CN112398142B - Power grid frequency intelligent control method based on empirical mode decomposition - Google Patents

Power grid frequency intelligent control method based on empirical mode decomposition Download PDF

Info

Publication number
CN112398142B
CN112398142B CN202011152690.4A CN202011152690A CN112398142B CN 112398142 B CN112398142 B CN 112398142B CN 202011152690 A CN202011152690 A CN 202011152690A CN 112398142 B CN112398142 B CN 112398142B
Authority
CN
China
Prior art keywords
control
frequency
signal
neural network
term memory
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
CN202011152690.4A
Other languages
Chinese (zh)
Other versions
CN112398142A (en
Inventor
殷林飞
吴云智
孙志响
高放
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.)
Guangxi University
Original Assignee
Guangxi University
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 Guangxi University filed Critical Guangxi University
Priority to CN202011152690.4A priority Critical patent/CN112398142B/en
Publication of CN112398142A publication Critical patent/CN112398142A/en
Application granted granted Critical
Publication of CN112398142B publication Critical patent/CN112398142B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • H02J3/24—Arrangements for preventing or reducing oscillations of power in networks
    • H02J3/241—The oscillation concerning frequency
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield 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
    • G06N3/04—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
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • 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
    • 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
    • H02J3/28—Arrangements for balancing of the load in a network by storage of energy
    • 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
    • H02J3/46—Controlling of the sharing of output between the generators, converters, or transformers
    • H02J3/466—Scheduling the operation of the generators, e.g. connecting or disconnecting generators to meet a given demand
    • 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
    • H02J2203/10—Power transmission or distribution systems management focussing at grid-level, e.g. load flow analysis, node profile computation, meshed network optimisation, active network management or spinning reserve management
    • 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]
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Power Engineering (AREA)
  • Control Of Eletrric Generators (AREA)
  • Supply And Distribution Of Alternating Current (AREA)
  • Feedback Control In General (AREA)

Abstract

The invention provides an intelligent power grid frequency control method based on empirical mode decomposition, which comprises two modules, wherein the first empirical mode decomposition module is based on a long-short term memory artificial neural network, the second empirical mode decomposition module is a control module comprising a reinforcement learning method and a deep neural network, and a double-layer step-by-step frequency control framework is provided. Firstly, the decomposition module decomposes the power grid frequency signal into a plurality of modal components in real time, and the regularity of the signal is highlighted. And secondly, the control module adjusts the output of the generator set according to the high-frequency signal and the low-frequency signal of the modal component, and the stability of the power grid frequency is maintained. Finally, in a multi-unit grid, the decomposition module and the control module can be effectively applied to a double-layer stepped frequency control framework to obtain optimal control performance.

Description

Power grid frequency intelligent control method based on empirical mode decomposition
Technical Field
The invention belongs to the field of power grid frequency control, and relates to a frequency control method based on an artificial intelligence technology, which is suitable for secondary frequency modulation of a power system.
Background
The frequency stability is one of important indexes for measuring the power quality of a power grid, and the frequency adjustment can be generally divided into primary frequency modulation, secondary frequency modulation and tertiary frequency modulation according to the frequency modulation period. The secondary frequency modulation is also called automatic power generation control, the power grid frequency is maintained in a stable range (50 +/-0.2 Hz in China) by adjusting the output of a generator set in real time, and the main indexes of the control performance are evaluated according to frequency deviation and regional control errors. The proportional-integral-derivative control method is a main traditional control method, but the method needs to be configured with proper parameters to achieve a satisfactory control effect. The intelligent method can optimize the parameters of the proportional-integral-derivative method and enhance the control performance, such as a particle swarm optimization method, a moth fire suppression optimization method or a genetic method. However, after the system parameters are changed, the parameters of the proportional-integral-derivative method need to be reset. In addition, the fixed parameter pid-i-d method has poor control performance in a microgrid with new energy and energy storage devices. The reinforcement learning method can realize the interaction between the controlled object and the system state, continuously updates the output action on line according to the change of the system state, and reinforces the Q value matrix, so that the more the output action quantity is, the more accurate the control is theoretically, but the dimension disaster is easily caused by the increase of the action quantity along with the complication of the system. Compared with an artificial neural network, the deep neural network has more hidden layers, stronger representation capability and better fitting capability, can extract the main characteristic relation between input data and output data, and reduces the complexity of the system. The proportional-integral-derivative method and the reinforcement learning method are generally embedded into a single-layer frequency control framework, and the method applied in the framework needs to analyze a frequency deviation signal and a issued power instruction simultaneously, so that the running performance of the method is reduced. Therefore, the conventional single-layer frequency control framework cannot meet the requirement of the power system on the high efficiency of frequency control. The invention brings the control performance standard index specified by the North American electric power reliability committee into double-layer step-by-step frequency control, and aims to better judge the stability of the frequency of the electric power system.
Empirical mode decomposition is a signal processing method of a time-frequency domain, can solve the problem that a basis function in a complex signal has no self-adaptability, and is commonly used for decomposition of irregular and nonlinear signals. The long-short term memory artificial neural network is a neural network capable of learning and memorizing long sequence data characteristics and is mainly used for wind power prediction, load prediction and fault diagnosis in a power system.
Disclosure of Invention
In order to cope with the impact and interference of new energy power generation equipment and energy storage equipment on the power grid frequency, the invention provides an intelligent power grid frequency control method based on empirical mode decomposition. The method provides a long-short term memory artificial neural network capable of carrying out real-time empirical mode decomposition, and frequency deviation signals of a power system are decomposed into a plurality of eigenmode function components and residual components, namely a decomposition module. And then, the reinforcement learning method and the deep neural network control the output of the generator set according to the eigenmode function and the residual component, namely a control module. The decomposition module and the control module are applied to a double-layer step-by-step frequency control framework to realize the optimal regulation and control of the frequency of the power system. The method comprises the following using processes: a double-layer step-by-step frequency control framework is provided, and a control performance standard index is brought into double-layer step-by-step frequency control; the decomposition module provided by the invention is used for empirical mode decomposition of the power grid frequency signal; the control module provided by the invention is used for adjusting the output of the generator set; the stepped decomposition module and the control module are applied to a stepped frequency control framework having a two-layer structure.
The method comprises the following specific steps:
(1) historical data M (t) of the frequency deviation signals of the power grid are collected and stored, and empirical mode decomposition is carried out on the data according to the following steps:
(1-1) identifying and extracting a maximum value point and a minimum value point of frequency deviation historical data M (t), and fitting an upper envelope line e1(t) and lower envelope e2(t) for e1(t) and e2(t) averaging the data of each point, and drawing a mean envelope e (t);
(1-2) subtracting the historical frequency deviation signal data M (t) from the mean envelope e (t) to obtain an intermediate signal H (t), and if the H (t) meets the constraint condition of the eigenmode function component, forming a first eigenmode function component IMF 1; if not, making M (t) H (t), and then repeating the steps (1-1) and (1-2);
(1-3) subtracting the obtained first eigenmode function component IMF1 from the original frequency deviation signal data M (t) to obtain a residual component R (t), making M (t) equal to R (t), and repeating the steps of (1-1), (1-2) and (1-3) until the number of eigenmode function components reaches a set number N or the residual component R (t) can not be decomposed any more;
(1-4) after completion of the decomposition, the following formula (1) is satisfied:
Figure BDA0002741681090000021
wherein m (t) represents a real-time grid frequency deviation signal; IMFiIs the ith eigenmode function component; n is the number of eigenmode function components; r is the residual component;
(2) establishing a long-short term memory artificial neural network:
the long-short term memory artificial neural network is formed by connecting a plurality of neurons. The forgetting gate can selectively forget the information transmitted by the previous neuron, and the forgetting degree is determined according to the following formula:
Zf＝σ(Wf·[ht-1,xt]+bf) (2)
in the formula, ZfA state indicating forgetting to gate; wfA weight matrix representing forgotten gate control input values; h is a total oft-1Is the short-term memory state value of the last neuron; x is the number oftIs the input signal of the current neuron; bfIs a forgetting gated bias signal; σ () is a Sigmoid function;
then selects the gated primary pair input signal xtSelectively memorizing:
Zi＝σ(Wi·[ht-1,xt]+bi) (3)
in the formula, ZiIndicating a state of selective gating; wiA weight matrix representing a selection gating input value; b is a mixture ofiIs a bias signal that selects gating;
the long-term memory state value of the current neuron is calculated by the following formula:
Figure BDA0002741681090000031
in the formula, CtRepresenting a long-term memory state value of a current neuron; ct-1Representing the long-term memory state value of the last neuron;
Figure BDA0002741681090000032
and representing the candidate long-term memory state value of the current neuron, wherein the expression is as follows:
Figure BDA0002741681090000033
in the formula (I), the compound is shown in the specification,
Figure BDA0002741681090000034
a weight matrix representing candidate input values;
Figure BDA0002741681090000035
a bias signal that is a candidate input value;
finally, the expression of the output gating is:
Zo＝σ(Wo·[ht-1,xt]+bo) (6)
in the formula, ZoIndicating the state of the output gate; woA weight matrix representing output gated input values; boIs an output gated bias signal;
when the input signal is xtThe output signal of the current neuron is calculated by the following formula:
ht＝Zo·tanh(Ct) (7)
in the formula, htIs the output signal of the current neuron;
(3) initializing a long-short term memory artificial neural network, taking all eigenmode function components and residual components obtained in the step (1) as output training data, taking historical data of a power grid frequency deviation signal as input training data, and performing off-line training on the long-short term memory artificial neural network;
(4) obtaining and storing a long-term and short-term memory artificial neural network with an empirical mode decomposition function to form a decomposition module;
(5) inputting a power grid real-time frequency deviation signal, and decomposing the signal in real time by using the long-short term memory artificial neural network in the step (4) to obtain a real-time eigenmode function component and a residual component;
(6) for IMF in real-time eigenmode function component1-IMF4Summing the 4 components to obtain a high-frequency signal S with frequency deviationH：
Figure BDA0002741681090000036
Then, the IMF in the real-time eigenmode function component is adjusted5-IMF9Summing the 5 components and the residual component r to obtain a low-frequency signal S with frequency deviationL：
Figure BDA0002741681090000037
(7) High frequency signal SHIntroducing a reinforcement learning method for optimization processing; low frequency signal SLThe afferent deep neural network carries out fitting processing;
(8) the reinforcement learning method and the deep neural network are jointly used as a control module and respectively send power instructions to the generator set;
(9) and (4) adjusting output of each generator set according to the power instruction, and obtaining a real-time CPS1 value according to the control performance standard.
Further, in the method of the present invention, the historical data m (t) in the step (1) is the grid frequency deviation data recorded every 1 second for 24 hours from the 0 th second in the previous day of the grid, and the dimension of m (t) is 86401 × 1.
Further, in the steps (1-4) of the method of the present invention, the number N of eigenmode function components is equal to 9.
Further, in the method of the present invention, the input feature dimension of the long short term memory artificial neural network in the step (3) is 1, the output feature dimension is 10, the hidden layer is 3 × 80, the learning rate of the off-line training is 0.008, the fading factor is 0.5, and the maximum iteration number is 500.
Furthermore, in the method, the decomposition module in the step (4) and the control module in the step (8) form a double-layer step-by-step frequency control framework, and a control performance standard index is incorporated into double-layer step-by-step frequency control to examine the control effect.
Further, in the step (9) of the method of the present invention, CPS1 is obtained by the following formula:
Figure BDA0002741681090000041
in the formula (I), the compound is shown in the specification,
Figure BDA0002741681090000042
is the average of the regional control errors over 1 minute;
Figure BDA0002741681090000043
is the average of the frequency deviations over 1 minute; b isiIs the frequency deviation coefficient (in 10MW/Hz) of the control region i; n istimeThe number of minutes within the statistical time; epsilon1The method comprises the following steps of (1) providing a control target of the mean square root of frequency deviation of an interconnected power grid per minute in one year; the closer CPS1 is to 200%, the better the control effect.
The invention has the advantages that: compared with the traditional proportional-integral-derivative method and the reinforcement learning method, the method disclosed by the invention can better adapt to the influence of new energy and energy storage equipment on the power grid frequency, and a better power grid frequency control effect is obtained.
Drawings
FIG. 1 is a flow chart of an embodiment of the method of the present invention.
Fig. 2 is a two-level step frequency control framework proposed by the method of the present invention.
Detailed Description
The invention provides an empirical mode decomposition-based power grid frequency intelligent control method, which is described in detail in combination with the accompanying drawings as follows:
FIG. 1 is a flow chart of an embodiment of the method of the present invention. In order to prove the superiority of the method, the invention method is further illustrated and explained by adopting an IEEE standard two-region load frequency control simulation embodiment comprising a wind power generation model, a photovoltaic power generation model and an energy storage device model for verification. The implementation steps of this example are as follows:
(1) recording frequency deviation signal data of one area in the two areas in the previous day, wherein the format of the data is that the frequency deviation is sampled every 1 second from 0:00 th to 24:00 th day, and recording power grid frequency deviation data M (t) in 24 hours of 86401 time nodes in total, and performing empirical mode decomposition on the data;
(2) establishing a long-short term memory artificial neural network;
(3) initializing a long-short term memory artificial neural network, wherein the input characteristic dimension of the long-short term memory artificial neural network is 1, the output characteristic dimension of the long-short term memory artificial neural network is 10, and the hidden layer is 3 x 80; taking all the eigenmode function components and residual components obtained in the step (1) as output training data, taking historical data of a power grid frequency deviation signal as input training data, and performing off-line training on the long-term and short-term memory artificial neural network, wherein the learning rate is 0.008, the fading factor is 0.5, and the maximum iteration number is 500;
(4) obtaining and storing a long-short term memory artificial neural network with an empirical mode decomposition function;
(5) inputting a power grid real-time frequency deviation signal, and decomposing the signal in real time by using the long-short term memory artificial neural network in the step (4) to obtain a real-time eigenmode function component and a residual component;
(6) for IMF in real-time eigenmode function component1-IMF4Summing the 4 components to obtain a high-frequency signal S with frequency deviationH：
Figure BDA0002741681090000051
Then, the IMF in the real-time eigenmode function component is adjusted5-IMF9Summing the 5 components and the residual component r to obtain a low-frequency signal S with frequency deviationL：
Figure BDA0002741681090000052
(7) High frequency signal SHTransmitting a reinforcement learning method for optimization processing; low frequency signal SLThe afferent deep neural network carries out fitting processing;
(8) the reinforcement learning method and the deep neural network respectively send power instructions to the generator set;
(9) adjusting output of each generator set according to the power instruction, and calculating a real-time CPS1 value; in this embodiment, the frequency deviation coefficient BiSet to 20 (unit 10 MW/Hz); statistical time ntimeSetting for 10 minutes; control target epsilon of frequency average deviation root mean square in each minute within one year1Set to 0.09 Hz.
Fig. 2 is a two-level step frequency control framework proposed by the method of the present invention. In the figure,. DELTA.f isA real-time frequency deviation signal; sHA high frequency signal that is a frequency deviation; sLA low frequency signal that is a frequency deviation; Δ P is a power command issued by the control module; { G1，G2，…，GNAnd is the number of the generator set. In order to improve the control performance, the decomposition module and the control module are embedded into a double-layer step frequency control framework. Compared with the traditional single-layer frequency control framework, the framework can process frequency analysis and frequency control by different modules, can improve the control precision, and has stronger expansibility and compatibility.

Claims (1)

1. An intelligent power grid frequency control method based on empirical mode decomposition is characterized by comprising a decomposition module and a control module, wherein the decomposition module is an empirical mode decomposition module based on a long-short term memory artificial neural network and the control module comprises a reinforcement learning method and a deep neural network, and a double-layer step-by-step frequency control framework is provided; the method is characterized in that a step-by-step decomposition module and a control module act on double-layer step-by-step frequency control; the method comprises the following steps in the using process:
(1) the control performance standard index is brought into double-layer step frequency control, the control target of the double-layer step frequency control is to make the CPS1 value in the control performance standard index approach 200%, and the CPS1 value is obtained by the following formula:
Figure FDA0003539103640000011
in the formula (I), the compound is shown in the specification,
Figure FDA0003539103640000012
is the average of the regional control errors over 1 minute;
Figure FDA0003539103640000013
is the average of the frequency deviations over 1 minute; b isiIs the frequency deviation coefficient of the control area i,the unit is 10 MW/Hz; n is a radical of an alkyl radicaltimeIs the number of minutes in the statistical time; epsilon1The method is a control target of the average frequency deviation root mean square of the interconnected power grid in each minute within one year; the closer CPS1 is to 200%, the better the control effect is;
(2) the decomposition module is used for empirical mode decomposition of the power grid frequency signal, a long-short term memory artificial neural network in the decomposition module is composed of a plurality of neurons containing gating, a forgetting gate selectively forgets information transmitted by the last neuron, and the forgetting degree is determined according to the following formula:
Zf＝σ(Wf·[ht-1,xt]+bf)
in the formula, ZfA state indicating forgetting to gate; wfA weight matrix representing forgotten gated input values; h ist-1Is the short-term memory state value of the last neuron; x is the number oftIs the input signal of the current neuron; bfIs a forgetting gated bias signal; σ () is a Sigmoid function;
then selects gating pair input signal xtSelectively memorizing:
Zi＝σ(Wi·[ht-1,xt]+bi)
in the formula, ZiIndicating a state of selective gating; wiA weight matrix representing a selection gating input value; biIs a bias signal that selects gating;
the long-term memory state value of the current neuron is calculated by the following formula:
Figure FDA0003539103640000014
in the formula, CtRepresenting a long-term memory state value of a current neuron; ct-1Representing the long-term memory state value of the last neuron;
Figure FDA0003539103640000015
candidate long-term memory state values representing a current neuron, whichThe expression is as follows:
Figure FDA0003539103640000021
in the formula (I), the compound is shown in the specification,
Figure FDA0003539103640000022
a weight matrix representing candidate input values;
Figure FDA0003539103640000023
a bias signal that is a candidate input value;
finally, the expression of the output gating is:
Zo＝σ(Wo·[ht-1,xt]+bo)
in the formula, ZoIndicating the state of the output gate; woA weight matrix representing output gated input values; b is a mixture ofoIs an output gated bias signal;
when the input signal is xtThe output signal h of the current neurontCalculated by the following formula:
ht＝Zo·tanh(Ct)
the trained long-short term memory artificial network decomposes the power grid frequency signal into 9 eigenmode function components IMF1-IMF9 and a residual component r in real time;
(3) the control module is used for adjusting the output of the generator set and comprises a reinforcement learning method and a deep neural network, the reinforcement learning method controls the output of the generator set according to a high-frequency signal obtained by the sum of 4 eigenmode function components of IMF1-IMF 4, and the deep neural network controls the output of the generator set according to a low-frequency signal obtained by the sum of 5 eigenmode functions of IMF 5-IMF 9 and 1 residual component; the high frequency signal and the low frequency signal are obtained by the following formula:
Figure FDA0003539103640000024
Figure FDA0003539103640000025
in the formula, SHIs a high frequency signal; sLIs a low frequency signal; r is the residual component;
(4) the step-by-step decomposition module and the control module are applied to a step-by-step frequency control framework with a two-layer structure, a long-short-term memory artificial neural network in the first layer of decomposition module carries out real-time empirical mode decomposition on a power grid frequency signal after offline training, a reinforcement learning method in the second layer of control module can optimize an output action value on line, and the strong representation capability of a deep neural network can improve the frequency control performance.
CN202011152690.4A 2020-10-26 2020-10-26 Power grid frequency intelligent control method based on empirical mode decomposition Active CN112398142B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011152690.4A CN112398142B (en) 2020-10-26 2020-10-26 Power grid frequency intelligent control method based on empirical mode decomposition

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011152690.4A CN112398142B (en) 2020-10-26 2020-10-26 Power grid frequency intelligent control method based on empirical mode decomposition

Publications (2)

Publication Number Publication Date
CN112398142A CN112398142A (en) 2021-02-23
CN112398142B true CN112398142B (en) 2022-05-13

Family

ID=74596108

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011152690.4A Active CN112398142B (en) 2020-10-26 2020-10-26 Power grid frequency intelligent control method based on empirical mode decomposition

Country Status (1)

Country Link
CN (1) CN112398142B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106126896A (en) * 2016-06-20 2016-11-16 中国地质大学（武汉） The mixed model wind speed forecasting method learnt based on empirical mode decomposition and the degree of depth and system
CN110738010A (en) * 2019-10-17 2020-01-31 湖南科技大学 Wind power plant short-term wind speed prediction method integrated with deep learning model

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103049798B (en) * 2012-12-05 2016-04-20 浙江大学城市学院 A kind of short-term power generation power Forecasting Methodology being applied to photovoltaic generating system
CN108521133B (en) * 2018-04-18 2022-02-01 国网山西省电力公司电力科学研究院 Fire-storage combined scheduling frequency modulation power distribution method based on ensemble empirical mode decomposition
CN109494766A (en) * 2018-11-23 2019-03-19 广西大学 A kind of intelligent power generation control method of manual depth's emotion game intensified learning
CN110059844B (en) * 2019-02-01 2021-10-08 东华大学 Energy storage device control method based on ensemble empirical mode decomposition and LSTM
CN111799820B (en) * 2020-05-27 2022-07-05 广西大学 Double-layer intelligent hybrid zero-star cloud energy storage countermeasure regulation and control method for power system

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106126896A (en) * 2016-06-20 2016-11-16 中国地质大学（武汉） The mixed model wind speed forecasting method learnt based on empirical mode decomposition and the degree of depth and system
CN110738010A (en) * 2019-10-17 2020-01-31 湖南科技大学 Wind power plant short-term wind speed prediction method integrated with deep learning model

Also Published As

Publication number Publication date
CN112398142A (en) 2021-02-23

Similar Documents

Publication Publication Date Title
CN112615379B (en) Power grid multi-section power control method based on distributed multi-agent reinforcement learning
An et al. Short-term wind power prediction based on particle swarm optimization-extreme learning machine model combined with AdaBoost algorithm
CN112733462B (en) Ultra-short-term wind power plant power prediction method combining meteorological factors
CN110766212B (en) Ultra-short-term photovoltaic power prediction method for historical data missing electric field
CN106600136B (en) Power section out-of-limit control efficiency evaluation method
CN110429652B (en) Intelligent power generation control method capable of expanding deep width self-adaptive dynamic planning
CN106920014A (en) A kind of short-term load forecasting method and device
CN116167531A (en) Photovoltaic power generation prediction method based on digital twin
CN111555297A (en) Unified time scale voltage control method with tri-state energy unit
CN113313306A (en) Elastic neural network load prediction method based on improved wolf optimization algorithm
CN112651519A (en) Secondary equipment fault positioning method and system based on deep learning theory
CN115932592A (en) Electric vehicle battery fault diagnosis method based on digital twin model
CN114742283A (en) Medium-voltage line loop closing current estimation and risk assessment method based on deep learning network
CN114429248A (en) Transformer apparent power prediction method
CN112398142B (en) Power grid frequency intelligent control method based on empirical mode decomposition
CN116885792A (en) Data-driven model prediction control voltage out-of-limit quantization method
CN114707613B (en) Layered depth strategy gradient network-based power grid regulation and control method
CN111967660A (en) Ultra-short-term photovoltaic prediction residual error correction method based on SVR
CN114204546B (en) Unit combination optimization method considering new energy consumption
CN112232714B (en) Deep learning-based risk assessment method for distribution network under incomplete structural parameters
CN114298429A (en) Power distribution network scheme aided decision-making method, system, device and storage medium
Huang et al. Short-term load forecasting based on similar day approach and intelligent algorithm using analytic hierarchy process
Liu et al. Short-Term Wind Speed and Wind Power Prediction Based on Meteorological Model Modification
Duan et al. Estimation of Solar Power by Combining Ground Measured and Satellite Data Using Artificial Neural Networks
Li Research on the application of artificial intelligence in diagnosing power system faults

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