CN109742773A - A kind of adaptive wide area damping control - Google Patents
A kind of adaptive wide area damping control Download PDFInfo
- Publication number
- CN109742773A CN109742773A CN201910086699.0A CN201910086699A CN109742773A CN 109742773 A CN109742773 A CN 109742773A CN 201910086699 A CN201910086699 A CN 201910086699A CN 109742773 A CN109742773 A CN 109742773A
- Authority
- CN
- China
- Prior art keywords
- time lag
- signal
- wide area
- grhdp
- adaptive
- 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.)
- Granted
Links
Landscapes
- Feedback Control In General (AREA)
Abstract
The present invention relates to a kind of adaptive wide area damping controls, comprising: adaptive time lag compensation device is used for from N number of first wide area measurement system signal of acquisition, and carries out adaptive time lag compensation to it, exports N number of second wide area measurement system signal；N input and output GrHDP unit is based on adaptive dynamic programming algorithm, obtains N number of first coordinated control signal for receiving N number of second wide area measurement system signal；Adaptive time lag compensation device is also used to receive N number of first coordinated control signal, and is based on GrHDP model algorithm, and N number of second coordinated control signal is calculated.The present invention carries out multi input and multi output using GrHDP unit, and can simultaneously multiple low-frequency oscillation mode of electric system be carried out with the coordination self adaptive control of multiple target, improve the adaptive ability to electric system.In addition, carrying out time lag compensation to the first wide area measurement system signal and the first coordinated control signal, time lag is greatly avoided to influence the control of electric system, improving controller improves dynamics to the damping of electric system.
Description
Technical field
The present invention relates to New-energy power system technical fields, more particularly to a kind of adaptive wide area damping control.
Background technique
With becoming increasingly conspicuous for energy problem, the permeability of the new energy such as wind-force, photovoltaic in the power system is higher and higher.
Due to the new energy such as wind, light power output mostly have the characteristics that it is common: by force randomness, fluctuation, intermittence, with its permeability
Increase, the safe and stable operation of electric system will be faced with formidable challenges, and particularly, the section of electric system may be made low
Frequency oscillation problem is even more serious.
Since wind-force, photovoltage model generally use vector control strategy, active power of output and reactive power have only
Adjustable feature is found, in conjunction with Wide Area Measurement System (wide-area measure system, WAMS) technology, can reasonably be set
It counts wide area damping control (wide-area damping controller, WADC), the new energy model such as control wind-force, photovoltaic
Reactive power power output, improve the damping of system, improve the transient stability of system.
However, current conventional wide area damping controller (conventional wide-area damping
Controller, C-WADC) it is poor to the operating condition adaptability of system real-time change, and existing adaptive time lag compensation device
(adaptive delay compensation, ADC) error of time lag compensation is very big under certain condition.Therefore, current skill
Art improves not enough the damping of electric system, cannot effectively inhibit the low-frequency oscillation of electric system.
Summary of the invention
The present invention provides a kind of adaptive wide area damping control, to solve in the prior art damping controller to electric power
The damping of system improves problem not in place.
The technical scheme to solve the above technical problems is that a kind of adaptive wide area damping control, comprising:
The adaptive time lag compensation device, for obtaining N number of first wide area measurement system signal from electric system, to described N number of
After first wide area measurement system signal carries out adaptive time lag compensation, output and N number of one-to-one N of first wide area measurement system signal
A second wide area measurement system signal；
N input and output GrHDP unit is calculated for receiving N number of second wide area measurement system signal, and based on GrHDP model
Method is calculated and the one-to-one N number of first coordinated control signal of N number of second wide area measurement system signal and exports；
The adaptive time lag compensation device is also used to receive N number of first coordinated control signal, and to described N number of
After one coordinated control signal carries out adaptive time lag compensation, obtain N number of correspondingly with N number of first coordinated control signal
Second coordinated control signal is simultaneously transmitted to the electric system, wherein the N is positive integer, and N > 1.
The beneficial effects of the present invention are: the present invention uses GrHDP unit, and multi input and multi output are carried out, it can be right simultaneously
Multiple low-frequency oscillation mode of electric system carry out the coordination self adaptive control of multiple target, improve to the adaptive of electric system
Ability.In addition, coordinated control signal is generated based on the input signal after time lag compensation after carrying out time lag compensation to input signal,
Because the control signal that GrHDP unit generates is adaptable to the control signal of electric system current state, when the control signal passes
When the control point being sent in electric system, control signal can have certain transmission delay, the letter for causing electric system to receive
Number it is not desired optimal control signal, influences control effect, therefore, time lag benefit further is carried out to the control signal
It repays.Time lag compensation twice, improving controller improves dynamics to the damping of electric system, effectively inhibits the low frequency vibration of electric system
It swings.
Based on the above technical solution, the present invention can also be improved as follows.
Further, the adaptive time lag compensation device includes: n time lag compensation submodule and additional gain computing unit；
The n time lag compensation submodule, for carrying out adaptive time lag compensation to each input signal；
It is first wide to calculate this for the communication delay according to each input signal for the additional gain computing unit
The corresponding additional gain value of domain measurement signal；
Wherein, the input signal is first wide area measurement system signal or the first coordinated control signal；
Then transmission function of the adaptive time lag compensation device to each input signal are as follows:
Wherein, K indicates the additional gain value of the input signal, SDCi(s) i-th of time lag compensation is indicated
The transmission function of submodule, βi(τ) is expressed as the weight of i-th of time lag compensation submodule, and n is the time lag compensation submodule
The number of block, value are that the order of input signal time lag transmission function after simplifying adds 1.
Further beneficial effect of the invention is: introducing additional gain computing unit, calculates additional gain value, and existing
Multiplied by the additional gain value on the basis of the transmission function calculating formula of adaptive time lag compensation device, i.e., mended according to each time lag
The transmission function of submodule and the gain size K of corresponding weight, additional gain link are repaid, the adaptive time lag of the present invention is calculated and mends
The transmission function ADC (s) for repaying device is reduced because carrying out simplifying error brought by processing to input signal time lag transmission function,
Improve time lag compensation accuracy.
Further, input signal time lag transmission function after the simplification are as follows:
Wherein, s indicates that complex variable, τ are expressed as the communication delay.
Further beneficial effect of the invention is: being carried out using time lag transmission function of the different methods to input signal close
Like simplification, then the simplified input signal time lag transmission function obtained is different, then its order also can be different.The present invention is to input
The time lag transmission function of signal carries out first approximation and simplifies, and compared with former ADC derivation formula, this expression formula order increases one
Times, approximate accuracy increases.
Further, the calculation formula of the additional gain value are as follows:
Wherein, τ indicates the communication delay.
Further beneficial effect of the invention is: ideally, input signal after the compensation of time lag compensation device, when
Stagnant transmission function meets equationI.e. the delayed phase of input signal only with set time constant Tc
(time constant relevant to system dynamic characteristic) is related, and unrelated with communication delay τ, thus for different communication delays,
Its lagging phase introduced is the same.Therefore, it is based on above-mentioned simplified input signal time lag transmission function, introduces K, is reduced
Because to input signal time lag transmission function carry out second it is approximate caused by approximate error so that expression formula more accurately at
It is vertical.
Further, the N input and output GrHDP unit includes: the first GrHDP neural network and its first weight；
The first GrHDP neural network is multiple groups operation work of the 2nd GrHDP neural network based on the electric system
Second weight of condition and fault condition information and the 2nd GrHDP neural network is trained to obtain；
Then the N input and output GrHDP unit is specifically used for: receiving N number of second wide area measurement system signal, and is based on
GrHDP model algorithm is adjusted first weight of the first GrHDP neural network, be calculated with it is described N number of
The one-to-one N number of first coordinated control signal of second wide area measurement system signal simultaneously exports.
Further beneficial effect of the invention is: the present invention is based on the typical multiple groups operating conditions and failure of electric system
Work information is trained GrHDP neural network, obtains the N input and output that can be adapted with Operation of Electric Systems operating condition
GrHDP unit improves the control effect of control signal.
Further, the 2nd GrHDP neural network includes executing network, target network and evaluation network, wherein described
The order for executing the weight matrix of network, target network and evaluation network is determined according to the N respectively, so that described second
The first GrHDP neural network obtained after GrHDP neural network is trained carries out N input and N output.
Further beneficial effect of the invention is: N is based on, to the weight of three neural networks in GrHDP neural network
Order of matrix number is adjusted, so that the first GrHDP nerve obtained after the 2nd GrHDP neural metwork training
Network, which receives N number of second wide area measurement system signal and calculates output described N number of first through the GrHDP model algorithm, to be coordinated to control
Signal processed.
Further, the N is 2.
Further beneficial effect of the invention is: when N input and output GrHDP unit be two-output impulse generator GrHDP unit,
Its coordinated control ability is preferable, and obtained coordinated control signal is preferable to the coordinated control effect of electric system.
Detailed description of the invention
Fig. 1 is a kind of structural schematic diagram of adaptive wide area damping control provided by one embodiment of the present invention；
Fig. 2 is a kind of structural schematic diagram for adaptive wide area damping control that another embodiment of the present invention provides；
Fig. 3 is a kind of workflow block diagram for adaptive wide area damping control that another embodiment of the present invention provides；
Fig. 4 is the New-energy power system schematic diagram containing blower and photovoltage model that another embodiment of the present invention provides；
Fig. 5 (a) is that time lag transmission function when additional gain unit not being added that another embodiment of the present invention provides is approximate
The three-dimensional figure that simplified error changes with time lag value and modal frequency；
Time lag transmission function approximate letter when the addition additional gain unit that Fig. 5 (b) provides for another embodiment of the present invention
The three-dimensional figure that the error of change changes with time lag value and modal frequency；
Fig. 6 is the active power transient state curve of system line 2-25 and 1-39；
Fig. 7 is the transient state curve of correlated variables inside Double input-double output GrHDP unit；
Fig. 8 is the active power transient state curve of system line 2-25 and 1-39 under the conditions of 100ms fixed time lag；
Fig. 9 is the active power transient state curve of system line 2-25 and 1-39 under the conditions of 200ms fixed time lag；
Figure 10 is the active power transient state curve of system line 2-25 and 1-39 under the conditions of stochastic Time-Delay；
Figure 11 is stochastic Time-Delay change curve；
Figure 12 is the weight change curve of time lag compensation submodule in adaptive time lag compensation device under the conditions of stochastic Time-Delay；
The additional gain that Figure 13 calculates for additional gain computing unit in time lag compensation device adaptive under the conditions of stochastic Time-Delay
The change curve of value K.
Specific embodiment
The principle and features of the present invention will be described below with reference to the accompanying drawings, and the given examples are served only to explain the present invention, and
It is non-to be used to limit the scope of the invention.
Embodiment one
A kind of adaptive wide area damping control, as shown in Figure 1, comprising: adaptive time lag compensation device and N input and output
GrHDP unit.Wherein, adaptive time lag compensation device, for obtaining N number of first wide area measurement system signal from electric system, to N number of the
After one wide area measurement system signal carries out adaptive time lag compensation, output and N number of first wide area measurement system signal one-to-one N number of second
Wide area measurement system signal；N input and output GrHDP unit is calculated for receiving N number of second wide area measurement system signal, and based on GrHDP model
Method is calculated and the one-to-one N number of first coordinated control signal of N number of second wide area measurement system signal and exports；When adaptive
Stagnant compensator is also used to receive N number of first coordinated control signal, and carries out adaptive time lag to N number of first coordinated control signal
After compensation, obtains and the one-to-one N number of second coordinated control signal of N number of first coordinated control signal and be transmitted to power train
System, wherein N is positive integer, and N > 1.
It should be noted that N input and output GrHDP unit design considerations GrHDP (Goal representation
Heuristic Dynamic Programming, GrHDP, the heuristic dynamic programming containing target network) model, by building
Three intrinsic nerve network (executing network, target network and evaluation network) Lai Shixian multi inputs and multi output, compared to existing
Single input export structure GrHDP model, the GrHDP model of N input/output structure calculating external strengthening signal r (t), interior
When portion enhanced signal S (t), cost function J (t), it can be related to the respective influence of N number of first wide area measurement system signal simultaneously,
So the N input and output GrHDP unit in the present embodiment can coordinate the information of N number of first wide area measurement system signal, conjunction is calculated
Suitable built-in variable (including external strengthening signal r (t), inside enhanced signal S (t), cost function J (t) etc.), finally obtains N
A control signal for being suitable for electric system current state.
The present invention uses GrHDP unit, and carries out multi input and multi output, can shake simultaneously to multiple low frequencies of electric system
The coordination self adaptive control that mode carries out multiple target is swung, the adaptive ability to electric system is improved.In addition, first believing input
Number carry out time lag compensation after, based on reflection real-time working condition time lag compensation after input signal generation system control signal, later
Time lag compensation is carried out to control signal again, twice time lag compensation, improving controller improves dynamics to the damping of electric system, has
Effect inhibits the low-frequency oscillation of electric system.
Embodiment two
On the basis of example 1, adaptive time lag compensation device includes: n time lag compensation submodule and additional gain meter
Calculate unit.Wherein, n time lag compensation submodule, for carrying out adaptive time lag compensation to each input signal；Additional gain meter
It calculates unit and calculates the corresponding additional gain value of the input signal, wherein defeated for the communication delay according to each input signal
Entering signal is the first wide area measurement system signal or the first coordinated control signal；Then adaptive time lag compensation device is to each input signal
Transmission function are as follows:
Wherein, K indicates the additional gain value of the input signal, SDCi(s) transmitting of i-th of time lag compensation submodule is indicated
Function, βi(τ) is expressed as the weight of i-th of time lag compensation submodule, and n is the number of time lag compensation submodule, and value is to simplify
The order of input signal time lag transmission function adds 1 afterwards.
It should be noted that since a kind of time lag compensation of existing adaptive time lag compensation device calculates that process is as follows:
Firstly, to input signal time lag transmission function Gd(s) it carries out time lag approximation twice and simplifies process:
Wherein, s is multiple
Variable, τ are expressed as communication delay.
Secondly, the transmission function of n time lag compensation submodule in adaptive time lag compensation device:
The weight computing formula of time lag compensation submodule is as follows:
Wherein, TcFor time constant relevant to system dynamic characteristic, general value range is [0.01s, 0.1s], βiFor
The weight of i-th of time lag compensation submodule, TiFor the time constant of i-th of time lag compensation submodule, and TiValue between system
Between maximum communication time lag and minimal communications time lag, to obtain preferable time lag compensation effect, s is complex variable.
So, the calculation formula of the transmission function of existing adaptive time lag compensation device are as follows:
Meet formulaI.e. at this time the delayed phase of input signal only with set time constant
Tc(time constant relevant to system dynamic characteristic) is related, and unrelated with communication delay τ, thus when for different communication
Stagnant, the lagging phase introduced is the same.The introduced permanent delay phase of time lag is low in electric system section by considering
Frequency vibration, which is swung, designs lead-lag link to compensate in band limits.
But due to GDIt (s) is approximately 1 by molecule, i.e., when will communicate when the further molecule of progress simplifies approximate processing
Stagnant relevant part is approximately 0, and when time lag is larger or signal modal frequency is larger, about there are large errors for equation for this, is influenced
Control accuracy.
Therefore, present invention introduces additional gain value K, reduce caused by simplifying approximate processing because of input signal time lag transmission function
Time lag compensation error.
Embodiment three
On the basis of example 2, input signal time lag transmission function after simplifying are as follows:
Wherein, s indicates that complex variable, τ are expressed as communication delay.
It should be noted that specifically, using single order Pade approximate transform to the time lag transmission function G of input signald(s)
It carries out approximation to simplify, simplified transmission function is GD(s), calculation formula is as follows:
To GD(s) it is further deformed, obtains following formula:
Wherein, when τ is the communication of input signal
Stagnant, s is complex variable.
The formula is the input signal time lag transmission function of the present embodiment, it should be noted that uses different methods pair
The time lag transmission function of input signal carries out approximation twice and simplifies, then the simplified input signal time lag transmission function obtained is not
Together, then its order also can be different.Compared with the present embodiment existing ADC derivation formula described in the embodiment two, this expression formula order
One times is increased, approximate accuracy increases.
Example IV
On the basis of embodiment three, the calculation formula of additional gain value are as follows:
Wherein, τ indicates communication delay.
Specifically, the derivation process of the K of the embodiment is shown below:
Wherein, K is the time lag transmission function G in embodiment threeD(s) inverse of molecule item enables s=2 π j, obtains
Expression formula, size are only related with time lag value.
CauseBased on the time lag transmission function simplified expression, can derive
The calculation formula of the transmission function of each time lag compensation submodule is as follows:
The weight of time lag compensation submodule is calculated, calculation formula is as follows:
Wherein, TcFor time constant relevant to system dynamic characteristic, general value range is [0.01s, 0.1s], βiFor
The weight of i-th of time lag compensation submodule, TiFor the time constant of i-th of time lag compensation submodule, and TiValue between system
Between maximum communication time lag and minimal communications time lag, to obtain preferable time lag compensation effect, s is complex variable
Therefore, the transmission function of the adaptive time lag compensation device of the present embodiment are as follows:N takes simplified input signal time lag transmission functionOrder add 1, i.e.,
9。
At this point, meeting formulaThat is the delayed phase of input signal is only normal with the set time
Number Tc(time constant relevant to system dynamic characteristic) is related, and unrelated with communication delay τ, thus when for different communication
Stagnant, the lagging phase introduced is the same.
Therefore, it is based on above-mentioned simplified input signal time lag transmission function, introduces K, reduces and input signal time lag is passed
Delivery function carries out approximate error caused by second of approximation, so that expression formula is more accurately set up.
It should be noted that according to expression formulaRight formula is permanent delay phase, considers area
Between low-frequency oscillation band limits, rationally design lead-lag compensator lagging phase is compensated.Due to K and time lag value and
Modal frequency is related, under the conditions of considering section low-frequency oscillation, can approximating assumption frequency be 1Hz.
Embodiment five
On the basis of one any embodiment into example IV of embodiment, N input and output GrHDP unit includes first
GrHDP neural network and its first weight；N input and output GrHDP unit is the 2nd GrHDP neural network based on electric system
Multiple groups operating condition and the second weight of fault condition information and the 2nd GrHDP neural network are trained to obtain；Then N is inputted
Output GrHDP unit is specifically used for: receiving N number of second wide area measurement system signal, and based on GrHDP model algorithm to the first GrHDP
First weight of neural network is adjusted, and is calculated and is coordinated with N number of second wide area measurement system signal one-to-one N number of first
Control signal simultaneously exports.
It should be noted that the training process of the 2nd GrHDP neural network can include: the 2nd GrHDP nerve net of setting
The parameter of network；The initial weight of the 2nd GrHDP neural network of random setting, and several groups electric system typical case's operating condition is set
And fault condition, off-line training is carried out to the 2nd GrHDP neural network using initial weight；By trained neural network weight
As the initial weight of on-line study, as the first weight, trained 2nd GrHDP neural network is the first GrHDP mind
Through network.First GrHDP neural network is receiving after the second wide area measurement system signal of time lag compensation, its net of on-line tuning
Network internal weights, to obtain the control signal by optimization.
Wherein, the parameter of the 2nd GrHDP neural network include: execute the input layer number of network, node in hidden layer,
Output layer number of nodes, learning rate, the number of iterations upper limit, error margin and weight range；Evaluate the input layer of network
Number, node in hidden layer, output layer number of nodes, learning rate, the number of iterations upper limit, error margin and weight range；Target
The input layer number of network, node in hidden layer, output layer number of nodes, learning rate, the number of iterations upper limit, error margin with
And weight range.
The initial weight of 2nd GrHDP neural network is each neural network (executing network, mesh network and evaluation network)
Corresponding initial weight.
The present embodiment is using several groups electric system typical case operating condition and fault condition information to the 2nd GrHDP nerve net
Network is trained, and the first GrHDP neural network that training obtains adapts to system conditions variation, and belongs to data-driven side
Formula has preferable practical value and control performance independent of system mathematic model.
Embodiment six
On the basis of embodiment five, the 2nd GrHDP neural network includes executing network, target network and evaluation network,
Wherein, the order for executing the weight matrix of network, target network and evaluation network is determined according to N respectively, so that through second
The first GrHDP neural network obtained after GrHDP neural metwork training receives N number of second wide area measurement system signal and through GrHDP mould
Type algorithm, which calculates, exports N number of first coordinated control signal.
It should be noted that in model: executing the input layer of network to the weight matrix W of hidden layera(1)Order (for
ma(1)×na(1), ma(1)For the number of input signal, na(1)For the number of hidden layer neuron), the weight of hidden layer to output layer
Matrix Wa(2)Order is (for ma(2)×na(2), ma(2)For the number of hidden layer neuron, na(2)For the number of output control signal)；It comments
Weight matrix W of the input layer of valence network to hidden layerc(1)Order (for mc(1)×nc(1), mc(1)Add for the number of input signal
The number of internal enhanced signal, nc(1)For the number of hidden layer neuron)；Weight matrix of the target network input layer to hidden layer
Wr(1)Order (for mr(1)×nr(1), mr(1)Add the number of output control signal, n for the number of input signalc(1)For hidden layer mind
Number through member)；The form of external strengthening signal r (t) includes the first wide area measurement system signal of all inputs, needs simultaneously
Illustrate, external strengthening signal is the function being manually set, and which includes external strengthening signals for the input of target network, originally
Function should include all input signals in matter, and be arranged to reasonable penalty term form.It is additionally based on side appropriate
Method carries out pre-training to model, and model is made to obtain suitable initial weight.
The present embodiment is based on N, adjusts to the order of the weight matrix of three neural networks in GrHDP neural network
It is whole, so that GrHDP neural network can be received and the N number of input signal of Coordination Treatment and be obtained based on adaptive dynamic programming algorithm
To and export N number of coordinated control signal.
Embodiment seven
On the basis of embodiment six, when N is 2, the structure of adaptive wide area damping control is as shown in Figures 2 and 3, and two
A input signal, two output signals, control process are as follows:
Adaptive time lag compensation device obtains two the first wide area measurement system signals from electric system, to two the first wide area measurements
After signal carries out adaptive time lag compensation, output and one-to-one two the second wide area measurements of two the first wide area measurement system signals
Signal；Two-output impulse generator GrHDP unit receives two the second wide area measurement system signals, and is based on adaptive dynamic programming algorithm,
It obtains and one-to-one two the first coordinated control signals of two the second wide area measurement system signals and exports；Adaptive time lag compensation
Device receives two the first coordinated control signals, and after carrying out adaptive time lag compensations to two the first coordinated control signals, obtains
With one-to-one two the second coordinated control signals of two the first coordinated control signals and be transmitted to electric system.
When N be 3 when, three input signals, three output signals, the course of work with N be 2 when, details are not described herein.
Preferably, 2 N.
For example, being illustrated in figure 4 containing blower and grid-connected 10 machine of New England, the 39 node power system of photovoltage model, in figure
Number 1~39 respectively represents 39 buses, and G1~G10 respectively represents 10 generators.Original system include 10 generators, 39
Node, 12 transformers (are not drawn, because will do it mark change processing, Suo Youbian in the calculating of electric system in Fig. 4
The no-load voltage ratio of depressor is all approximately 1, and being equivalent to directly to omit), 34 routes represent one of New England, U.S. state
345kV electric power networks represent the Canadian part being connected with system wherein the generator on No. 39 buses is an equal check-ins
Electric power networks, inertia time constant is very big.The reference power of system is 100MVA, reference voltage 345V.The 4 of original system
Doubly-fed wind turbine model and grid-connected model are respectively connected at number bus and No. 16 buses to form new energy electric power
System, initially injecting active power is respectively 215MW and 200MW, and reactive power is 0.Wherein, the nothing of grid-connected model
Function power control circuit realizes that the Reactive Power Control circuit of blower model then passes through output by the control of grid entry point voltage close loop
Reactive power closed-loop control is realized.New-energy power system equivalence mould as shown in Figure 4 is built in MATLAB/Simulink
Type, the test object system as the embodiment of the present invention.
Two-output impulse generator GrHDP unit can pass through the first wide area of electric system based on adaptive dynamic programming algorithm
Measuring signal (i.e. status feedback information) carries out on-line study to obtain the first optimal coordinated control signal (i.e. control letter
Number).Dual input export GrHDP unit model be to be realized based on neural network, the neural network include: execute network,
Target network and evaluation network, wherein execute network and be used to be generated correspondingly according to two the second wide area measurement system signals
Two optimum control signals；Evaluating network is actually a function approximator, and output is cost function J (t), for approximation
Indicate the summation of following all internal enhanced signals, guidance executes the weighed value adjusting of network, to optimize output control signal；Mesh
Mark network is for automatically generating internal enhanced signal S (t), compared with external strengthening signal r (t), can preferably adjust input letter
Mapping relations number between output signal, so that cost function J (t) can preferably evaluate the superiority and inferiority of output signal.
It is illustrated in figure 2 the control structure block diagram of the present embodiment, adaptive wide area damping control provided in this embodiment
Including two-output impulse generator GrHDP unit and improve adaptive time lag compensation device；The input terminal of two-output impulse generator GrHDP unit
For receiving by improving adaptive compensated second wide area measurement system signal of time lag compensation device；Two-output impulse generator GrHDP is mono-
Member output end also pass through the compensation for improving adaptive time lag compensation device after obtain the second coordinated control signal, and send respectively to
In the Reactive Power Control circuit of blower model and photovoltage model, to realize the control to the idle power output of its transient state, inhibit system
The section low-frequency oscillation of system.
Adaptive time lag compensation device includes n time lag compensation submodule (sub-delay compensator, SDC) and one
A additional gain computing unit.Additional gain computing unit can reduce the approximate processing error to time lag signal, such as Fig. 5 (a) and
It is respectively the approximate error for simplify to time lag transmission function processing shown in Fig. 5 (b), value illustrates close closer to 1
It is more accurate like handling.Comparison diagram 5 (a) and Fig. 5 (b), it is found that carry out time lag transmission function at approximation after gain K is added
The caused error size of reason makes moderate progress.The transmission function for improving adaptive time lag compensation device is n time lag compensation submodule
Transmission function weighted sum and K product, and the weight coefficient of each time lag compensation submodule and the size of K with
The communication delay τ of wide area signal is related, so that different communication delays can be made by improving adaptive time lag compensation device
Corresponding compensation, to effectively eliminate the lagging phase that communication delay introduces in wide area signal；The number n of time lag compensation submodule
Order to simplify obtained transmission function to time lag approximation adds one；In the present embodiment, time lag simplifies the order of transmission function
9 are taken as 8, n.
Adaptive time lag compensation device is transferred to two-output impulse generator GrHDP unit after carrying out time lag compensation to input signal.
Two-output impulse generator GrHDP unit is calculated and the current operating condition of power grid according to the second wide area measurement system signal and network parameter
The first adaptable coordinated control signal；Specifically comprise the following steps:
(21) parameter of GrHDP neural network is set, can generally use trial and error procedure, in the present embodiment relevant parameter
Specific setting is as shown in table 1:
The parameter setting of 1 GrHDP neural network of table
(22) the first initial weight of GrHDP neural network is set at random, and several typical operating conditions events of system are set
Barrier carries out off-line training to GrHDP neural network；It is initially weighed trained neural network weight as the second of on-line study
Value；
(23) using the second wide area measurement system signal by time lag compensation as the input of GrHDP neural network, intrinsic nerve
Network adjusts network weight by on-line study, obtains the first coordinated control signal by optimization；Wherein, the first coordinated control
Signal includes additional reactive power reference qref Δ Qref(t) and auxiliary voltage reference value Δ Vref(t), it is respectively used to control blower mould
Power output that the transient state of type and photovoltage model is idle；
(24) adaptive time lag compensation is carried out to the first coordinated control signal of GrHDP unit output, and finally sent to wind
In the control loop of machine model and photovoltage model.
By carrying out linearisation model analysis it is found that not putting into WADC (wide-area damping to steady-state operation system
Control, wide area damping control) when, the section mode 2 of system and the damping ratio of section mode 3 be respectively 1.38%,
2.14%, belong to underdamping section mode, needs to be promoted the damping ratio of mode by wide-area damping control, improve the transient state of system
Stability.Therefore, the principle here based on considerable degree, respectively in selection system it is considerable for two target modalities degree it is higher and
Two weaker loop line road power Ps of intercoupling property26-27、P1-39As the input signal of wide area damping control, so that wide area hinders
Buddhist nun's controller can effectively improve the damping ratio of target modalities.
Specifically, being based on above-mentioned two target modalities, two examples are set, are provided with the first example to verify this implementation
The second example is arranged to the adaptability of system condition to verify the adaptive of the present embodiment proposition in the adaptive wide area damping control of example
Answer time lag compensator to the compensation effect of wide area measurement system signal communication delay.
Wherein, in the first example, wide area network traffic time lag is not considered, system operates in initial steady state point, in order to test
Card the present embodiment A-WADC (adaptive wide-area damping control, adaptive wide area damping control) is to being
It unites the adaptability of different operating conditions, following scene is set: three-phase shortcircuit, 1.1s occurs when 1s at the nearly bus 17 of route 17-27
When cut off faulty line, reclosing success when 2.1s；Payload when 30s at bus 26 is uprushed from 1.39pu as 5pu；60s
When bus 20 at payload from 6.28pu anticlimax be 1pu；Three-phase shortcircuit, 90.1s occurs when 90s at the nearly bus 2 of route 2-3
When cut off faulty line.
Select two higher line powers of degree considerable for target underdamping section mode as observation object (such as Fig. 6
It is shown), in addition, executing network error Ea, external strengthening signal r (t), internal reinforcing letter inside dual input output GrHDP unit
Number S (t), cost function J (t), execute network input layer to hidden layer weight Wa(1)And network hidden layer is executed to output
The weight W of layera(2)Transient state curve difference it is as shown in Figure 7.
The transient response characteristic under different condition after the system failure is compared respectively, as shown in fig. 6, not putting into any wide area resistance
When Buddhist nun's controller (wide-area damping control, WADC), the vibration of low frequency by a relatively large margin can occur after a failure for system
It swings, the duration is longer, and transient stability is poor；Put into conventional wide area damping controller (conventional wide-area
Damping control, C-WADC) when, the low-frequency oscillation of post-fault system makes moderate progress；Put into adaptive wide-area damping control
When device (adaptive wide-area damping control, A-WADC), the low-frequency oscillation of system after a failure has obviously
Improve, system vibrates after continuous four failures of time of origin calms down rapidly, and transient stability is effectively promoted, emulation knot
Fruit embodies the good control performance of A-WADC and the adaptability to system difference operating condition.Wherein, the upper following figure in Fig. 6
Represent different line fault.
As shown in fig. 7, the transient state curve of Double input-double output GrHDP unit inside correlated variables, comprising: execute network
Error E a, external strengthening signal r (t), inside enhanced signal S (t), cost function J (t), execute network input layer to imply
The weight W of layera(1)With the weight W of execution network hidden layer to output layera(2).Every time after system jam, dual input output
Wave can correspondingly occur in external strengthening signal r (t), inside enhanced signal S (t) and cost function J (t) in GrHDP unit
It is dynamic, so that the error E a for executing network is more than error margin value, therefore executes network and start corrective networks weight Wa(1)And Wa(2)；
By adjusting, the output control signal of A-WADC is optimized, after the modified weight for executing network is basically completed, A-WADC energy
The new operating status of system is enough adapted to, meanwhile, the low-frequency oscillation of post-fault system is also calmed down substantially.Simulation results show
A-WADC can be updated by neural network weight realizes on-line study, and then adapts to the variation of system conditions, and can be not
With under operating condition and different faults, preferable low frequency oscillations rejection ability is kept.
In addition, the second example is used to verify the compensation effect for improving adaptive time lag compensation device to wide area signal communication delay
Following two groups of scenes: fixed time lag and stochastic Time-Delay are respectively set in fruit.
Scene I: fixed time lag
Four go back tos wide area measurement system signal channels are respectively set, and (twice wide area measurement system signal channels, twice wide-area control signals are logical
Road) communication delay be 100ms and 200ms, system operates in initial steady state point, occurs when 1s at the nearly bus 17 of route 17-27
Three-phase shortcircuit, when 1.1s, cut off faulty line, reclosing success when 2.1s.
Under the conditions of 100ms fixed time lag, comparison investment improve ADC before and after A-WADC and without C- under the conditions of time lag
WADC control effect, as shown in Figure 8.Under the conditions of 200ms fixed time lag, the A-WADC of comparison investment original ADC and ADC after improvement
And the control effect without C-WADC under the conditions of time lag, as shown in Figure 9.
As shown in figure 8, A-WADC is after system jam when not putting into improvement ADC under the conditions of 100ms fixed time lag
Original control performance cannot be maintained, system cannot keep transient stability；When investment improves ADC, A-WADC is in system event
It still is able to maintain good control performance in transient process after barrier, and its control effect will be substantially better than C-WADC.Emulation
The result shows that improving the influence that ADC model is capable of effective compensation wide-area communication time lag, A-WADC is made to maintain good control performance.
Wherein, the upper following figure in Fig. 8 represents different line fault.
As shown in figure 9, the A-WADC of investment original ADC can not after system jam under the conditions of 200ms fixed time lag
Original control performance is maintained, system loses transient stability；Investment improves transient process of the A-WADC of ADC after the system failure
In good control performance can still be maintained, calm down low frequency oscillations rapidly, effect will be substantially better than C-WADC.It is imitative
Very the result shows that, under the conditions of time lag is biggish, former ADC model may cannot keep ideal time lag compensation due to approximate error
Ability, and still be able to play good time lag compensation effect by improved improvement ADC model, it improves to a certain extent
The ability of its time lag compensation.Wherein, the upper following figure in Fig. 9 represents different line fault.
It can be seen from simulation result in this scene communicate time lag presence meeting so that wide area damping control control
Degradation, the system that results even in when serious lose transient stability；Fixed communication time lag is coped with by putting into ADC
It influences, A-WADC is made to keep good control performance, and compared to former ADC model, improved improvement ADC model has more preferably
Time lag compensation ability.
Scene II: stochastic Time-Delay
Four go back tos wide area signal channels (two go back tos wide area measurement measuring signal channels, two go back tos wide-area control signal paths) are arranged to deposit
In the stochastic Time-Delay of size time-varying, system operates in initial steady state point, and it is short at the nearly bus 17 of route 17-27 three-phase to occur when 1s
Road, when 1.1s, cut off faulty line, reclosing success when 2.1s.
A-WADC before and after comparison investment improvement ADC and the control effect without the A-WADC under the conditions of time lag, such as Figure 10
It is shown, wherein the upper following figure in Figure 10 represents different line fault；The transient state curve of stochastic Time-Delay is as shown in figure 11；Transient state
In the process improve ADC inside each compensation submodule weight coefficient and additional gain K change curve respectively as Figure 12 with
Shown in Figure 13.
Under conditions of stochastic Time-Delay as shown in figure 11, as shown in Figure 10, the A-WADC for not putting into improvement ADC has been lost
Control ability is gone, system is unable to maintain that transient stability after a failure；The A-WADC that investment improves ADC can be tieed up in transient process
Good control performance is held, system is made to keep transient stability after a failure, is compared with the A-WADC under the conditions of no time lag, control
Effect wants a bit weaker, and the oscillation amplitude of system transient modelling process is somewhat larger, but still oscillation can be made rapid within a short period of time
It calms down.
As shown in figure 12, the weight coefficient for improving each compensation submodule (9) inside ADC becomes with the variation of time lag
Change, enables improvement ADC model to adaptively compensate for different time lag values, maintain good time lag compensation effect.Such as Figure 13
It is shown, the gain size K of additional gain link likewise as time lag variation and change so as to time lag transmission function
Approximation simplification is more accurate, improves the effect of time lag compensation.
The foregoing is merely presently preferred embodiments of the present invention, is not intended to limit the invention, it is all in spirit of the invention and
Within principle, any modification, equivalent replacement, improvement and so on be should all be included in the protection scope of the present invention.
Claims (7)
1. a kind of adaptive wide area damping control characterized by comprising
Adaptive time lag compensation device, for obtaining N number of first wide area measurement system signal from electric system, and it is wide to described N number of first
Domain measurement signal carries out adaptive time lag compensation, and output and N number of first wide area measurement system signal one-to-one N number of second are wide
Domain measurement signal；
N input and output GrHDP unit for receiving N number of second wide area measurement system signal, and is based on GrHDP model algorithm, meter
Calculation is obtained and the one-to-one N number of first coordinated control signal of N number of second wide area measurement system signal and is exported；
The adaptive time lag compensation device is also used to receive N number of first coordinated control signal, and to N number of first association
Regulate and control after signal processed carries out adaptive time lag compensation, obtains and N number of first coordinated control signal one-to-one N number of second
Coordinated control signal is simultaneously transmitted to the electric system, wherein the N is positive integer, and N > 1.
2. a kind of adaptive wide area damping control according to claim 1, which is characterized in that the adaptive time lag is mended
Repaying device includes: n time lag compensation submodule and additional gain computing unit；
The n time lag compensation submodule, for carrying out adaptive time lag compensation to each input signal；
The additional gain computing unit calculates the input signal pair for the communication delay according to each input signal
The additional gain value answered；
Wherein, the input signal is first wide area measurement system signal or the first coordinated control signal；
Then transmission function of the adaptive time lag compensation device to each input signal are as follows:
I is 1,2,3....., n
Wherein, K indicates the additional gain value of the input signal, SDCi(s) i-th of time lag compensation submodule is indicated
The transmission function of block, βi(τ) is expressed as the weight of i-th of time lag compensation submodule, and n is the time lag compensation submodule
Number, value are that the order of input signal time lag transmission function after simplifying adds 1.
3. a kind of adaptive wide area damping control according to claim 2, which is characterized in that input letter after the simplification
Number time lag transmission function are as follows:
Wherein, s indicates that complex variable, τ are expressed as the communication delay.
4. a kind of adaptive wide area damping control according to claim 3, which is characterized in that the additional gain value
Calculation formula are as follows:
Wherein, τ indicates the communication delay.
5. a kind of adaptive wide area damping control according to any one of claims 1 to 4, which is characterized in that the N is defeated
Entering to export GrHDP unit includes: the first GrHDP neural network and its first weight；
The first GrHDP neural network be multiple groups operating condition of the 2nd GrHDP neural network based on the electric system and
Second weight of fault condition information and the 2nd GrHDP neural network is trained to obtain；
Then the N input and output GrHDP unit is specifically used for:
N number of second wide area measurement system signal is received, and based on GrHDP model algorithm to the first GrHDP neural network
First weight is adjusted, and is calculated and is coordinated to control with N number of second wide area measurement system signal one-to-one N number of first
Signal processed simultaneously exports.
6. a kind of adaptive wide area damping control according to claim 5, which is characterized in that the 2nd GrHDP mind
It include executing network, target network and evaluation network through network, wherein described to execute network, target network and evaluate network
The order of weight matrix is determined according to the N respectively, so that described in obtaining after the 2nd GrHDP neural metwork training
First GrHDP neural network receives N number of first coordinated control signal and calculates described in output through the GrHDP model algorithm
N number of second coordinated control signal.
7. a kind of adaptive wide area damping control according to claim 5, which is characterized in that the N is 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910086699.0A CN109742773B (en) | 2019-01-29 | 2019-01-29 | Self-adaptive wide area damping controller |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910086699.0A CN109742773B (en) | 2019-01-29 | 2019-01-29 | Self-adaptive wide area damping controller |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109742773A true CN109742773A (en) | 2019-05-10 |
| CN109742773B CN109742773B (en) | 2020-07-28 |
Family
ID=66366640
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910086699.0A Active CN109742773B (en) | 2019-01-29 | 2019-01-29 | Self-adaptive wide area damping controller |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109742773B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112338914A (en) * | 2020-10-27 | 2021-02-09 | 东北大学 | Single-link manipulator fuzzy control algorithm based on random system under output limitation and input hysteresis |
| CN112836818A (en) * | 2021-01-14 | 2021-05-25 | 西安热工研究院有限公司 | Dynamic response artificial neural network suitable for process control modeling |
| WO2021184753A1 (en) * | 2020-03-16 | 2021-09-23 | 西安建筑科技大学 | Self-adaptive real-time hybrid simulation test loading method for space framework model |
| US11150159B2 (en) | 2020-03-16 | 2021-10-19 | Xi'an University Of Architecture And Technology | Adaptive loading method for real-time hybrid simulation testing of space frame model |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107346892A (en) * | 2017-04-24 | 2017-11-14 | 国家电网公司 | PSS design methods based on the input hierarchical control of wide area multi signal |
| CN108365615A (en) * | 2018-02-08 | 2018-08-03 | 华中科技大学 | A kind of adaptive wide area damping control and control method |
-
2019
- 2019-01-29 CN CN201910086699.0A patent/CN109742773B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107346892A (en) * | 2017-04-24 | 2017-11-14 | 国家电网公司 | PSS design methods based on the input hierarchical control of wide area multi signal |
| CN108365615A (en) * | 2018-02-08 | 2018-08-03 | 华中科技大学 | A kind of adaptive wide area damping control and control method |
Non-Patent Citations (1)
| Title |
|---|
| 曾令康等: "基于数据驱动的背靠背柔性直流自适应广域阻尼控制", 《中国电机工程学报》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021184753A1 (en) * | 2020-03-16 | 2021-09-23 | 西安建筑科技大学 | Self-adaptive real-time hybrid simulation test loading method for space framework model |
| US11150159B2 (en) | 2020-03-16 | 2021-10-19 | Xi'an University Of Architecture And Technology | Adaptive loading method for real-time hybrid simulation testing of space frame model |
| CN112338914A (en) * | 2020-10-27 | 2021-02-09 | 东北大学 | Single-link manipulator fuzzy control algorithm based on random system under output limitation and input hysteresis |
| CN112338914B (en) * | 2020-10-27 | 2022-03-04 | 东北大学 | Single-link manipulator fuzzy control algorithm based on random system under output limitation and input hysteresis |
| CN112836818A (en) * | 2021-01-14 | 2021-05-25 | 西安热工研究院有限公司 | Dynamic response artificial neural network suitable for process control modeling |
| CN112836818B (en) * | 2021-01-14 | 2023-07-25 | 西安热工研究院有限公司 | Dynamic response artificial neural network suitable for process control modeling |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109742773B (en) | 2020-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109742773A (en) | A kind of adaptive wide area damping control | |
| Lu et al. | Direct heuristic dynamic programming for damping oscillations in a large power system | |
| Rodríguez-Amenedo et al. | Damping low-frequency oscillations in power systems using grid-forming converters | |
| CN106340907A (en) | Power system security and stability control strategy determining method and device | |
| CN108365615B (en) | Self-adaptive wide area damping controller and control method | |
| CN108923460A (en) | The method for parameter configuration that microgrid virtual synchronous machine multi-machine parallel connection dynamic unanimously responds | |
| Zacharia et al. | Integration of renewables into the wide area control scheme for damping power oscillations | |
| Fu et al. | Power oscillation suppression in multi-VSG grid by adaptive virtual impedance control | |
| CN104362653A (en) | Power system stabilizer | |
| CN105337290A (en) | Reactive adjustment method applicable to low-frequency oscillation aid decision of electric system | |
| Dong et al. | A time-delay correction control strategy for HVDC frequency regulation service | |
| CN113394828B (en) | Improved VSG parallel control method based on secondary voltage regulation | |
| CN107171335B (en) | Wind power plant voltage coordination control method based on local reactive power regulation | |
| CN109755968A (en) | A kind of neural network guaranteed cost virtual synchronous control method of double-fed fan motor unit | |
| Morovati et al. | Control coordination between DFIG-based wind turbines and synchronous generators for optimal primary frequency response | |
| Park et al. | New external neuro-controller for series capacitive reactance compensator in a power network | |
| Sanz et al. | Coordinated corrective control for transient stability enhancement in future Great Britain transmission system | |
| Liang et al. | Adaptive critic design based dynamic optimal power flow controller for a smart grid | |
| Li et al. | EEOP-Based Interaction Path Differentiation and Control Parameter Optimization to Mitigate Oscillation in Multi-DFIG Wind Farm | |
| Zhang et al. | Adaptive inter-area power oscillation damping from offshore wind farm and MMC-HVDC using deep reinforcement learning | |
| CN115207988B (en) | Control method and system for new energy power generation converter | |
| CN111525617A (en) | VSG power decoupling control method and system based on excitation regulation | |
| Kumar et al. | A robust damping control for battery energy storage integrated power systems to mitigate inter-area oscillations | |
| CN114421531A (en) | Active power distribution network distributed voltage control method considering global sensitivity | |
| Barros et al. | An optimal control strategy for DFIG |
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 |