CN111723908A - Real-time scheduling model of wind power-containing power system based on deep learning - Google Patents

Real-time scheduling model of wind power-containing power system based on deep learning Download PDF

Info

Publication number
CN111723908A
CN111723908A CN202010532253.9A CN202010532253A CN111723908A CN 111723908 A CN111723908 A CN 111723908A CN 202010532253 A CN202010532253 A CN 202010532253A CN 111723908 A CN111723908 A CN 111723908A
Authority
CN
China
Prior art keywords
neural network
deep neural
real
cost
scheduling
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
CN202010532253.9A
Other languages
Chinese (zh)
Inventor
陈新建
于杰
朱轶伦
王彬任
陈秋临
夏敏燕
高慧英
周洪青
洪骋怀
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.)
Taizhou Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Original Assignee
Taizhou Power Supply Co of State Grid Zhejiang 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 Taizhou Power Supply Co of State Grid Zhejiang Electric Power Co Ltd filed Critical Taizhou Power Supply Co of State Grid Zhejiang Electric Power Co Ltd
Priority to CN202010532253.9A priority Critical patent/CN111723908A/en
Publication of CN111723908A publication Critical patent/CN111723908A/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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06312—Adjustment or analysis of established resource schedule, e.g. resource or task levelling, or dynamic rescheduling
    • 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
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
    • G06Q10/06315—Needs-based resource requirements planning or analysis
    • 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
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J3/00—Circuit arrangements for ac mains or ac distribution networks
    • H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
    • H02J3/381—Dispersed generators
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
    • H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
    • H02J2300/20—The dispersed energy generation being of renewable origin
    • H02J2300/28—The renewable source being wind energy
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00—Energy generation through renewable energy sources
    • Y02E10/70—Wind energy
    • Y02E10/76—Power conversion electric or electronic aspects

Abstract

The invention discloses a deep learning-based real-time scheduling model of a wind power-containing power system, which adopts a deep neural network to obtain a scheduling strategy: firstly, considering environmental economy in a traditional optimization scheduling model, and calculating a scheduling strategy and cost as the output of a deep neural network; then, according to the characteristics of input and output data of power grid dispatching, different activation functions among layers of the deep neural network are designed, and wider output is captured; finally, an improved deep neural network parameter initialization method is provided, and convergence speed is improved. The real-time scheduling model fully considers the environmental economic influence of the wind power-containing power system, reduces the scheduling strategy acquisition time by using a data driving method, and ensures that the system acquires a scheduling plan in real time.

Description

Real-time scheduling model of wind power-containing power system based on deep learning
Technical Field
The invention relates to the technical field of power system correlation, in particular to a wind power-containing power system real-time scheduling model based on deep learning.
Background
With the rapid development of wind power, the large-scale wind power integration can effectively relieve the power utilization pressure. However, in a wind power system, how to coordinate the output of the traditional thermal power generating unit and reduce the consumption of coal resources and environmental pollution is an important subject in the current research.
In the current scheduling calculation method for the wind power system, a model method is adopted. However, the traditional optimal scheduling model of the wind power system has more constraint conditions, relatively longer calculation time and complex calculation; and (3) solving the dynamic scheduling model considering the wind energy uncertainty by adopting a genetic algorithm so as to easily converge to a local optimal solution. The above problems can not complete the real-time scheduling of the system according to the wind power change, and the actual operation requirements can not be met. Considering that a deep neural network is the most effective machine learning method and is widely applied to the problems of prediction and classification of a power system, a data driving method considering the time-space correlation characteristic of data is urgently needed to be applied to real-time scheduling optimization calculation to guide actual engineering.
Disclosure of Invention
The invention aims to solve the problems in the prior art, and provides a wind power-containing power system real-time scheduling model based on deep learning.
A wind power-containing electric power system real-time scheduling model based on deep learning comprises the following steps:
step 1, obtaining the generated energy of a wind generating set and a thermal generating set at the current moment according to a system topological structure;
step 2, substituting the data of the wind generating set and the thermal generating set at the current moment into the trained real-time scheduling model based on the deep neural network; the real-time scheduling model training process based on the deep neural network is as follows:
system topology information and generating capacity data at different moments are input as a real-time scheduling model, the minimum cost of coal resource consumption and environmental pollution of a traditional thermal power generating unit, standby of a wind generating unit and the like is considered as an optimization target, and corresponding scheduling decisions and cost are calculated and serve as output of the real-time scheduling model;
wherein, the optimization objective function can be expressed as:
Figure BDA0002534489720000021
in the formula (1)
Figure BDA0002534489720000022
Represents the electricity purchasing cost and C of the thermal generator setEFIs defined as the environmental pollution cost of the thermal generator set,
Figure BDA0002534489720000023
Representing the electricity purchasing cost and C of the wind generating setWRRepresenting the standby cost of the wind power generator set; wherein n isGIndicates the number of thermal power generation groups in the system, nWRepresenting the number of wind power generation groups in the system;
the system constraints are:
Figure BDA0002534489720000024
Figure BDA0002534489720000025
Figure BDA0002534489720000026
Figure BDA0002534489720000027
in the formula:
Figure BDA0002534489720000028
the initial output of the generator node is characterized,
Figure BDA0002534489720000029
characterizing the initial load of a node, Gij、BijRespectively representing the real part and the imaginary part of corresponding elements in the admittance matrix; u shapei、UjRepresenting the voltage amplitudes, theta, of nodes i, j, respectivelyij＝θi-θjRepresenting the voltage angle difference between the node i and the node j, i and j are the first end node and the tail end node of the line where the tangent point needs to be adjusted,
Figure BDA00025344897200000212
iU、 iθrespectively representing the voltage amplitude value and the upper limit value and the lower limit value of the phase angle of the node i;
Figure BDA00025344897200000210
ijPrespectively representing the upper limit value and the lower limit value of the power of the lines i-j; GiP、 GiQ、
Figure BDA00025344897200000211
respectively representing the upper limit and the lower limit of active power output and reactive power output of a generator node;
designing a deep neural network model activation function: the activation function of the last layer of the deep neural network is designed to be linear, the activation functions of the feedforward transfer functions of other layers are designed to be ReLU activation functions, and then the form of the feedforward transfer function can be expressed as follows:
Figure BDA0002534489720000031
wherein l represents the number of layers of the deep neural network, n is the total number of layers of the deep neural network,
Figure BDA0002534489720000036
is the transfer function of the l-th layer, xl-1For the input of the l layer, theta is an optimized parameter in the calculation process of the deep neural network, and W islAnd blA weight matrix and an offset vector between the l layer and the l-1 layer are obtained;
improving initialization parameters of the deep neural network model:
design of forward propagation initialization:
the activation function for top-layer forward propagation is as follows:
zi＝wiyi+bi，yi＝Ri(zi-1) (7)
suppose yi、ziAre independent of each other, provided with wi-1A symmetrical distribution around 0, bi-1＝0，zi-1Has a mean value of zero and has a symmetrical distribution; due to yiIs derived from the ReLU function, then:
Figure BDA0002534489720000032
the sufficient conditions for the above formula to initialize correctly are:
Figure BDA0002534489720000033
thus, it follows:
Figure BDA0002534489720000034
and (3) reverse propagation initialization design:
for the back propagation process, the corresponding partial derivatives are:
Figure BDA0002534489720000035
in the formula, superscript T represents matrix transposition; same forward propagation, assuming R'
Figure BDA0002534489720000037
Independently of each other, considering that the gradient is not a sufficient condition for being exponentially large or small, there are:
Figure BDA0002534489720000041
w is aligned and propagated in opposite directions by the formulas (10) and (12)iThe variance is obtained, and the balance needs to be obtained:
Figure BDA0002534489720000042
then, the weight parameter w of the deep neural network is initialized to zero mean gaussian distribution, the standard deviation thereof is only required to solve the square root of the formula (13), and the deviation b is initialized to 0;
completing the offline training of the deep neural network model by using the confirmed input data, output data, activation function and initial data;
step 3, obtaining a real-time scheduling strategy according to the calculation result of the deep neural network model:
and during on-line calculation, substituting the real-time data into the trained real-time scheduling model to obtain the scheduling plan at the current moment and corresponding cost.
Step 4, judging whether the scheduling result meets all constraint requirements of the system: if the current time is in line with the preset time, outputting a scheduling plan and corresponding cost; if not, returning to the recalculation.
The invention has the beneficial effects that:
(1) the method obtains the power system scheduling strategy based on the DNN model, and gives full play to the superiority of data driving. The DNN can be trained and learned according to mass historical data, only time is spent in offline training, a scheduling strategy can be given quickly in an online application stage, real-time scheduling problems can be effectively solved, a real-time scheduling scheme is given to a decision maker, and the safe and stable operation of the system is guaranteed more effectively;
(2) the deep neural network has strong learning ability, and can be suitable for other different working condition scenes by adding corresponding samples according to different topologies and application scenes for supplementary training, so that the method has strong applicability;
(3) the invention takes the power generation cost of the traditional thermal power generation, the environmental impact cost of a thermal power unit, the installation and operation cost of wind power operation, the wind power standby cost caused by wind power uncertainty and the like into consideration, ensures that the obtained scheduling strategy is more economical, takes the environmental cost and the environmental impact into more consideration, and accords with the aim of green development of the current society.
Drawings
FIG. 1 is a diagram of a real-time scheduling model based on a deep neural network.
FIG. 2 is a flow chart of real-time scheduling of a wind-powered electrical power system in view of environmental economics.
Detailed Description
The following detailed description of embodiments of the invention is intended to be illustrative, but not limiting, of the invention. In order to make the technical means, creation features, achievement purposes and effects of the invention easy to understand, the invention is further described by combining the following specific embodiments:
example 1
As shown in the first figure, the present invention comprises the following steps:
step 1, obtaining the generated energy of a wind generating set and a thermal generating set at the current moment according to a system topological structure;
step 2, substituting the data of the wind generating set and the thermal generating set at the current moment into the trained real-time scheduling model based on the deep neural network; the real-time scheduling model training process based on the deep neural network is as follows:
system topology information and generating capacity data at different moments are input as a real-time scheduling model, the minimum cost of coal resource consumption and environmental pollution of a traditional thermal power generating unit, standby of a wind generating unit and the like is considered as an optimization target, and corresponding scheduling decisions and cost are calculated and serve as output of the real-time scheduling model;
wherein, the optimization objective function can be expressed as:
Figure BDA0002534489720000051
in the formula (1)
Figure BDA0002534489720000052
Represents the electricity purchasing cost and C of the thermal generator setEFIs defined as the environmental pollution cost of the thermal generator set,
Figure BDA0002534489720000053
Representing the electricity purchasing cost and C of the wind generating setWRRepresenting the standby cost of the wind power generator set; wherein n isGIndicates the number of thermal power generation groups in the system, nWRepresenting the number of wind power generation groups in the system.
Wherein, the electricity purchasing cost of the thermal generator set in the formula (1)
Figure BDA0002534489720000054
Can be expressed as:
Figure BDA0002534489720000061
in the formula, ai、bi、ciRepresenting a characteristic coefficient of cost, P, of the thermal power generating unitGiThe power generation amount of the ith thermal power generator;
environmental pollution cost C of medium-sized thermal generator set in formula (1)EFCan be expressed as:
Figure BDA0002534489720000062
in the formula, muiTo representα cost coefficient of environmental pollution of thermal generator seti、βi、χi、i、iConverting the pollutants discharged in unit time of each unit into discharged nitride weight expression for the environmental pollution characteristic index of the ith thermal generator set;
electricity purchasing cost of wind generating set in formula (1)
Figure BDA0002534489720000063
Can be expressed as:
Figure BDA0002534489720000064
in the formula, ηWjRepresenting the cost coefficient of the wind generating set, ten thousand yuan/pu; pWjThe actual power generation amount of the jth wind generating set;
in formula (1), the standby cost C of the wind power electric generating setWRCan be expressed as:
Figure BDA0002534489720000065
in the formula, PWPjIndicating the planned power generation of the jth wind turbine generator system, ξWjRepresenting a wind power generation standby cost coefficient, ten thousand yuan/pu, and ξWjThe relationship with the amount of electricity generation is expressed as:
Figure BDA0002534489720000066
the system constraints are:
Figure BDA0002534489720000071
Figure BDA0002534489720000072
Figure BDA0002534489720000073
Figure BDA0002534489720000074
in the formula:
Figure BDA0002534489720000075
the initial output of the generator node is characterized,
Figure BDA0002534489720000076
characterizing the initial load of a node, Gij、BijRespectively representing the real part and the imaginary part of corresponding elements in the admittance matrix; u shapei、UjRepresenting the voltage amplitudes, theta, of nodes i, j, respectivelyij＝θi-θjRepresenting the voltage angle difference between the node i and the node j, i and j are the first end node and the tail end node of the line where the tangent point needs to be adjusted,
Figure BDA0002534489720000077
iU、 iθrespectively representing the voltage amplitude value and the upper limit value and the lower limit value of the phase angle of the node i;
Figure BDA0002534489720000078
ijPrespectively representing the upper limit value and the lower limit value of the power of the lines i-j; GiP、 GiQ、
Figure BDA0002534489720000079
respectively representing the upper limit and the lower limit of active power output and reactive power output of a generator node;
designing a deep neural network model activation function: the activation function of the last layer of the deep neural network is designed to be linear, the activation functions of the feedforward transfer functions of other layers are designed to be ReLU activation functions, and then the form of the feedforward transfer function can be expressed as follows:
Figure BDA00025344897200000710
wherein l represents the number of layers of the deep neural network, n is the total number of layers of the deep neural network,
Figure BDA00025344897200000711
is the transfer function of the l-th layer, xl-1For the input of the l layer, theta is an optimized parameter in the calculation process of the deep neural network, and W islAnd blA weight matrix and an offset vector between the l layer and the l-1 layer are obtained;
improving initialization parameters of the deep neural network model:
design of forward propagation initialization:
the activation function for top-layer forward propagation is as follows:
zi＝wiyi+bi，yi＝Ri(zi-1) (7)
suppose yi、ziAre independent of each other, provided with wi-1A symmetrical distribution around 0, bi-1＝0，zi-1Has a mean value of zero and has a symmetrical distribution; due to yiIs derived from the ReLU function, then:
Figure BDA0002534489720000081
the sufficient conditions for the above formula to initialize correctly are:
Figure BDA0002534489720000082
thus, it follows:
Figure BDA0002534489720000083
and (3) reverse propagation initialization design:
for the back propagation process, the corresponding partial derivatives are:
Figure BDA0002534489720000084
in the formula, superscript T represents matrix transposition; same forward propagation, assuming R'
Figure BDA0002534489720000085
Independently of each other, considering that the gradient is not a sufficient condition for being exponentially large or small, there are:
Figure BDA0002534489720000086
w is aligned and propagated in opposite directions by the formulas (10) and (12)iThe variance is obtained, and the balance needs to be obtained:
Figure BDA0002534489720000087
then, the weight parameter w of the deep neural network is initialized to zero mean gaussian distribution, the standard deviation thereof is only required to solve the square root of the formula (13), and the deviation b is initialized to 0;
completing the offline training of the deep neural network model by using the confirmed input data, output data, activation function and initial data;
step 3, obtaining a real-time scheduling strategy according to the calculation result of the deep neural network model:
and during on-line calculation, substituting the real-time data into the trained real-time scheduling model to obtain the scheduling plan at the current moment and corresponding cost.
Step 4, judging whether the scheduling result meets all constraint requirements of the system: if the current time is in line with the preset time, outputting a scheduling plan and corresponding cost; if not, returning to the recalculation.
The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can understand that the modifications or substitutions within the technical scope of the present invention are included in the scope of the present invention, and therefore, the scope of the present invention should be subject to the protection scope of the claims.

Claims (1)

1. A wind power-containing electric power system real-time scheduling model based on deep learning is characterized by comprising the following steps:
step 1, obtaining the generated energy of a wind generating set and a thermal generating set at the current moment according to a system topological structure;
step 2, substituting the data of the wind generating set and the thermal generating set at the current moment into the trained real-time scheduling model based on the deep neural network; the real-time scheduling model training process based on the deep neural network is as follows:
system topology information and generating capacity data at different moments are input as a real-time scheduling model, the minimum cost of coal resource consumption and environmental pollution of a traditional thermal power generating unit, standby of a wind generating unit and the like is considered as an optimization target, and corresponding scheduling decisions and cost are calculated and serve as output of the real-time scheduling model;
wherein, the optimization objective function can be expressed as:
Figure FDA0002534489710000011
in the formula (1)
Figure FDA0002534489710000012
Represents the electricity purchasing cost and C of the thermal generator setEFIs defined as the environmental pollution cost of the thermal generator set,
Figure FDA0002534489710000013
Representing the electricity purchasing cost and C of the wind generating setWRRepresenting the standby cost of the wind power generator set; wherein n isGIndicates the number of thermal power generation groups in the system, nWRepresenting the number of wind power generation groups in the system;
the system constraints are:
Figure FDA0002534489710000014
Figure FDA0002534489710000015
Figure FDA0002534489710000016
Figure FDA0002534489710000017
in the formula:
Figure FDA0002534489710000018
the initial output of the generator node is characterized,
Figure FDA0002534489710000019
characterizing the initial load of a node, Gij、BijRespectively representing the real part and the imaginary part of corresponding elements in the admittance matrix; u shapei、UjRepresenting the voltage amplitudes, theta, of nodes i, j, respectivelyij＝θi-θjRepresenting the voltage angle difference between the node i and the node j, i and j are the first end node and the tail end node of the line where the tangent point needs to be adjusted,
Figure FDA0002534489710000021
iU、 iθrespectively representing the voltage amplitude value and the upper limit value and the lower limit value of the phase angle of the node i;
Figure FDA0002534489710000022
ijPrespectively representing the upper limit value and the lower limit value of the power of the lines i-j; GiP、 GiQ、
Figure FDA0002534489710000023
respectively representing the upper limit and the lower limit of active power output and reactive power output of a generator node;
designing a deep neural network model activation function: the activation function of the last layer of the deep neural network is designed to be linear, the activation functions of the feedforward transfer functions of other layers are designed to be ReLU activation functions, and then the form of the feedforward transfer function can be expressed as follows:
Figure FDA0002534489710000024
wherein l represents the number of layers of the deep neural network, n is the total number of layers of the deep neural network,
Figure FDA0002534489710000025
is the transfer function of the l-th layer, xl-1For the input of the l layer, theta is an optimized parameter in the calculation process of the deep neural network, and W islAnd blA weight matrix and an offset vector between the l layer and the l-1 layer are obtained;
improving initialization parameters of the deep neural network model:
design of forward propagation initialization:
the activation function for top-layer forward propagation is as follows:
zi＝wiyi+bi，yi＝Ri(zi-1) (7)
suppose yi、ziAre independent of each other, provided with wi-1A symmetrical distribution around 0, bi-1＝0，zi-1Has a mean value of zero and has a symmetrical distribution; due to yiIs derived from the ReLU function, then:
Figure FDA0002534489710000026
the sufficient conditions for the above formula to initialize correctly are:
Figure FDA0002534489710000027
thus, it follows:
Figure FDA0002534489710000031
and (3) reverse propagation initialization design:
for the back propagation process, the corresponding partial derivatives are:
Figure FDA0002534489710000032
in the formula, superscript T represents matrix transposition; same forward propagation, assuming R'
Figure FDA0002534489710000033
Independently of each other, considering that the gradient is not a sufficient condition for being exponentially large or small, there are:
Figure FDA0002534489710000034
w is aligned and propagated in opposite directions by the formulas (10) and (12)iThe variance is obtained, and the balance needs to be obtained:
Figure FDA0002534489710000035
then, the weight parameter w of the deep neural network is initialized to zero mean gaussian distribution, the standard deviation thereof is only required to solve the square root of the formula (13), and the deviation b is initialized to 0;
completing the offline training of the deep neural network model by using the confirmed input data, output data, activation function and initial data;
step 3, obtaining a real-time scheduling strategy according to the calculation result of the deep neural network model:
and during on-line calculation, substituting the real-time data into the trained real-time scheduling model to obtain the scheduling plan at the current moment and corresponding cost.
Step 4, judging whether the scheduling result meets all constraint requirements of the system: if the current time is in line with the preset time, outputting a scheduling plan and corresponding cost; if not, returning to the recalculation.
CN202010532253.9A 2020-06-11 2020-06-11 Real-time scheduling model of wind power-containing power system based on deep learning Pending CN111723908A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010532253.9A CN111723908A (en) 2020-06-11 2020-06-11 Real-time scheduling model of wind power-containing power system based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010532253.9A CN111723908A (en) 2020-06-11 2020-06-11 Real-time scheduling model of wind power-containing power system based on deep learning

Publications (1)

Publication Number Publication Date
CN111723908A true CN111723908A (en) 2020-09-29

Family

ID=72566479

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010532253.9A Pending CN111723908A (en) 2020-06-11 2020-06-11 Real-time scheduling model of wind power-containing power system based on deep learning

Country Status (1)

Country Link
CN (1) CN111723908A (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170091615A1 (en) * 2015-09-28 2017-03-30 Siemens Aktiengesellschaft System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies
US20190012598A1 (en) * 2017-07-04 2019-01-10 Taiwan Power Company Power generation prediction system and a method thereof
CN109241630A (en) * 2018-09-11 2019-01-18 国网河北能源技术服务有限公司 The method for optimizing scheduling and device of electric system
CN111105090A (en) * 2019-12-18 2020-05-05 沈阳鼓风机集团自动控制系统工程有限公司 Distributed energy system optimal scheduling method and device based on intelligent algorithm

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170091615A1 (en) * 2015-09-28 2017-03-30 Siemens Aktiengesellschaft System and method for predicting power plant operational parameters utilizing artificial neural network deep learning methodologies
US20190012598A1 (en) * 2017-07-04 2019-01-10 Taiwan Power Company Power generation prediction system and a method thereof
CN109241630A (en) * 2018-09-11 2019-01-18 国网河北能源技术服务有限公司 The method for optimizing scheduling and device of electric system
CN111105090A (en) * 2019-12-18 2020-05-05 沈阳鼓风机集团自动控制系统工程有限公司 Distributed energy system optimal scheduling method and device based on intelligent algorithm

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
TAHIR, MUHAMMAD FAIZAN； ET AL: "《Optimal scheduling of electrical power in energy-deficient scenarios using artificial neural network and Bootstrap aggregating》", 《INTERNATIONAL JOURNAL OF ELECTRICAL POWER & ENERGY SYSTEMS》 *
杨迎，等: "《基于RBF神经网络的梯级电站优化调度规则研究》", 《水电能源科学》 *

Similar Documents

Publication Publication Date Title
CN104408260B (en) A kind of tidal current energy water turbine vane airfoil profile method for designing
CN108241782B (en) Mixed transient analysis method and mixed transient analysis system
CN108471143A (en) Micro-grid multi-energy method for optimizing scheduling based on positive and negative feedback particle cluster algorithm
Yang et al. DISTRIBUTED OPTIMAL DISPATCH OF VIRTUAL POWER PLANT BASED ON ELM TRANSFORMATION.
Amundarain et al. Neural rotational speed control for wave energy converters
CN107482618A (en) Electricity pricing method with the generating of wind fire and the power network of flexible load containing polymorphic type
CN107666155A (en) System Stochastic Stability Analysis method of providing multiple forms of energy to complement each other based on Markov model
Xi et al. A deep reinforcement learning algorithm for the power order optimization allocation of AGC in interconnected power grids
CN110601177A (en) Economic optimization method for micro-grid containing wind power and photovoltaic power generation
Aly Forecasting, modeling, and control of tidal currents electrical energy systems
CN103904641A (en) Method for controlling intelligent power generation of island micro grid based on correlated equilibrium reinforcement learning
CN108594658A (en) A kind of electric-gas coupled system maximum probability load margin Model for Multi-Objective Optimization and its method for solving
Zangeneh et al. A survey: Fuzzify parameters and membership function in electrical applications
CN112084680B (en) Energy internet optimization strategy method based on DQN algorithm
CN112214883B (en) Micro-energy-grid comprehensive energy system operation optimization method and system
CN106100002A (en) A kind of optimizing operation method of alternating current-direct current mixing microgrid
CN112990674A (en) Multi-target operation scheduling method for offshore floating wind power plant
Zhang et al. Intelligent control for large-scale variable speed variable pitch wind turbines
CN111723908A (en) Real-time scheduling model of wind power-containing power system based on deep learning
Wang et al. Analysis of Coordinated Operation of the Clean Energy System Based on the Multiobjective Optimization Model
CN112600202B (en) Method for calculating optimal power flow of power grid with controllable phase shifter considering randomness of new energy
CN114421468A (en) Primary frequency modulation capacity planning method considering wind power cluster shared energy storage joint participation
Simani et al. Fuzzy control techniques applied to wind turbine systems and hydroelectric plants
Oudah et al. Wind turbines control: Features and trends
Ouhaddach et al. New MPPT optimizer for wind turbine energy sources

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
WD01 Invention patent application deemed withdrawn after publication
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20200929