condition of limited power demand, and takes the secondary power generation equipment with relatively low safety as standby equipment, so that the method can be helpful for improving the safety and the stability of the power grid.

Description

Intelligent energy management system and method for power grid power equipment
Technical Field
The invention relates to the field of artificial intelligence and automatic management of power generation equipment in the power industry, in particular to a system and a method for intelligently managing energy of power grid power equipment.
Background
In the prior art, for monitoring the state of the power equipment of the power grid, indexes such as output voltage, current and power of the power equipment are generally monitored, taking power as an example, the monitored power is an actual value of power output in the running process of the equipment, the state quantity and the state variation of the power are continuously monitored and compared with a correspondingly set safety threshold, and if the monitored power exceeds the threshold, corresponding protection is started, for example, a line connecting the power equipment with the power grid is cut off, the power equipment is shut down, and then after-the-fact fault detection and maintenance processing are performed.
The monitoring and management mode of the power equipment has the disadvantages that the reason that the unstable fault of the power grid can occur is found out from the power equipment after the emergency treatment is carried out through the relay protection device on the line only when the unstable fault of the power grid line occurs, namely the unstable fault of the power grid line is finally caused due to the reasons of equipment aging, overlong running time and the like of the power equipment, so that the safety and the stability of the power grid are reduced by using the power equipment with the potential safety hazard.
Disclosure of Invention
The invention aims to provide a power grid power equipment energy intelligent management system and method, which are used for solving the problem that the safety and the stability of a power grid are reduced by the existing power equipment monitoring and management method.
Therefore, the adopted technical scheme is as follows:
the data acquisition module is used for acquiring basic information of the power equipment, wherein the basic information comprises equipment aging degree, equipment model and equipment running time; the device is also used for acquiring the actual output power of the equipment when the running time is t;
the electric energy output quantity prediction module is used for randomly selecting the basic information of a set number of electric equipment in batches from the basic information of all the electric equipment, the basic information of all batches of electric equipment is used as training samples, the training samples of all batches are respectively used as input, the actual output power of the equipment in all the batches of the training samples when the running time is t is used as output, and the actual output power is respectively and independently used as outputTraining an output power prediction network, wherein the output power prediction network adopts a neural network, outputting the accuracy of each output power prediction network after training, and selecting a reference network according to the accuracy of each output power prediction network as an optimal network for predicting the output power of the power equipment; the basic information of the power equipment is specifically equipment aging degree
Figure DEST_PATH_IMAGE001
Type of device
Figure 676302DEST_PATH_IMAGE002
Length of operation of the apparatus
Figure 633107DEST_PATH_IMAGE003
；
The equipment volatility evaluation module is used for selecting corresponding electric equipment in the training samples participating in the training of the optimal network as primary stable power generation equipment; selecting power equipment which does not participate in the optimal network training as secondary stable power generation equipment;
and the equipment control module is used for selecting the primary stable power generation equipment to carry out grid-connected input control according to the power requirement and a preset priority input sequence for each power generation equipment in all the primary stable power generation equipment determined by the equipment volatility evaluation module, and using the rest primary stable power generation equipment and the rest secondary stable power generation equipment as standby equipment.
Preferably, the specific construction process of the neural network is as follows:
collecting training samples including basic information of a plurality of electric power equipment to be screened, wherein the basic information is expressed as
Figure 157498DEST_PATH_IMAGE004
Wherein, in the step (A),
Figure 912221DEST_PATH_IMAGE005
is as follows
Figure 828747DEST_PATH_IMAGE005
The device identification of the individual power device,
Figure 10199DEST_PATH_IMAGE004
three symbols in the sequence are
Figure 669107DEST_PATH_IMAGE005
Equipment aging degree, equipment model and equipment running time of each power equipment;
training output power prediction network, including power output quantity prediction encoder and full connection layer, network inputs basic information of electric power equipment, i.e.
Figure 760559DEST_PATH_IMAGE004
Obtaining the characteristic vector after the coding by the power output quantity prediction coder, sending the characteristic vector into a full connection layer, wherein the running time of the network output equipment is as long astAnd the actual value of the output power of the hour is obtained through detection.
Preferably, the selecting a reference network according to the accuracy of each output power prediction network as the optimal network for predicting the output power of the power equipment includes:
(A) using a sample of basic information of the first batch of electrical equipment as a first training set to predict the output power of the network W1Training and judging the network W1Whether the predicted value of each output quantity is accurate or not is determined according to the network W1Outputting the number of accurate power output predicted values, and determining the network W after the training of the first training set1Accuracy A of1；
(B) In the amount of
Figure 309703DEST_PATH_IMAGE006
Taking the second batch of samples as a second training set, and sending the second training set serving as a first verification set into the trained network W1And obtaining the verified network W of the first verification set1Accuracy of
Figure 244030DEST_PATH_IMAGE007
(ii) a Then using the second training set as trainingSample, output Power prediction network W for initial parameters2Training is carried out, and a second training set is obtained to train the network W2Accuracy of
Figure 487054DEST_PATH_IMAGE008
；
(C) Obtaining a trained network W of a first training set1Accuracy of
Figure 289138DEST_PATH_IMAGE009
Network W after verification with first verification set1Accuracy of
Figure 340009DEST_PATH_IMAGE007
As a first accuracy deviation; obtaining a trained network W of a first training set1Accuracy of
Figure 498982DEST_PATH_IMAGE009
Network W after training with the second training set2Accuracy of
Figure 981916DEST_PATH_IMAGE008
As a second accuracy deviation; judging whether a set condition is met or not according to the first accuracy deviation and the second accuracy deviation, and determining a reference network;
(D) then the number is
Figure 605884DEST_PATH_IMAGE006
Taking the next batch of samples as a new second training set, and taking the reference network W determined for the first time1' the corresponding training set is used as a new first training set, the processing of the first training set and the second training set in the above steps (A) - (C) is repeated, and the reference network W is updated again1’；
(E) Repeating step (D) above until the last group is of quantity
Figure 440372DEST_PATH_IMAGE006
The last batch of samples of the individuals is used as a new second training setLast determined reference network W1' the corresponding training set is used as a new first training set, the processing of the first training set and the second training set in the steps (A) - (C) is repeated, and the optimal reference network W is obtained by updating1' to this point.
Preferably, in the step (C), whether the set condition is satisfied is determined according to the magnitudes of the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is as follows:
when the first accuracy deviation and the second accuracy deviation meet the first set condition, the network is retrained again by using the second training set to obtain the reference network W1', the formula of the first setting condition is as follows:
Figure 349291DEST_PATH_IMAGE010
Figure 983770DEST_PATH_IMAGE011
in the formula (I), the compound is shown in the specification,
Figure 638743DEST_PATH_IMAGE012
in order to preset the first accuracy threshold,
Figure 644132DEST_PATH_IMAGE013
a second accuracy threshold is preset.
Preferably, in the step (C), whether the set condition is satisfied is determined according to the magnitudes of the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is as follows:
when the first accuracy deviation and the second accuracy deviation satisfy a second set condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1', the formula of the second setting condition is as follows:
Figure 899402DEST_PATH_IMAGE010
Figure 111422DEST_PATH_IMAGE014
in the formula (I), the compound is shown in the specification,
Figure 122365DEST_PATH_IMAGE015
in order to be the first deviation of the accuracy,
Figure 561306DEST_PATH_IMAGE016
for the second accuracy, the first accuracy is biased,
Figure 769784DEST_PATH_IMAGE017
in order to preset the first accuracy threshold,
Figure 663790DEST_PATH_IMAGE018
a second accuracy threshold is preset.
Preferably, in the step (C), whether the set condition is satisfied is determined according to the magnitudes of the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is as follows:
when the first accuracy deviation and the second accuracy deviation meet a third set condition, combining the first training set and the second training set, and training the output power prediction network of the initial parameters by using the combined training set to obtain a reference network W1', the formula of the third setting condition is as follows:
Figure 545552DEST_PATH_IMAGE019
Figure 729628DEST_PATH_IMAGE011
in the formula (I), the compound is shown in the specification,
Figure 631594DEST_PATH_IMAGE020
in order to be the first deviation of the accuracy,
Figure 332221DEST_PATH_IMAGE021
for the second accuracy, the first accuracy is biased,
Figure 472084DEST_PATH_IMAGE022
in order to preset the first accuracy threshold,
Figure 273335DEST_PATH_IMAGE023
a second accuracy threshold is preset.
Preferably, in the step (C), whether the set condition is satisfied is determined according to the magnitudes of the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is as follows:
when the first accuracy deviation and the second accuracy deviation satisfy a fourth setting condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1', the formula of the fourth setting condition is as follows:
Figure 305007DEST_PATH_IMAGE019
Figure 931029DEST_PATH_IMAGE014
in the formula (I), the compound is shown in the specification,
Figure 250365DEST_PATH_IMAGE020
in order to be the first deviation of the accuracy,
Figure 280638DEST_PATH_IMAGE021
for the second accuracy, the first accuracy is biased,
Figure 799607DEST_PATH_IMAGE022
in order to preset the first accuracy threshold,
Figure 23128DEST_PATH_IMAGE023
to prepareA second accuracy threshold is set.
In a second aspect, the invention provides a method for intelligently managing energy of power grid power equipment, which comprises the following steps:
s1, acquiring basic information of the power equipment, wherein the basic information comprises equipment aging degree, equipment model and equipment running time; the actual output power of the equipment when the running time is t is also obtained;
s2, randomly selecting the basic information of a set number of electric power equipment in batches from the basic information of all the electric power equipment, using the basic information of all the electric power equipment in each batch as training samples, respectively using the training samples in each batch as input, using the actual output power of the equipment in the training samples in each batch when the operation time is t as output, respectively and independently training an output power prediction network, wherein the output power prediction network adopts a neural network, outputting the accuracy of each output power prediction network after training, and selecting a reference network according to the accuracy of each output power prediction network as an optimal network for predicting the output power of the electric power equipment; the basic information of the power equipment comprises equipment aging degree
Figure 747370DEST_PATH_IMAGE024
Type of device
Figure 872846DEST_PATH_IMAGE025
Length of operation of the apparatus
Figure 767858DEST_PATH_IMAGE026
；
S3, selecting corresponding power equipment in the training samples participating in the optimal network training as primary stable power generation equipment; selecting power equipment which does not participate in the optimal network training as secondary stable power generation equipment;
and S4, selecting the first-stage stable power generation equipment to carry out grid-connected input control according to the power requirement and the preset priority input sequence for each power generation equipment in all the first-stage stable power generation equipment, and taking the rest first-stage stable power generation equipment and the rest second-stage stable power generation equipment as standby equipment.
Preferably, in step S2, the selecting a reference network according to the accuracy of each output power prediction network as the optimal network for predicting the output power of the electrical power equipment includes:
(A) using a sample of basic information of the first batch of electrical equipment as a first training set to predict the output power of the network W1Training and judging the network W1Whether the predicted value of each output quantity is accurate or not is determined according to the network W1Outputting the number of accurate power output predicted values, and determining the network W after the training of the first training set1Accuracy A of1；
(B) In the amount of
Figure 44338DEST_PATH_IMAGE027
Taking the second batch of samples as a second training set, and sending the second training set serving as a first verification set into the trained network W1And obtaining the verified network W of the first verification set1Accuracy of
Figure 953912DEST_PATH_IMAGE028
(ii) a Then, the second training set is used as a training sample, and the network W is predicted according to the output power of the initial parameters2Training is carried out, and a second training set is obtained to train the network W2Accuracy of
Figure 981780DEST_PATH_IMAGE029
；
(C) Obtaining a trained network W of a first training set1Accuracy of
Figure 56701DEST_PATH_IMAGE030
Network W after verification with first verification set1Accuracy of
Figure 687271DEST_PATH_IMAGE028
As a first accuracy deviation; obtaining a trained network W of a first training set1Accuracy of
Figure 356412DEST_PATH_IMAGE030
Network W after training with the second training set2Accuracy of
Figure 552251DEST_PATH_IMAGE029
As a second accuracy deviation; judging whether a set condition is met or not according to the first accuracy deviation and the second accuracy deviation, and determining a reference network;
(D) then the number is
Figure 421856DEST_PATH_IMAGE027
Taking the next batch of samples as a new second training set, and taking the reference network W determined for the first time1' the corresponding training set is used as a new first training set, the processing of the first training set and the second training set in the above steps (A) - (C) is repeated, and the reference network W is updated again1’；
(E) Repeating step (D) above until the last group is of quantity
Figure 875359DEST_PATH_IMAGE027
Taking the last batch of samples as a new second training set, and taking the reference network W determined last time1' the corresponding training set is used as a new first training set, the processing of the first training set and the second training set in the steps (A) - (C) is repeated, and the optimal reference network W is obtained by updating1' to this point.
Preferably, in the step (C), whether the set condition is satisfied is determined according to the magnitudes of the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is as follows:
when the first accuracy deviation and the second accuracy deviation meet the first set condition, the network is retrained again by using the second training set to obtain the reference network W1', the formula of the first setting condition is as follows:
Figure 553333DEST_PATH_IMAGE010
Figure 920074DEST_PATH_IMAGE011
in the formula (I), the compound is shown in the specification,
Figure 808133DEST_PATH_IMAGE031
in order to be the first deviation of the accuracy,
Figure 32704DEST_PATH_IMAGE032
for the second accuracy, the first accuracy is biased,
Figure 31097DEST_PATH_IMAGE033
in order to preset the first accuracy threshold,
Figure 40510DEST_PATH_IMAGE034
a second accuracy threshold is preset;
when the first accuracy deviation and the second accuracy deviation satisfy a second set condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1', the formula of the second setting condition is as follows:
Figure 763003DEST_PATH_IMAGE010
Figure 273488DEST_PATH_IMAGE014
in the formula (I), the compound is shown in the specification,
Figure 54885DEST_PATH_IMAGE031
in order to be the first deviation of the accuracy,
Figure 300446DEST_PATH_IMAGE032
for the second accuracy, the first accuracy is biased,
Figure 100780DEST_PATH_IMAGE033
in order to preset the first accuracy threshold,
Figure 162759DEST_PATH_IMAGE034
a second accuracy threshold is preset;
when the first accuracy deviation and the second accuracy deviation meet a third set condition, combining the first training set and the second training set, and training the output power prediction network of the initial parameters by using the combined training set to obtain a reference network W1', the formula of the third setting condition is as follows:
Figure 263308DEST_PATH_IMAGE019
Figure 148611DEST_PATH_IMAGE011
in the formula (I), the compound is shown in the specification,
Figure 213739DEST_PATH_IMAGE031
in order to be the first deviation of the accuracy,
Figure 472550DEST_PATH_IMAGE032
for the second accuracy, the first accuracy is biased,
Figure 696115DEST_PATH_IMAGE033
in order to preset the first accuracy threshold,
Figure 624756DEST_PATH_IMAGE034
a second accuracy threshold is preset;
when the first accuracy deviation and the second accuracy deviation satisfy a fourth setting condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1', the formula of the fourth setting condition is as follows:
Figure 396754DEST_PATH_IMAGE019
Figure 304929DEST_PATH_IMAGE014
in the formula (I), the compound is shown in the specification,
Figure 786595DEST_PATH_IMAGE031
in order to be the first deviation of the accuracy,
Figure 296859DEST_PATH_IMAGE032
for the second accuracy, the first accuracy is biased,
Figure 496765DEST_PATH_IMAGE033
in order to preset the first accuracy threshold,
Figure 224943DEST_PATH_IMAGE034
a second accuracy threshold is preset.
The invention has the following beneficial effects:
the invention discloses an intelligent management system and method for standby energy, which divide basic information of different power equipment into a plurality of batches, perform independent neural network training on equipment information of each batch, determine an optimal network according to the accuracy of each neural network, and finally utilize a sample participating in the optimal network training to reasonably classify the power equipment into primary power generation equipment with small volatility and secondary power generation equipment with relatively large volatility, namely the power equipment in the sample participating in the optimal network training is primary power generation equipment, and the power equipment in the sample not participating in the optimal network training is secondary power generation equipment.
Therefore, in the control process of the power equipment which is put into operation, the power equipment which is beneficial to the safety and the stability of the power grid can be screened in a targeted mode, the equipment is preferentially adopted to be put into use under the condition that the power demand is limited, the secondary power generation equipment with relatively low safety is used as standby equipment, and output power fluctuation is more likely to occur to the power generation equipment relative to the primary power generation equipment, so that the power grid is broken down. In addition, the intelligent automatic classification method can realize the intelligent automatic classification of the primary power generation equipment and the secondary power generation equipment, and the automatic classification speed is higher and more reasonable.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a schematic diagram of an intelligent energy management system for grid power equipment in embodiment 1 of the present invention;
fig. 2 is a flowchart of a method for intelligently managing energy of grid power equipment in embodiment 2 of the present invention.
Detailed Description
The embodiments provided by the invention are specifically described below with reference to the accompanying drawings.
Example 1:
the invention discloses an intelligent energy management system for power grid power equipment, which mainly aims to realize the following steps: the invention relates to efficient and quick-response intelligent management of power equipment, which aims at the following specific scenes: under the intelligent power grid scene, demand side demand electric quantity is known, does not consider transmission loss, carries out overall management to power equipment's switch.
Specifically, the system for intelligently managing the energy of the grid power equipment as shown in fig. 1 includes:
the data acquisition module is used for acquiring basic information of the power equipment, wherein the basic information comprises equipment aging degree, equipment model and current running time of the equipment.
The basic information of the above electric power equipment is known as the basic attribute of the electric power equipment, wherein the rated power of the equipment corresponding to the equipment can be obtained according to the model of the equipment.
And (II) an electric energy output quantity prediction module, which is used for randomly selecting the basic information of a set number of electric power equipment in batches from the basic information of all the electric power equipment, using the basic information of all the batches of electric power equipment as training samples, respectively using the training samples of all the batches as input, using the actual output power of the equipment in the training samples of all the batches when the running time is t as output, respectively and independently training output power prediction networks, outputting the accuracy of all the output power prediction networks after training, and selecting a reference network according to the accuracy of all the output power prediction networks to be used as an optimal network for predicting the output power of the electric power equipment.
In the module, an output power prediction network for predicting theoretical output power is realized by adopting a neural network, and the specific construction process of the neural network is as follows:
collecting training samples including basic information of a plurality of power equipment to be screened (screening primary stable power generation equipment), specifically the aging degree of the equipment
Figure 295536DEST_PATH_IMAGE035
Type of device
Figure 297471DEST_PATH_IMAGE036
Length of operation of the apparatus
Figure 610772DEST_PATH_IMAGE037
The basic information is expressed as
Figure 605623DEST_PATH_IMAGE038
Wherein, in the step (A),
Figure DEST_PATH_IMAGE039
is as follows
Figure 658286DEST_PATH_IMAGE039
The equipment identification of each power equipment is marked by an implementer, and all the equipment needs to be ensured to be marked and not to be repeatedly marked. In addition, the collected information also includesThe running time of the power equipment is
Figure 224265DEST_PATH_IMAGE037
Actual output power of the time.
For a single device, based on its initial operating duration
Figure 723730DEST_PATH_IMAGE040
Obtaining the long information of the running time of the equipment with continuous n periods, i.e.
Figure 777399DEST_PATH_IMAGE041
，
Figure 88164DEST_PATH_IMAGE042
N =1,2, … N for the sampling interval of the device basis information. Thus, basic information of different operation time lengths of a single device and corresponding actual output power can be collected.
The training output power prediction network comprises a power output quantity prediction coder and a full connection layer, and the specific training process is as follows: network input of basic information of electric power equipment, i.e.
Figure 970186DEST_PATH_IMAGE038
Acquiring a characteristic vector after the characteristic vector is coded by a power output quantity prediction coder, and sending the characteristic vector to a full connection layer, wherein network output is actual output power when the running time of equipment is t; t is a fixed duration, set by the implementer. Taking t as six hours as an example, the output quantity is an output power detection value of the device which continuously operates for six hours.
Further, in order to improve the prediction accuracy of the output power prediction network, that is, according to the basic information of the input power device, the error between the theoretical output power output by the prediction network and the actual output power is smaller, a prediction network with good prediction performance needs to be determined as a reference network for predicting the theoretical output power of each power device.
The specific determination process of the reference network is as follows:
(A) using a sample of basic information of a first batch of equipment as a first training set to predict the output power of the network W1Training and judging the network W1Whether the predicted value of each output quantity is accurate or not is determined according to the network W1Outputting the number of accurate power output predicted values, and determining the network W after the training of the first training set1Accuracy A of1。
Understandably, the number of samples in a single batch in the training process is set as
Figure 445030DEST_PATH_IMAGE043
Taking a single batch of samples as a first training set to perform network training, wherein a loss function of the network adopts a mean square error function, namely
Figure 928488DEST_PATH_IMAGE044
Wherein, in the step (A),
Figure 952814DEST_PATH_IMAGE045
is as follows
Figure 965988DEST_PATH_IMAGE046
The device is at
Figure 649167DEST_PATH_IMAGE047
The predicted value of the power output quantity at each moment,
Figure 871069DEST_PATH_IMAGE048
is as follows
Figure 199502DEST_PATH_IMAGE046
The device is at
Figure 405748DEST_PATH_IMAGE049
Obtaining the actual value of the output power (namely the detected value of the output power) at each moment, and obtaining the network W after the training of the first training set1Accuracy of
Figure 324026DEST_PATH_IMAGE050
。
The accuracy rate
Figure 268061DEST_PATH_IMAGE050
Is defined as: when it comes to
Figure 407925DEST_PATH_IMAGE046
The device is at
Figure DEST_PATH_IMAGE051
When the difference between the predicted value and the actual value of the output power at each moment satisfies the following setting conditions, the predicted value of the output power is judged to be accurate. The setting conditions are as follows:
Figure 772957DEST_PATH_IMAGE052
in the formula (I), the compound is shown in the specification,
Figure 958957DEST_PATH_IMAGE045
is as follows
Figure 197696DEST_PATH_IMAGE046
The device is at
Figure 519962DEST_PATH_IMAGE049
The predicted value of the power output quantity at each moment,
Figure 45840DEST_PATH_IMAGE048
is as follows
Figure 190907DEST_PATH_IMAGE046
The device is at
Figure 699249DEST_PATH_IMAGE049
The actual value of the output power at each time instant, M is the number of samples of a single training batch, N is the sampling period of the power output, N =1,2, …, N is an integer.
Then, according to the above setting conditions, the number of the predicted values which are accurate in statistical judgment is counted and judged
Figure 404250DEST_PATH_IMAGE053
I.e. the number of predicted values satisfying the above formula, the ratio of the number of accurate predicted values to the total number of predicted values
Figure 372468DEST_PATH_IMAGE054
Accuracy as a first training set
Figure 267480DEST_PATH_IMAGE050
(ii) a The higher the accuracy, the more uniform the distribution of the samples representing the selected training set, and the lower the fluctuation of the actual value of the corresponding output power of the selected training set, i.e. the closer to the predicted value of the power output.
(B) In the amount of
Figure 763534DEST_PATH_IMAGE043
Taking the second batch of samples as a second training set, and sending the second training set serving as a first verification set into the trained network W1And obtaining the verified network W of the first verification set1Accuracy of
Figure 545545DEST_PATH_IMAGE055
(ii) a Then, the second training set is used as a training sample, and the network W is predicted according to the output power of the initial parameters2Training is carried out, and a second training set is obtained to train the network W2Accuracy of
Figure 700976DEST_PATH_IMAGE056
。
(C) Obtaining a trained network W of a first training set1Accuracy of
Figure 224231DEST_PATH_IMAGE050
Network W after verification with first verification set1Accuracy of
Figure 124576DEST_PATH_IMAGE055
As a first accuracy deviation; obtaining a trained network W of a first training set1Accuracy of
Figure 278871DEST_PATH_IMAGE050
Network W after training with the second training set2Accuracy of
Figure 87427DEST_PATH_IMAGE056
As a second accuracy deviation. And judging whether the set conditions are met or not according to the first accuracy deviation and the second accuracy deviation, and determining the reference network. The judgment process is as follows:
(1) when the first accuracy deviation and the second accuracy deviation meet the first set condition, the network is retrained again by using the second training set to obtain the reference network W1'. Wherein, the formula of the first setting condition is as follows:
Figure 360626DEST_PATH_IMAGE010
Figure 374981DEST_PATH_IMAGE011
in the formula (I), the compound is shown in the specification,
Figure 990639DEST_PATH_IMAGE057
in order to be the first deviation of the accuracy,
Figure 465702DEST_PATH_IMAGE058
for the second accuracy, the first accuracy is biased,
Figure 294374DEST_PATH_IMAGE059
in order to preset the first accuracy threshold,
Figure 751900DEST_PATH_IMAGE060
a second accuracy threshold is preset.
(2) When the first accuracy deviation and the second accuracy deviation satisfy a second set condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1'. WhereinThe formula of the second setting condition is as follows:
Figure 281452DEST_PATH_IMAGE010
Figure 681078DEST_PATH_IMAGE014
in the formula (I), the compound is shown in the specification,
Figure 43051DEST_PATH_IMAGE057
in order to be the first deviation of the accuracy,
Figure 793068DEST_PATH_IMAGE058
for the second accuracy, the first accuracy is biased,
Figure 304690DEST_PATH_IMAGE059
in order to preset the first accuracy threshold,
Figure 284672DEST_PATH_IMAGE060
a second accuracy threshold is preset.
(3) When the first accuracy deviation and the second accuracy deviation meet a third set condition, combining the first training set and the second training set, and training the output power prediction network of the initial parameters by using the combined training set to obtain a reference network W1'. Wherein, the formula of the third setting condition is as follows:
Figure 101318DEST_PATH_IMAGE019
Figure 678143DEST_PATH_IMAGE011
in the formula (I), the compound is shown in the specification,
Figure 765310DEST_PATH_IMAGE057
in order to be the first deviation of the accuracy,
Figure 444422DEST_PATH_IMAGE058
for the second accuracy, the first accuracy is biased,
Figure 276593DEST_PATH_IMAGE059
in order to preset the first accuracy threshold,
Figure 128881DEST_PATH_IMAGE060
a second accuracy threshold is preset.
(4) When the first accuracy deviation and the second accuracy deviation satisfy a fourth setting condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1'. Wherein, the formula of the fourth setting condition is as follows:
Figure 70554DEST_PATH_IMAGE019
Figure 245534DEST_PATH_IMAGE014
in the formula (I), the compound is shown in the specification,
Figure 161406DEST_PATH_IMAGE057
in order to be the first deviation of the accuracy,
Figure 102205DEST_PATH_IMAGE058
for the second accuracy, the first accuracy is biased,
Figure 911766DEST_PATH_IMAGE059
in order to preset the first accuracy threshold,
Figure 156451DEST_PATH_IMAGE060
a second accuracy threshold is preset.
In order to facilitate understanding of the meaning of the above four judgment conditions, a condition specific judgment process is provided, which specifically comprises the following steps:
first, it is judged
Figure 202030DEST_PATH_IMAGE057
And a preset first accuracy threshold
Figure 661699DEST_PATH_IMAGE059
In a relation of (1), if
Figure 463783DEST_PATH_IMAGE061
The sample distribution of the first training set is considered to be consistent with the sample distribution of the second training set, and if the sample distribution of the first training set is consistent with the sample distribution of the second training set, the sample distribution of the second training set is considered to be consistent with the sample distribution of the first training set
Figure 734227DEST_PATH_IMAGE062
，
Figure 752255DEST_PATH_IMAGE060
If the actual value label fluctuation is the second accuracy threshold value, judging that the actual value label fluctuation corresponding to the first training set and the second training set is approximate, namely the fluctuation degrees of the actual values of the output power of the electric power equipment in the two training sets are similar, retraining the network trained by the first training set through the second training set, and taking the retrained network as a reference network; if it is
Figure 730795DEST_PATH_IMAGE063
And the fluctuation degrees of the actual values of the output power of the power equipment in the two training sets are different, and the trained network with higher accuracy is reserved as the reference network.
If it is
Figure 780528DEST_PATH_IMAGE064
The distribution of the first training set is not uniform, if it is
Figure 615016DEST_PATH_IMAGE063
And explaining that the fluctuation degrees of the actual values of the output power of the electric power equipment in the two training sets are similar, at this time, the sample distribution of the first training set and the sample distribution of the second training set are judged to have a complementary relation, so that the first training set and the second training set are combined to be used as the training setTraining the initial network, and taking the trained network as a reference network; if it is
Figure 868143DEST_PATH_IMAGE063
And keeping the trained network with higher accuracy as a reference network.
(D) Then the number is
Figure 666685DEST_PATH_IMAGE065
Taking the next batch of samples as a new second training set, and taking the reference network W determined for the first time1' the corresponding training set is used as a new first training set, the processing of the first training set and the second training set in the above steps (A) - (C) is repeated, and the reference network W is updated again1’。
(E) Repeating step (D) above until the last group is of quantity
Figure 291964DEST_PATH_IMAGE065
Taking the last batch of samples as a new second training set, and taking the reference network W determined last time1' the corresponding training set is used as a new first training set, the processing of the first training set and the second training set in the steps (A) - (C) is repeated, and the optimal reference network W is obtained by updating1' to this point.
Reference network W finally determined by the above steps1' the training data set corresponding to the network is distributed uniformly, the fluctuation of the true value is low, and the generalization performance of the network is better. Therefore, the optimal reference network is the optimal network, which is the basis for the device volatility evaluation module to be mentioned later to screen the primary stable power generation device.
The electric energy output quantity prediction module has the advantages that the accuracy of the network is utilized, the four set conditions are combined, some batches of samples participating in training are screened, the samples of each batch are traversed, the samples meeting the corresponding set conditions participate in network training, the purpose of continuously updating the reference network is achieved, the updated reference network can uniquely correspond to a certain batch or several batches of training samples each time, after all batches of samples are traversed, the finally updated reference network is used as the optimal network, and the corresponding samples participating in training can be determined according to the trained optimal network.
(III) a device volatility evaluation module for evaluating the volatility of the reference network W according to the finally determined reference network W1' i.e. an optimal network, selecting corresponding power equipment in a training sample participating in optimal network training as primary stable power generation equipment, i.e. power equipment with good stability and less volatility; and selecting corresponding power equipment in the sample which does not participate in the optimal network training as secondary stable power generation equipment, namely equipment with poor stability and large volatility.
That is, by continuously searching for a more optimized reference network process in the electric energy output quantity prediction module, samples participating in training are continuously searched and expanded in a step length manner of batch samples, and the training samples corresponding to the finally determined reference network are a set of a plurality of batch samples (i.e., a set formed by selected samples) screened from all batches of samples (basic information of all to-be-selected electric power devices), and compared with the remaining electric power devices corresponding to the electric power devices not participating in network training, the electric power devices in the set have better stability and smaller volatility of output power, so that the classification of the electric power devices can be reasonably realized.
And the equipment control module is used for selecting a set number of primary power generation equipment from all the primary stable power generation equipment determined by the equipment volatility evaluation module according to the power requirement and a preset priority input sequence for each power generation equipment, carrying out grid-connected input control, and taking the rest primary stable power generation equipment and the rest secondary stable power generation equipment as standby equipment.
It can be understood that, assuming that a total of 10 primary stable power generation devices are determined at present, on the basis of presetting a priority input sequence for each power generation device in advance, and according to the power demand condition in the power grid, it is determined that only 2 power generation devices with set rated power are needed, then 2 of the 10 primary stable power generation devices can be selected to be operated according to the priority input sequence and the power model of the power generation device. The invention is concerned with recommending the primary stable power generation equipment with higher priority level under the condition of limited quantity of the input operation equipment, and can ensure that the operation environment of the whole power grid is safer after the primary stable power generation equipment is selected for use than after the secondary stable power generation equipment is selected for use.
Preferably, in the system, the power equipment used for classification is solar household power generation equipment, equipment meeting the power grid requirement can be selected, meanwhile, stable power grid equipment is guaranteed to be connected to the power grid, and rapid identification of the power equipment which is beneficial to stable operation of the power grid is achieved. As another embodiment, the power plant for classification may be a large power plant, such as a photovoltaic power plant, a wind power plant, or the like.
Example 2:
the embodiment provides an intelligent energy management method for power grid power equipment, as shown in fig. 2, including the following steps:
s1, acquiring basic information of the power equipment, wherein the basic information comprises equipment aging degree, equipment model and equipment running time; the actual output power of the equipment when the running time is t is also obtained;
s2, randomly selecting the basic information of a set number of electric power equipment in batches from the basic information of all the electric power equipment, using the basic information of all the electric power equipment in each batch as training samples, respectively using the training samples in each batch as input, using the actual output power of the equipment in the training samples in each batch when the operation time is t as output, respectively and independently training an output power prediction network, wherein the output power prediction network adopts a neural network, outputting the accuracy of each output power prediction network after training, and selecting a reference network according to the accuracy of each output power prediction network as an optimal network for predicting the output power of the electric power equipment; the basic information of the power equipment comprises equipment aging degree
Figure 28844DEST_PATH_IMAGE066
Type of device
Figure 304622DEST_PATH_IMAGE067
Length of operation of the apparatus
Figure 394938DEST_PATH_IMAGE068
；
S3, selecting corresponding power equipment in the training samples participating in the optimal network training as primary stable power generation equipment; selecting power equipment which does not participate in the optimal network training as secondary stable power generation equipment;
and S4, selecting the first-stage stable power generation equipment to carry out grid-connected input control according to the power requirement and the preset priority input sequence for each power generation equipment in all the first-stage stable power generation equipment, and taking the rest first-stage stable power generation equipment and the rest second-stage stable power generation equipment as standby equipment.
The intelligent management method for device energy in this embodiment corresponds to the management method in the intelligent management system in embodiment 1, and the specific implementation process refers to the relevant records in embodiment 1, which is not described in detail in this embodiment.
It should be noted that: the above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (10)

1. An intelligent management system for power grid power equipment energy, the system comprising:
the data acquisition module is used for acquiring basic information of the power equipment, wherein the basic information comprises equipment aging degree, equipment model and equipment running time; the device is also used for acquiring the actual output power of the equipment when the running time is t;
an electric energy output quantity prediction module used for randomly selecting the basic information of a set number of electric equipment in batches from the basic information of all the electric equipment, wherein the basic information of each batch of electric equipment is used asTraining samples, namely respectively taking the training samples of each batch as input, taking actual output power of equipment in the training samples of each batch when the running time is t as output, respectively and independently training an output power prediction network, wherein the output power prediction network adopts a neural network, outputting the accuracy of each output power prediction network after training, and selecting a reference network according to the accuracy of each output power prediction network to be used as an optimal network for predicting the output power of the electric power equipment; the basic information of the power equipment is specifically equipment aging degree
Figure DEST_PATH_IMAGE002
Type of device
Figure DEST_PATH_IMAGE004
Length of operation of the apparatus
Figure DEST_PATH_IMAGE006
；
The equipment volatility evaluation module is used for selecting corresponding electric equipment in the training samples participating in the training of the optimal network as primary stable power generation equipment; selecting power equipment which does not participate in the optimal network training as secondary stable power generation equipment;
and the equipment control module is used for selecting the primary stable power generation equipment to carry out grid-connected input control according to the power requirement and a preset priority input sequence for each power generation equipment in all the primary stable power generation equipment determined by the equipment volatility evaluation module, and using the rest primary stable power generation equipment and the rest secondary stable power generation equipment as standby equipment.
2. A grid power equipment energy intelligent management system according to claim 1, wherein the neural network is constructed by the following specific process:
collecting training samples, basic information of the power equipment to be screened, wherein the basic information is expressed as
Figure DEST_PATH_IMAGE008
Wherein, in the step (A),
Figure DEST_PATH_IMAGE010
is as follows
Figure 83462DEST_PATH_IMAGE010
The device identification of the individual power device,
Figure 797340DEST_PATH_IMAGE008
three symbols in the sequence are
Figure 373815DEST_PATH_IMAGE010
Equipment aging degree, equipment model and equipment running time of each power equipment;
training output power prediction network, including power output quantity prediction encoder and full connection layer, network inputs basic information of electric power equipment, i.e.
Figure 565762DEST_PATH_IMAGE008
Obtaining the characteristic vector after the coding by the power output quantity prediction coder, sending the characteristic vector into a full connection layer, wherein the running time of the network output equipment is as long astAnd the actual value of the output power of the hour is obtained through detection.
3. The grid power equipment energy intelligent management system according to claim 1 or 2, wherein the reference network is selected according to the accuracy of each output power prediction network, and the selection as the optimal network for predicting the output power of the power equipment comprises:
A) taking a sample of basic information of the first batch of electrical equipment as a first training set, training an output power prediction network W1, judging whether each output quantity predicted value output by a network W1 is accurate, and determining the accuracy A1 of the network W1 after the training of the first training set according to the number of accurate power output predicted values output by the network W1;
B) in the amount of
Figure DEST_PATH_IMAGE012
Taking a second batch of samples as a second training set, and sending the second training set serving as a first verification set into the trained network W1And obtaining the verified network W of the first verification set1Accuracy of
Figure DEST_PATH_IMAGE014
(ii) a Then, the second training set is used as a training sample, and the network W is predicted according to the output power of the initial parameters2Training is carried out, and a second training set is obtained to train the network W2Accuracy of
Figure DEST_PATH_IMAGE016
；
C) Obtaining a trained network W of a first training set1Accuracy of
Figure DEST_PATH_IMAGE018
Network W after verification with first verification set1Accuracy of
Figure 163490DEST_PATH_IMAGE014
As a first accuracy deviation; obtaining a trained network W of a first training set1Accuracy of
Figure 782690DEST_PATH_IMAGE018
Network W after training with the second training set2Accuracy of
Figure 112040DEST_PATH_IMAGE016
As a second accuracy deviation; judging whether a set condition is met or not according to the first accuracy deviation and the second accuracy deviation, and determining a reference network;
D) then the number is
Figure 107678DEST_PATH_IMAGE012
The next batch of samplesFor a new second training set, the reference network W determined for the first time above is used1' the corresponding training set is used as a new first training set, the processing of the first training set and the second training set in the above steps A) -C) is repeated, and the reference network W is updated again1’；
E) Repeating step D) above until the last group is of quantity
Figure 104453DEST_PATH_IMAGE012
Taking the last batch of samples as a new second training set, and taking the reference network W determined last time1Taking the corresponding training set as a new first training set, repeating the processing of the first training set and the second training set in the steps A) to C), and updating to obtain the optimal reference network W1' to this point.
4. The system according to claim 3, wherein in step C), the determination of whether the set condition is satisfied is performed according to the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is:
when the first accuracy deviation and the second accuracy deviation meet the first set condition, the network W1 trained by the first training set is retrained again through the second training set to obtain a reference network W1', the formula of the first setting condition is as follows:
Figure DEST_PATH_IMAGE020
Figure DEST_PATH_IMAGE022
in the formula (I), the compound is shown in the specification,
Figure DEST_PATH_IMAGE024
in order to be the first deviation of the accuracy,
Figure DEST_PATH_IMAGE026
for the purpose of the second accuracy deviation,
Figure DEST_PATH_IMAGE028
in order to preset the first accuracy threshold,
Figure DEST_PATH_IMAGE030
a second accuracy threshold is preset.
5. The system according to claim 3, wherein in step C), the determination of whether the set condition is satisfied is performed according to the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is:
when the first accuracy deviation and the second accuracy deviation satisfy a second set condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1', the formula of the second setting condition is as follows:
Figure DEST_PATH_IMAGE020A
Figure DEST_PATH_IMAGE032
in the formula (I), the compound is shown in the specification,
Figure 989494DEST_PATH_IMAGE024
in order to be the first deviation of the accuracy,
Figure 809070DEST_PATH_IMAGE026
for the purpose of the second accuracy deviation,
Figure 342820DEST_PATH_IMAGE028
is preset toA threshold value of the accuracy rate is set,
Figure 662943DEST_PATH_IMAGE030
a second accuracy threshold is preset.
6. The system according to claim 3, wherein in step C), the determination of whether the set condition is satisfied is performed according to the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is:
when the first accuracy deviation and the second accuracy deviation meet a third set condition, combining the first training set and the second training set, and training the output power prediction network of the initial parameters by using the combined training set to obtain a reference network W1', the formula of the third setting condition is as follows:
Figure DEST_PATH_IMAGE034
Figure DEST_PATH_IMAGE022A
in the formula (I), the compound is shown in the specification,
Figure 342055DEST_PATH_IMAGE024
in order to be the first deviation of the accuracy,
Figure 645997DEST_PATH_IMAGE026
for the purpose of the second accuracy deviation,
Figure 240228DEST_PATH_IMAGE028
in order to preset the first accuracy threshold,
Figure 680436DEST_PATH_IMAGE030
a second accuracy threshold is preset.
7. The system according to claim 3, wherein in step C), the determination of whether the set condition is satisfied is performed according to the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is:
when the first accuracy deviation and the second accuracy deviation satisfy a fourth setting condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1', the formula of the fourth setting condition is as follows:
Figure DEST_PATH_IMAGE034A
Figure DEST_PATH_IMAGE032A
in the formula (I), the compound is shown in the specification,
Figure 733712DEST_PATH_IMAGE024
in order to be the first deviation of the accuracy,
Figure 259371DEST_PATH_IMAGE026
for the purpose of the second accuracy deviation,
Figure 400502DEST_PATH_IMAGE028
in order to preset the first accuracy threshold,
Figure 695218DEST_PATH_IMAGE030
a second accuracy threshold is preset.
8. An intelligent energy management method for power grid power equipment is characterized by comprising the following steps:
s1, acquiring basic information of the power equipment, wherein the basic information comprises equipment aging degree, equipment model and equipment running time; the actual output power of the equipment when the running time is t is also obtained;
s2, randomly selecting the basic information of a set number of electric power equipment in batches from the basic information of all the electric power equipment, using the basic information of all the electric power equipment in each batch as training samples, respectively using the training samples in each batch as input, using the actual output power of the equipment in the training samples in each batch when the operation time is t as output, respectively and independently training an output power prediction network, wherein the output power prediction network adopts a neural network, outputting the accuracy of each output power prediction network after training, and selecting a reference network according to the accuracy of each output power prediction network as an optimal network for predicting the output power of the electric power equipment; the basic information of the power equipment is specifically equipment aging degree
Figure 532111DEST_PATH_IMAGE002
Type of device
Figure 545066DEST_PATH_IMAGE004
Length of operation of the apparatus
Figure 224309DEST_PATH_IMAGE006
；
S3, selecting corresponding power equipment in the training samples participating in the optimal network training as primary stable power generation equipment; selecting power equipment which does not participate in the optimal network training as secondary stable power generation equipment; and S4, selecting the first-stage stable power generation equipment to carry out grid-connected input control according to the power requirement and the preset priority input sequence for each power generation equipment in all the first-stage stable power generation equipment, and taking the rest first-stage stable power generation equipment and the rest second-stage stable power generation equipment as standby equipment.
9. The grid power equipment energy intelligent management method according to claim 8, wherein in step S2, an output power prediction network with optimal prediction performance is determined as a reference network, and the specific determination process of the reference network is as follows:
A) taking a sample of basic information of the first batch of electrical equipment as a first training set, training an output power prediction network W1, judging whether each output quantity predicted value output by a network W1 is accurate, and determining the accuracy A1 of the network W1 after the training of the first training set according to the number of accurate power output predicted values output by the network W1;
B) in the amount of
Figure 373531DEST_PATH_IMAGE012
Taking a second batch of samples as a second training set, and sending the second training set serving as a first verification set into the trained network W1And obtaining the verified network W of the first verification set1Accuracy of
Figure 112817DEST_PATH_IMAGE014
(ii) a Then, the second training set is used as a training sample, and the network W is predicted according to the output power of the initial parameters2Training is carried out, and a second training set is obtained to train the network W2Accuracy of
Figure 347489DEST_PATH_IMAGE016
；
C) Obtaining a trained network W of a first training set1Accuracy of
Figure 830423DEST_PATH_IMAGE018
Network W after verification with first verification set1Accuracy of
Figure 365310DEST_PATH_IMAGE014
As a first accuracy deviation; obtaining a trained network W of a first training set1Accuracy of
Figure 541076DEST_PATH_IMAGE018
Network W after training with the second training set2Accuracy of
Figure 531553DEST_PATH_IMAGE016
As a second accuracy deviation; judging whether a set condition is met or not according to the first accuracy deviation and the second accuracy deviation, and determining a reference network;
D) then the number is
Figure 552599DEST_PATH_IMAGE012
Taking the next batch of samples as a new second training set, and taking the reference network W determined for the first time1' the corresponding training set is used as a new first training set, the processing of the first training set and the second training set in the above steps A) -C) is repeated, and the reference network W is updated again1’；
E) Repeating step D) above until the last group is of quantity
Figure 676413DEST_PATH_IMAGE012
Taking the last batch of samples as a new second training set, and taking the reference network W determined last time1Taking the corresponding training set as a new first training set, repeating the processing of the first training set and the second training set in the steps A) to C), and updating to obtain the optimal reference network W1' to this point.
10. The grid power equipment energy intelligent management method according to claim 9, wherein in step C), whether the set condition is met is judged according to the magnitude of the first accuracy deviation and the second accuracy deviation, and the process of determining the reference network is as follows:
when the first accuracy deviation and the second accuracy deviation meet the first set condition, the network W1 trained by the first training set is retrained again through the second training set to obtain a reference network W1', the formula of the first setting condition is as follows:
Figure DEST_PATH_IMAGE020AA
Figure DEST_PATH_IMAGE022AA
in the formula (I), the compound is shown in the specification,
Figure 210031DEST_PATH_IMAGE024
in order to be the first deviation of the accuracy,
Figure 419296DEST_PATH_IMAGE026
for the purpose of the second accuracy deviation,
Figure 509611DEST_PATH_IMAGE028
in order to preset the first accuracy threshold,
Figure 490861DEST_PATH_IMAGE030
a second accuracy threshold is preset;
when the first accuracy deviation and the second accuracy deviation satisfy a second set condition, the network W is compared1And W2The accuracy of (2) selecting the network W with higher accuracy1Or W2As a reference network W1', the formula of the second setting condition is as follows:
Figure DEST_PATH_IMAGE020AAA
Figure DEST_PATH_IMAGE032AA
in the formula (I), the compound is shown in the specification,
Figure 664223DEST_PATH_IMAGE024
in order to be the first deviation of the accuracy,
Figure 360783DEST_PATH_IMAGE026
for the purpose of the second accuracy deviation,
Figure 723631DEST_PATH_IMAGE028
in order to preset the first accuracy threshold,
Figure 87617DEST_PATH_IMAGE030
a second accuracy threshold is preset;
when the first accuracy deviation and the second accuracy deviation meet a third set condition, combining the first training set and the second training set, and training the output power prediction network of the initial parameters by using the combined training set to obtain a reference network W1', the formula of the third setting condition is as follows:
Figure DEST_PATH_IMAGE034AA
Figure DEST_PATH_IMAGE022AAA
in the formula (I), the compound is shown in the specification,
Figure 977686DEST_PATH_IMAGE024
in order to be the first deviation of the accuracy,
Figure 692701DEST_PATH_IMAGE026
for the purpose of the second accuracy deviation,
Figure 859240DEST_PATH_IMAGE028
in order to preset the first accuracy threshold,
Figure 812153DEST_PATH_IMAGE030
a second accuracy threshold is preset;
when the first accuracy deviation and the second accuracy deviation satisfy a fourth setting condition, the network W is compared1And W2The accuracy of the process is high and low,selecting a more accurate network W1Or W2As a reference network W1', the formula of the fourth setting condition is as follows:
Figure DEST_PATH_IMAGE034AAA
Figure DEST_PATH_IMAGE032AAA
in the formula (I), the compound is shown in the specification,
Figure 126984DEST_PATH_IMAGE024
in order to be the first deviation of the accuracy,
Figure 798137DEST_PATH_IMAGE026
for the purpose of the second accuracy deviation,
Figure 33946DEST_PATH_IMAGE028
in order to preset the first accuracy threshold,
Figure 106944DEST_PATH_IMAGE030
a second accuracy threshold is preset.
CN202110921796.4A 2021-08-12 2021-08-12 Intelligent energy management system and method for power grid power equipment Active CN113379005B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110921796.4A CN113379005B (en) 2021-08-12 2021-08-12 Intelligent energy management system and method for power grid power equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110921796.4A CN113379005B (en) 2021-08-12 2021-08-12 Intelligent energy management system and method for power grid power equipment

Publications (2)

Publication Number Publication Date
CN113379005A CN113379005A (en) 2021-09-10
CN113379005B true CN113379005B (en) 2021-10-29

Family

ID=77576785

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110921796.4A Active CN113379005B (en) 2021-08-12 2021-08-12 Intelligent energy management system and method for power grid power equipment

Country Status (1)

Country Link
CN (1) CN113379005B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115022381B (en) * 2022-08-08 2022-11-18 广东邦盛新能源科技发展有限公司 Intelligent networking method of photovoltaic panel data acquisition equipment
CN115632487B (en) * 2022-12-16 2023-03-14 北京国网电力技术有限公司 Smart park electric power coordination control method and system

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105117591A (en) * 2015-08-15 2015-12-02 国家电网公司 Power device running status analysis method
CN106679847A (en) * 2016-12-14 2017-05-17 云南电网有限责任公司电力科学研究院 Electric power equipment fault diagnosing method and apparatus

Family Cites Families (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5469528A (en) * 1992-09-16 1995-11-21 Syseca, Inc. Neural load disturbance analyzer
JP6384065B2 (en) * 2014-03-04 2018-09-05 日本電気株式会社 Information processing apparatus, learning method, and program
CN104121949A (en) * 2014-08-18 2014-10-29 中国船舶重工集团公司第七一二研究所 Condition monitoring method of ship electric propulsion system
CN104535865A (en) * 2014-12-30 2015-04-22 西安工程大学 Comprehensive diagnosing method for operation troubles of power transformer based on multiple parameters
US20190115753A1 (en) * 2017-10-13 2019-04-18 Schneider Electric It Corporation Method for automatic cloud control of energy storage systems
CN108229827A (en) * 2018-01-04 2018-06-29 南京航空航天大学 A kind of power equipment quality problems modeling and analysis methods
CN109558979A (en) * 2018-11-29 2019-04-02 广东电网有限责任公司 Power equipments defect prediction technique and device
CN109615131A (en) * 2018-12-07 2019-04-12 国能日新科技股份有限公司 Photovoltaic power prediction technique and device
CN109376972B (en) * 2019-01-01 2021-07-27 中南大学 Smart grid short-term power load prediction method based on block clustering
CN110175386B (en) * 2019-05-21 2022-11-25 陕西科技大学 Method for predicting temperature of electrical equipment of transformer substation
CN110334948B (en) * 2019-07-05 2023-04-07 上海交通大学 Power equipment partial discharge severity evaluation method and system based on characteristic quantity prediction
CN111009893B (en) * 2019-11-08 2022-07-01 广西电网有限责任公司电力科学研究院 Household user short-term load prediction method based on load decomposition technology
CN112348052A (en) * 2020-09-30 2021-02-09 国网山西省电力公司大同供电公司 Power transmission and transformation equipment abnormal sound source positioning method based on improved EfficientNet
CN112636328B (en) * 2020-11-25 2022-06-21 国网湖南省电力有限公司 Medium-voltage distribution network equipment load state identification method
CN112990130B (en) * 2021-04-27 2021-07-23 南京甄视智能科技有限公司 Data cleaning method and device for training sample and classifier

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105117591A (en) * 2015-08-15 2015-12-02 国家电网公司 Power device running status analysis method
CN106679847A (en) * 2016-12-14 2017-05-17 云南电网有限责任公司电力科学研究院 Electric power equipment fault diagnosing method and apparatus

Also Published As

Publication number Publication date
CN113379005A (en) 2021-09-10

Similar Documents

Publication Publication Date Title
CN113379005B (en) Intelligent energy management system and method for power grid power equipment
CN107274105B (en) Linear discriminant analysis-based multi-attribute decision tree power grid stability margin evaluation method
CN107395121B (en) Based on Grubbs test method and outlier detection photovoltaic array fault detection method
CN105740975B (en) A kind of equipment deficiency assessment and prediction technique based on data correlation relation
CN113298422B (en) Pollution source enterprise illegal production monitoring method based on electricity consumption data
CN111369168B (en) Associated feature selection method suitable for multiple regulation and control operation scenes of power grid
CN113344346B (en) Power utilization abnormity detection method and system based on non-intrusive load decomposition
CN115425680B (en) Power prediction model construction and prediction method of multi-energy combined power generation system
CN106817708A (en) One kind patrols and examines cluster construction method and device
CN113904322A (en) Low-voltage distribution network topology generation method based on current and voltage
CN116108402A (en) Method, equipment and storage medium based on electric power multi-source heterogeneous data fusion analysis
CN115733258A (en) Control method of all-indoor intelligent substation system based on Internet of things technology
CN108694475B (en) Short-time-scale photovoltaic cell power generation capacity prediction method based on hybrid model
CN111061152A (en) Attack recognition method based on deep neural network and intelligent energy power control device
CN114936808A (en) Cloud-edge cooperative task management system and method for substation fault detection
CN112241812B (en) Topology identification method for low-voltage distribution network based on single-side optimization and genetic algorithm cooperation
CN114819362A (en) Power grid power load balancing method for wind-solar power system
CN112485690B (en) Energy storage battery health state identification method and system based on vibration signals
CN113418632A (en) Conceptual drift detection method for oil temperature prediction of oil-immersed transformer
CN115224684A (en) Intelligent power distribution network risk state identification method and system based on immune hazard theory
Li et al. Study on the extraction method of high-risk operation scenarios under high proportion wind power grid-connected power system
CN113537528B (en) Preprocessing method and system for state monitoring data of power transmission and transformation equipment
CN116596194B (en) Photovoltaic array running state dividing method, system and device
CN113837486B (en) RNN-RBM-based distribution network feeder long-term load prediction method
CN110297145B (en) Voltage sag detection method based on multi-user electric energy data deep analysis

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