CN106532691B - Single regional power system frequency multiplexed control method based on adaptive Dynamic Programming - Google Patents
Single regional power system frequency multiplexed control method based on adaptive Dynamic Programming Download PDFInfo
- Publication number
- CN106532691B CN106532691B CN201611104714.2A CN201611104714A CN106532691B CN 106532691 B CN106532691 B CN 106532691B CN 201611104714 A CN201611104714 A CN 201611104714A CN 106532691 B CN106532691 B CN 106532691B
- Authority
- CN
- China
- Prior art keywords
- network
- behavior
- weight
- power system
- indicate
- 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
Links
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 83
- 238000000034 method Methods 0.000 title claims abstract description 36
- 239000002131 composite material Substances 0.000 claims abstract description 26
- 230000006399 behavior Effects 0.000 claims description 126
- 238000011156 evaluation Methods 0.000 claims description 85
- 210000002569 neuron Anatomy 0.000 claims description 83
- 230000006870 function Effects 0.000 claims description 31
- 238000012549 training Methods 0.000 claims description 17
- 210000004205 output neuron Anatomy 0.000 claims description 11
- 238000004088 simulation Methods 0.000 claims description 11
- 238000012546 transfer Methods 0.000 claims description 11
- 210000002364 input neuron Anatomy 0.000 claims description 9
- 230000017105 transposition Effects 0.000 claims description 9
- 230000010354 integration Effects 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 6
- 210000004027 cell Anatomy 0.000 claims description 4
- 210000005036 nerve Anatomy 0.000 claims description 4
- 230000001105 regulatory effect Effects 0.000 claims description 4
- 238000005070 sampling Methods 0.000 claims description 4
- 150000001875 compounds Chemical class 0.000 claims description 3
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000001276 controlling effect Effects 0.000 claims description 2
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 230000005540 biological transmission Effects 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 230000008859 change Effects 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 12
- 230000000694 effects Effects 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 6
- 230000005611 electricity Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000010248 power generation Methods 0.000 description 2
- 230000008929 regeneration Effects 0.000 description 2
- 238000011069 regeneration method Methods 0.000 description 2
- 230000008054 signal transmission Effects 0.000 description 2
- 230000006641 stabilisation Effects 0.000 description 2
- 238000011105 stabilization Methods 0.000 description 2
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 1
- 240000002853 Nelumbo nucifera Species 0.000 description 1
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001311 chemical methods and process Methods 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000009415 formwork Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
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
-
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Feedback Control In General (AREA)
Abstract
Single regional power system frequency multiplexed control method based on adaptive Dynamic Programming that the present invention relates to a kind of.Wherein, this method comprises: obtaining following measuring signal: single regional power system governor time constant, generator time constant, region duration of load application constant, region load gain, single regional power system frequency departure;Then proportion of utilization integral controller and adaptive dynamics programming control device generate proportional plus integral control signal and adaptive dynamics programming control signal according to measuring signal, and the two is superimposed, obtain composite signals;Composite signals are applied in single regional power system again, carry out frequency control.The present invention is by the way that PI control amount to be added with adaptive dynamics programming control amount, constitute complex controll amount, single regional power system comprising governor, generator and local load is solved under random load situation of change, system actual frequency deviates the technical issues of nominal value, eliminates frequency departure.
Description
Technical field
The present invention relates to power system frequency control technology field, more particularly, to a kind of based on adaptive Dynamic Programming
Single regional power system frequency multiplexed control method.
Background technique
Large-scale power system is usually made of multiple regions, and each region is connected with each other by interconnection and adjacent area.
With electrical energy demands growth and renewable energy power generation technology it is increasingly mature, modern power systems gradually develop for incorporate point
The smart grid of cloth power generation and random load, workload demand and generated energy become more and more random, and electrical component is also increasingly
It is abundant.
When electric system scale is smaller as, entire electric system can be regarded to a region, i.e., single regional power system.
When single regional power system encounters the variation of multiple random loads, the frequency of system may occur seriously to vibrate, and at this moment using has
The power system frequency control method of effect maintains power system frequency stabilization to become to be even more important.In conjunction with single regional power system
The characteristics of, some control methods gradually attract attention.For example, surrounding proportional integral differential (Proportion-
Integration-differentiation, PID) control, internal model control, sliding formwork control, the methods of fuzzy logic control,
Through the research for having carried out some power system load frequency controls.The application of these methods, from improving electric power in varying degrees
The performance of system frequency control, but still can be improved, because these methods lack on-line study and adaptive ability.
Therefore, under random load situation of change, guarantee the safety and frequency stabilization of electric system, have become intelligent electricity
Net a significant challenge of development.
Summary of the invention
In order to solve the above problem in the prior art, change feelings in order to solve single regional power system in random load
Under condition, system actual frequency deviate nominal value the technical issues of and a kind of single region electric power based on adaptive Dynamic Programming is provided
System frequency composite control method.
To achieve the goals above, following technical scheme is provided:
A kind of single regional power system frequency multiplexed control method based on adaptive Dynamic Programming, which comprises
Obtain following measuring signal: single regional power system governor time constant, generator time constant, region load
Time constant, region load gain, single regional power system frequency departure;
Proportion of utilization integral controller and adaptive dynamics programming control device generate proportional integration according to the measuring signal
Signal and adaptive dynamics programming control signal are controlled, and the two is superimposed, obtains composite signals;
The composite signals are applied in single regional power system, frequency control is carried out.
Preferably, the generation proportional plus integral control signal specifically includes:
Determine the transmitting of governor equivalent model, generator equivalent model and local load equivalent model respectively according to the following formula
Function model:
Wherein, the Gg(s) transfer function model of the governor equivalent model is indicated;The s indicates to draw general
Complex variable in the transformation of Lars;The TgIndicate governor time constant;The Gt(s) the generator equivalent model is indicated
The transfer function model;The TtIndicate generator time constant;The Gp(s) the local load equivalent model is indicated
The transfer function model;The kpIndicate the gain of region load;The TpIndicate region duration of load application constant;
The equivalent model based on the governor, the generator and the local load, establishes single region electric power
System simulation model;
According to single regional power system simulation model, the parameter of pi controller is obtained using examination method is gathered;
Parameter based on single regional power system simulation model and the pi controller, with single region
Power system frequency deviation is regulated variable, calculates the proportional plus integral control signal according to the following formula:
Wherein, the u1(t) the proportional plus integral control signal is indicated;The KpIndicate proportionality coefficient;Δ f (t) table
Show single regional power system frequency departure;The KiIndicate integral coefficient;The τ indicates integration variable;When the t is indicated
Between variable.
Preferably, the adaptive dynamics programming control signal specifically includes:
The input signal of behavior network is determined according to the following formula:
mf=max | Δ f (t) |, | Δ f (t- Δ t) | }
Wherein, the xa(t) input signal of the behavior network is indicated;The t indicates current time;The t- Δ t
Indicate the One-step delay at current time;The mfIndicate normalization coefficient;It is described | Δ f (t) | indicate the absolute of the Δ f (t)
Value;The Δ f (t) indicates the corresponding single regional power system frequency departure of t moment;It is described | and Δ f (t- Δ t) | indicate the Δ
F (the absolute value of t- Δ t);(t- Δ t) indicates the corresponding single regional power system frequency departure of the t- time Δt to the Δ f;
The adaptive dynamics programming control signal is calculated according to the following formula:
Wherein, paj(t) input of j-th of hidden neuron of behavior network is indicated;xai(t) indicate i-th of behavior network it is defeated
Enter the input of neuron,wa1,ij(t) indicate i-th of input neuron of behavior network to j-th of hidden neuron
Weight, Indicate hidden neuron number;Expression behavior network input layer neuron number;qaj
(t) output of j-th of hidden neuron of behavior network is indicated;σaThe activation primitive of expression behavior network;vak(t) behavior net is indicated
The input of k-th of output neuron of network, Indicate output layer neuron number;wa2,jk(t) behavior net is indicated
Weight of j-th of the hidden neuron of network to k-th of output neuron;u2k(t) the defeated of k-th of output neuron of behavior network is indicated
Out.
Preferably, the adaptive dynamics programming control signal is specific further include:
The minimum cost of the adaptive dynamics programming control device is determined using evaluation network;
Update the weight of the behavior network;
Update the weight of the evaluation network.
Preferably, the minimum cost that the adaptive dynamics programming control device is determined using evaluation network is specific to wrap
It includes:
The input signal of the evaluation network is determined according to the following formula:
Wherein, the ci(t) input signal of the evaluation network is indicated;It is describedIndicate the behavior network
The transposition of input signal;It is describedIndicate the transposition of the adaptive dynamics programming control signal;
The minimum cost is calculated according to the input signal of the evaluation network:
Wherein, the pcj(t) input of described evaluation j-th of hidden neuron of network is indicated, it is describedInstitute
It statesIndicate hidden neuron number;The xai(t) input of i-th of the behavior network input neuron is indicated, it is describedIt is describedIt indicates to judge network input layer neuron number,
The wC1, ij(t) weight of i-th of the network input neuron of the expression evaluation to j-th of hidden neuron;The qcj(t)
Indicate the output of described evaluation j-th of hidden neuron of network;The σcIndicate the activation primitive of the evaluation network;It is describedIndicate the minimum cost of t moment;The wc2,j(t) indicate j-th of hidden neuron of the evaluation network to output nerve
The weight of member.
Preferably, the weight for updating the behavior network specifically includes:
Behavior network error is determined according to the following formula:
Wherein, the Eat(l) the behavior network error is indicated;The l indicates the behavior network weight in t moment
Inner iterative number;It is describedIndicate the minimum cost of t moment;
Weight Training permissible error and maximum number of iterations are set, and carry out inner iterative according to the following formula, updates the row
For network weight:
Wherein, describedIt is describedIt is describedWith it is describedTable respectively
Show the weight gradient of the l times iteration of t moment behavior network hidden neuron and input layer;It is describedAnd institute
It statesRespectively indicate the weight of the l+1 times iteration of t moment behavior network hidden neuron and input layer;λaTable
Show the learning rate of behavior network;
When the behavior network error meets first error threshold value or the behavior network internal the number of iterations reaches first
When frequency threshold value, stops the behavior network internal iteration, obtain behavior network hidden neuron and input layer weight.
Preferably, the weight for updating the evaluation network specifically includes:
Utility function is calculated according to the following formula:
Wherein, the R (t) indicates the utility function;The Q=diag (1,0.5);The xa(t) row is indicated
For the input signal of network;Indicate the transposition of the behavior network input signal;
Based on the utility function, setting evaluation network weight training permissible error and maximum number of iterations, according to the following formula
Determine evaluation network error:
Wherein, describedIt indicates to theThe evaluation network error of secondary iteration;It is describedIndicate the evaluation network weight
In the inner iterative number of t moment;It is describedIndicate the minimum cost of t moment;It is describedIndicate t- time Δt most
Small cost;The γ indicates discount factor;The Δ t indicates the sampling time;
Using gradient descent method, inner iterative is carried out according to the following formula, updates the evaluation network weight:
Wherein, describedIt is describedIt is describedWith it is describedRespectively
Indicate t moment evaluation network hidden neuron and input layer theThe weight gradient of secondary iteration;It is described
With it is describedIndicate t moment evaluation network hidden neuron and input layer theThe weight of secondary iteration;
The λcIndicate the learning rate of evaluation network;
Reach second number when the evaluation network error meets the second error threshold or evaluates network internal the number of iterations
When threshold value, stop the evaluation network internal iteration, obtains evaluation network hidden neuron and input layer weight.
Preferably, described that the composite signals are applied in single regional power system, frequency control is carried out,
It specifically includes:
Step 1: carving t at the beginning0, single regional power system manages and controls cell S MMC and receives frequency deviation f
(t0), calculate the t0The proportional plus integral control signal at moment, and determine the input of behavior networkWherein, the mf=1.2 | Δ f (t0) |, calculate the t0The adaptive Dynamic Programming at moment
Signal is controlled, then the proportional plus integral control signal and the adaptive dynamics programming control signal are sent to single region
Each participation unit in electric system;
Step 2: the random initializtion behavior network weight w on [0,1] sectionA2, jk(t0) and wa1,ij(t0) and evaluation net
Network weight wc2,jk(t0) and wC1, ij(t0);
Step 3: initializing adaptive Dynamic Programming parameter: cost function target value, the input layer of behavior network
Number, hidden neuron number, output layer neuron number judge input layer number, the hidden nodes of network
Mesh, output layer neuron number, behavior e-learning rate judge e-learning rate, behavior network weight training permissible error, power
It is worth training maximum number of iterations, judges network weight training permissible error and maximum number of iterations;
Step 4: receiving the frequency deviation f (t) in moment t, the SMMC, calculate the proportional plus integral control letter
Number;After data prediction, the input x of the behavior network is obtaineda(t), using the behavior network weight wA2, jk(t) and
wa1,ij(t), the adaptive dynamics programming control signal is calculated;Use the evaluation network weight wC2, jk(t) and wC1, ij
(t) minimum cost is exported;
Step 5: calculating behavior network error, utility function and evaluation network error, update the behavior network weight respectively
Value wa2,jk(t+ Δ t) and wa1,ij(t+ Δ t) and the evaluation network weight wC2, jk(t+ Δ t) and wc1,ij(t+ Δ t), and
New weight is used when next time step;
Step 6: the proportional plus integral control signal and the adaptive dynamics programming control signal being overlapped, obtained
Composite signals are simultaneously sent to each participation unit in single regional power system by the SMMC, and enter next
A time step t+ Δ t, repeats step 4 to step 6.
The contemplated technical solution of the present invention compared with prior art, has the advantages that
The present invention provides a kind of single regional power system frequency multiplexed control method based on adaptive Dynamic Programming.Its
In, this method comprises: obtaining following measuring signal: single regional power system governor time constant, generator time constant, area
Domain duration of load application constant, region load gain, single regional power system frequency departure;Proportion of utilization integral controller and adaptive
Then Dynamic Programming controller generates proportional plus integral control signal and adaptive dynamics programming control signal according to measuring signal,
And be superimposed the two, obtain composite signals;Composite signals are applied in single regional power system again, carry out frequency
Control.The present invention constitutes complex controll amount by the way that PI control amount to be added with adaptive dynamics programming control amount, solves and includes
Under random load situation of change, system actual frequency deviates single regional power system of governor, generator and local load
The technical issues of nominal value, realizes quickly and effectively single regional power system frequency and adjusts, realizes adaptive frequency control,
Frequency departure (frequency fluctuation) is eliminated, is the application demand and development trend for meeting intelligent power grid technology.
Detailed description of the invention
Fig. 1 is single regional power system frequency multiplexed control according to an embodiment of the present invention based on adaptive Dynamic Programming
The flow diagram of method;
Fig. 2 is the structure and signal transmission schematic diagram of single regional power system equivalent model according to an embodiment of the present invention;
Fig. 3 is adaptive dynamics programming control device schematic illustration according to an embodiment of the present invention;
Fig. 4 is according to an embodiment of the present invention by single regional power system frequency multiplexed control based on adaptive Dynamic Programming
Method processed is applied to carry out the schematic diagram of frequency control in single regional power system;
Fig. 5 is the random load variable signal schematic diagram of single regional power system according to an embodiment of the present invention;
Fig. 6 is single regional power system according to an embodiment of the present invention under random load variation, using proportional integration control
The control effect comparison schematic diagram of the frequency departure for the composite controller that device processed and the embodiment of the present invention are mentioned;
Fig. 7 is compound control of the single regional power system according to an embodiment of the present invention in random load variation lower frequency deviation
Amount processed and PI control amount comparison schematic diagram;
Fig. 8 is composite controller under the random load interference according to an embodiment of the present invention in 10s, 30s, 50s, 70s
In adaptive dynamics programming control amount schematic diagram.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, this hair is described with reference to the accompanying drawings
Bright preferred embodiment.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to limit
The fixed present invention.As long as in addition, technical characteristic involved in invention described below embodiment non-structure each other
It can be combined with each other at conflict.
With the appearance of intelligent control technology, adaptive Dynamic Programming (Adaptive dynamic programming,
ADP) method is applied in multiple industrial circles, such as robot, aircraft, chemical process and smart grid.This side
The on-line study ability for being a major advantage that it of method makes controlled device have adaptive ability in disturbed situation.It will be this
Adaptive control technology is applied in the LOAD FREQUENCY control (Load frequency control, LFC) of electric system, solves
The problems such as modern power network random load changes, has a very important significance.
The basic thought of the embodiment of the present invention is based on single regional power system frequency control problem, with proportional plus integral control
Device is basic controller, and multiple random loads are changed, and keeps PI controller parameter constant, with adaptive dynamics programming control
Device be upper controller Compound Control Strategy, for multiple random loads change, according to electric system current frequency deviation into
The adjustment of row online adaptive, it is online to carry out adaptive dynamics programming control device right value update, it obtains corresponding adaptive dynamic and advises
Control amount is drawn, power system frequency is made to return to specified value.
For this purpose, the embodiment of the present invention provides a kind of single regional power system frequency multiplexed control based on adaptive Dynamic Programming
Method processed.As shown in Figure 1, this method can be realized by step S100 to step S120.Wherein:
S100: following measuring signal is obtained: single regional power system governor time constant, generator time constant, area
Domain duration of load application constant, region load gain, single regional power system frequency departure.
This step can measure all signals of single regional power system by distributed sensor, then by communication channel
It is transferred to micro-capacitance sensor and manages and controls system (Smart Micro-Grid Management and Control, SMMC).So
Afterwards, it is handled by SMMC, generates control signal, then send into single regional power system in each participation unit by communication channel.
Single regional power system includes generator, governor and single region locality load.Wherein, generator provides electric energy and supplies
It gives;Governor controls the speed of generator, prevents generator from damaging;Local load is the Demand-side of the electric system, consumption electricity
Energy.
S110: proportion of utilization integral controller and adaptive dynamics programming control device generate ratio according to above-mentioned measuring signal
Example integral control signal and adaptive dynamics programming control signal, and the two is superimposed, obtain composite signals.
This step is basic controller with pi controller (PI controller), is with adaptive dynamics programming control device
Upper controller.
Wherein, the step of generating PI (proportional integration) control signal can specifically include:
S111: the transfer function model of governor, generator and local load equivalent model is determined respectively according to the following formula:
Wherein, Gg(s) transfer function model of governor equivalent model is indicated;S indicates the plural number in Laplace transform
Variable;TgIndicate governor time constant;Gt(s) transfer function model of generator equivalent model is indicated;TtWhen indicating generator
Between constant;Gp(s) transfer function model of local load equivalent model is indicated;kpIndicate the gain of region load;TpIndicate that region is negative
Lotus time constant.
S112: the equivalent model based on governor, generator and local load establishes single regional power system and emulates mould
Type.
Specifically, this step can further include:
SA1: according to single regional power system frequency departure and its integral parameter, state vector is determined.
As an example, state vector x=[the Δ f (t), Δ P of single regional power systemt(t),ΔXg(t),ΔE(t)]T。
SA2: being based on state vector, establishes single regional power system simulation model according to the following formula:
Wherein,Indicate Δ f (t) to the first differential of time, i.e., Indicate single region electricity
Force system frequency departure;Δ d (t) indicates the load disturbance that single regional power system random load variation generates;ΔPt(t) it indicates
Single regional power system generated output power variable quantity;Indicate Δ Pt(t) to the first differential of time, i.e.,ΔXg(t) single regional power system adjuster position deviation value is indicated;Indicate Δ Xg(t) clock synchronization
Between first differential, i.e.,R indicates the equivalent impedance of single regional power system;Δ E (t) indicates region control
Deviation processed, i.e., the integral of single regional power system frequency departure, Indicate that Δ E (t) is right
The first differential of time, i.e.,U (t) indicates control signal;keIndicate integration gain factor.
Wherein, consumption electric energy is equivalent to as Δ d (t) > 0 to increase;It is equivalent to as Δ d (t) < 0 to power grid and inputs electricity
Energy.
Fig. 2 schematically illustrates the structure and signal transmission figure of single regional power system equivalent model.
S113: according to single regional power system simulation model, the parameter of pi controller is obtained using examination method is gathered.
Specifically, this step may include:
Step a1: the proportionality coefficient of adjuster and the value of integral coefficient are determined.
Step a2: one disturbance is added to system by changing given value, observes curve shape.
Step a3: the value by changing proportionality coefficient or integral coefficient gathers examination until controlled volume meets dynamic process repeatedly
Until quality requirements.
Step a4: retaining the finally obtained proportionality coefficient of step a3 and integral coefficient is the parameter of PI controller.
S114: the parameter based on single regional power system simulation model and pi controller, with single region power train
System frequency departure is regulated variable, calculates ratio integral control signal according to the following formula:
Wherein, u1(t) proportional plus integral control signal is indicated;KpIndicate proportionality coefficient;Δ f (t) indicates single regional power system
Frequency departure;KiIndicate integral coefficient;τ indicates integration variable;T indicates time variable, such as moment.
Wherein, the step of generating adaptive dynamics programming control signal can specifically include:
S115: the input signal of behavior network is determined according to the following formula:
Wherein, xa(t) input signal of behavior network is indicated;T indicates current time;T- Δ t indicates the one of current time
Step delay;mfIndicate normalization coefficient, mf=max | Δ f (t) |, | Δ f (t- Δ t) | };| Δ f (t) | indicate the exhausted of Δ f (t)
To value;Δ f (t) indicates the corresponding single regional power system frequency departure of t moment;| Δ f (t- Δ t) | expression Δ f be (t- Δ t's)
Absolute value;(t- Δ t) indicates the corresponding single regional power system frequency departure of t- time Δt to Δ f.
Preferably, behavior network is realized using Multilayer perceptron network.
The parameter of behavior network can be carried out following initial: be set as cost function target value UC=0, behavior network is defeated
Enter a layer neuron number, hidden neuron number, output layer neuron number, behavior e-learning rate, behavior network weight instruction
Practice permissible error and Weight Training maximum number of iterations.
S116: adaptive dynamics programming control signal is calculated according to the following formula:
Wherein, paj(t) input of j-th of hidden neuron of behavior network is indicated;xai(t) indicate i-th of behavior network it is defeated
Enter the input of neuron,wa1,ij(t) indicate i-th of input neuron of behavior network to j-th of hidden layer nerve
The weight of member, Indicate hidden neuron number;Expression behavior network input layer neuron number;qaj
(t) output of j-th of hidden neuron of behavior network is indicated;σaThe activation primitive of expression behavior network;vak(t) behavior net is indicated
The input of k-th of output neuron of network, Indicate output layer neuron number;wa2,jk(t) behavior net is indicated
Weight of j-th of the hidden neuron of network to k-th of output neuron;u2k(t) the defeated of k-th of output neuron of behavior network is indicated
Out.
Due to single regional power system adaptive dynamics programming control device output be it is one-dimensional, i.e., behavior network output mind
It is 1 through first number, i.e. k=1.Therefore, adaptive dynamics programming control signal u2(t)=u21(t).Certainly using the output of behavior network
It adapts to Dynamic Programming and controls signal.
On the basis of the above embodiments, method provided in an embodiment of the present invention can also include:
S117: the minimum cost of adaptive dynamics programming control device is determined using evaluation network.
Preferably, evaluation network is realized using Multilayer perceptron network.
Fig. 3 schematically illustrates adaptive dynamics programming control device schematic diagram.
Can be initialized as follows to evaluation network: network input layer neuron number, hidden neuron are judged in setting
Number, output layer neuron number judge e-learning rate, judge network weight training permissible error and maximum number of iterations.
Specifically, this step may include:
SB1: the input signal of evaluation network is determined according to the following formula:
Wherein, ci(t) input signal of evaluation network is indicated;The transposition of expression behavior network input signal;
Indicate the transposition of adaptive dynamics programming control signal.
The input of above-mentioned evaluation network isWithThe vector of composition.
SB2: minimum cost is calculated according to the input signal of evaluation network:
Wherein, pcj(t) input of evaluation j-th of hidden neuron of network is indicated, Indicate hidden layer mind
Through first number;xai(t) input of i-th of behavior network input neuron is indicated, It indicates to judge network inputs
Layer neuron number, wC1, ij(t) evaluation i-th of input neuron of network is indicated
To the weight of j-th of hidden neuron;qcj(t) output of evaluation j-th of hidden neuron of network is indicated;σcIndicate evaluation network
Activation primitive;Indicate that the minimum cost of t moment, the minimum cost can be output valve of the evaluation network in t moment;
wc2,j(t) weight of expression evaluation j-th of the hidden neuron of network to output neuron.
Multiple random loads are changed, keep PI controller parameter constant.It is upper with adaptive dynamics programming control device
Layer controller, changes multiple random loads, online to carry out adaptive dynamics programming control device power according to system frequency deviation
Value updates, and obtains corresponding adaptive dynamics programming control amount.
S118: the weight w of regeneration behavior networka1,ij(t) and wa2,jk(t)。
Step b1: behavior network error is determined according to the following formula:
Wherein, Eat(l) behavior network error is indicated;L indicates behavior network weight in the inner iterative number of t moment;Indicate the minimum cost of t moment.
Step b2: setting Weight Training permissible error and maximum number of iterations, and inner iterative is carried out according to the following formula, it updates
Behavior network weight:
Wherein, WithRespectively indicate t moment behavior network
The weight gradient of the l times iteration of hidden neuron and input layer;WithRespectively indicate t moment
The weight of the l+1 times iteration of behavior network hidden neuron and input layer;λaThe learning rate of expression behavior network, preferably
Ground, λa> 0.
Wherein,WithAlso illustrate that the behavior network weight w for carrying out inner iterativea1,ij(t) and wa2,jk
(t)。
Step b3: when behavior network error meets first error threshold value or behavior network internal the number of iterations reaches for the first time
When number threshold value, stops behavior network internal iteration, obtain behavior network hidden neuron and input layer weight.
For example, when the weight of acquisitionWithMake behavior network error Eat(l) meet Eat
(l)≤εaWhen, stop behavior network internal iteration, exports weightWithFor t+1 moment behavior network
Hidden neuron and input layer weight, it may be assumed that
Or when the weight obtainedWithBehavior network internal the number of iterations l is set to meet l=Ma
When, stop behavior network internal iteration, exports weightWithFor t+1 moment behavior network hidden layer nerve
Member and input layer weight similarly haveWith
S119: the more weight of New Appraisement network.
Specifically, this step may include:
Step c1: utility function is calculated according to the following formula:
Wherein, R (t) indicates utility function;Q=diag (1,0.5);xa(t) input signal of behavior network is indicated;Table
Show the transposition of behavior network input signal.
Step c2: it is based on utility function, setting evaluation network weight training permissible error and maximum number of iterations, under
Formula determines evaluation network error:
Wherein,It indicates to theThe evaluation network error of secondary iteration;Indicate evaluation network weight in t moment
Portion's the number of iterations;Indicate the minimum cost of t moment;Indicate the minimum cost of t- time Δt;γ indicate discount because
Son, it is preferable that 0 < γ < 1;T indicates the moment;Δ t indicates the sampling time.
Step c3: gradient descent method is used, carries out inner iterative, more New Appraisement network weight according to the following formula:
Wherein, WithRespectively indicate t moment evaluation net
Network hidden neuron and input layerThe weight gradient of secondary iteration;WithWhen indicating t
Carve evaluation network hidden neuron and input layer theThe weight of secondary iteration;λcIndicate the learning rate of evaluation network, it is excellent
Selection of land, λc> 0.
WithAlso the evaluation network weight w for carrying out inner iterative is respectively indicatedc1,ij(t) and wc2,j
(t)。
Step c4: reach second when evaluation network error meets the second error threshold or evaluates network internal the number of iterations
When number threshold value, stops evaluation network internal iteration, obtain evaluation network hidden neuron and input layer weight.
For example, when the weight of acquisitionWithMake to evaluate network errorMeetWhen, stop evaluation network internal iteration, exports weightWithNet is evaluated for the t+1 moment
Network hidden neuron and input layer weight, it may be assumed that
Alternatively, when acquisitionWithMake to evaluate network internal the number of iterationsMeetWhen,
Stop evaluation network internal iteration, exports weightWithNetwork hidden neuron is evaluated for the t+1 moment
Similarly have with input layer weightWith
S120: above-mentioned composite signals are applied in single regional power system, carry out frequency control.
Fig. 4, which is schematically illustrated, to be applied to method provided in an embodiment of the present invention to carry out frequency in single regional power system
The schematic diagram of rate control.
It is applied in single regional power system with a preferred embodiment to by composite signals below with reference to Fig. 4, into
The step of line frequency controls is described in detail.Wherein, which may include:
S121: t is carved at the beginning0, single regional power system manages and controls cell S MMC and receives frequency deviation f (t0),
Calculate t0The proportional plus integral control signal at moment, and determine the input of behavior networkWherein, mf=
1.2|Δf(t0) |, calculate t0The adaptive dynamics programming control signal at moment, then by proportional plus integral control signal and adaptively
Dynamic Programming control signal is sent to each participation unit in single regional power system.
S122: the random initializtion behavior network weight w on [0,1] sectiona2,jk(t0) and wa1,ij(t0), and evaluation net
Network weight wc2,jk(t0) and wC1, ij(t0)。
S123: adaptive Dynamic Programming parameter: cost function target value U is initializedC=0, the input layer mind of behavior network
Through first numberHidden neuron numberOutput layer neuron numberJudge the input layer number of networkHidden neuron numberOutput layer neuron numberBehavior e-learning rate λa, judge e-learning rate λc,
Behavior network weight trains permissible error εa, Weight Training maximum number of iterations Ma, judge network weight training permissible error εcAnd
Maximum number of iterations Mc。
S124: frequency deviation f (t) is received in moment t, SMMC, calculates u1(t);After data prediction, obtain
The input x of behavior networka(t), usage behavior network weight wa2,jk(t) and wa1,ij(t), adaptive Dynamic Programming control is calculated
Signal u processed2(t);In-service evaluation network weight wc2,jk(t) and wc1,ij(t) minimum cost is exported
S125: behavior network error E is calculatedat(l), utility function R (t) and evaluation network errorIt updates respectively
Behavior network weight wa2,jk(t+ Δ t) and wa1,ij(t+ Δ t) and evaluation network weight wC2, jk(t+ Δ t) and wc1,ij(t+Δ
T), and in next time step use new weight.
S126: proportional plus integral control signal and adaptive dynamics programming control signal are overlapped, complex controll is obtained
Signal is simultaneously sent to each participation unit in single regional power system by SMMC, and enters next time step t+ Δ t, repeats
Step S204 to step S206.
In order to enable those skilled in the art to better understand the present invention, below in conjunction with specific embodiment, to single region electric power
System frequency composite control method is described in detail.
S301: for single regional power system, it is arranged as follows: governor time constant Tg=0.1, generator time
Constant Tt=0.3, the time constant T of region loadp=10, region load gain kp=1, circuit impedance r=0.05, integral increase
Beneficial coefficient ke=0.4.
S302: for single regional power system, be added respectively in 10s, 30s, 50s, 70s amplitude be+0.15, -0.3 ,+
0.25, -0.1 random step signal is as random load variable signal.
Random load interference of this step namely in 10s, 30s, 50s, 70s is Δ d=0.15, Δ d=- respectively
0.15, Δ d=0.1, Δ d=0.
Fig. 5 schematically illustrates the random load variable signal figure of single regional power system.
S303: the Proportional coefficient K usedp=10 and integral coefficient Ki=50 design adaptive dynamics programming control devices.
S304: random initializtion behavior network weight wa2,jk(0) and wa1,ij(0) and evaluation network weight wc2,j(0) and
wc1,ij(0), t0=0.
S305: the parameter of setting adaptive dynamics programming control device: sampling time Δ t=0.05s, cost function target value
UC=0, behavior network input layer neuron numberHidden neuron numberOutput layer neuron numberJudge network input layer neuron numberJudge network hidden neuron numberJudge network output
Layer neuron numberBehavior e-learning rate λa=0.05, e-learning rate λ is judgedc=0.05, behavior network weight is instructed
Practice permissible error εa=10-6, Weight Training maximum number of iterations Ma=80, network weight training permissible error ε is judgedc=10-7
And maximum number of iterations Mc=50.
Fig. 6 schematically illustrates single regional power system under random load variation, using pi controller and
The frequency departure for the composite controller (pi controller and adaptive dynamics programming control device) that the embodiment of the present invention is mentioned
Control effect comparison schematic diagram.Wherein dotted line is the control effect using the frequency departure of pi controller;Solid line is
Using the control effect of the frequency departure of composite controller.As it can be seen that frequency departure is adjusted with smaller using composite controller
Overshoot, the speed of frequency departure to 0 is also faster.
Fig. 7 schematically illustrates single regional power system in the complex controll amount of random load variation lower frequency deviation
(PI controls signal and adaptive dynamics programming control signal) and PI control amount comparison schematic diagram.The figure illustrates composite controller
Compared to PI controller, it is capable of providing better control performance.When frequency departure occurs, under the action of composite controller,
Frequency departure reduces comparatively fast, and overshoot is also smaller.
Fig. 8 schematically illustrate the random load in 10s, 30s, 50s, 70s interference under, in composite controller from
Adapt to the schematic diagram of Dynamic Programming control amount.Variation of the control amount according to frequency departure, adaptively provides control amount, so that
Composite controller has better control performance.
S306: the transfer function model of governor, generator and local load equivalent model is determined respectively.
S307: the equivalent model based on governor, generator and local load establishes single regional power system and emulates mould
Type.
S308: according to single regional power system simulation model, the parameter of PI controller is obtained using examination method is gathered.
S309: the parameter based on single regional power system simulation model and PI controller, with single regional power system frequency
Deviation is regulated variable, generates PI and controls signal.
S310: the input signal of behavior network is determined.
S311: adaptive dynamics programming control signal is calculated.
S312: the minimum generation of adaptive dynamics programming control device is determined using evaluation network.
S313: the weight of regeneration behavior network.
S314: the more weight of New Appraisement network.
S315: signal is controlled using the output adaptive Dynamic Programming of behavior network.
S316: controlling signal and adaptive dynamics programming control signal for PI, and the two be superimposed, and obtains complex controll letter
Number.
S317: composite signals are applied in single regional power system, carry out frequency control.
The random load of single regional power system changes, and leads to the appearance of load disturbance Δ d (t), so that system frequency
There is deviation delta f (t) in rate.Due to the uncertainty of load disturbance Δ d (t), the embodiment of the present invention is used based on PI controller
Controller uses adaptive dynamics programming control device as upper controller, generates adaptive dynamics programming control signal u2(t)
Signal u is controlled with PI1(t), and the two is managed and controlled to be added in cell S MMC in single regional power system and generates control letter
Number, then given the control signal into single regional power system in each participation unit by communication channel.It solves comprising adjusting
For single regional power system of fast device, generator and local load under random load situation of change, system actual frequency deviates mark
The technical issues of title value, realizes that quickly and effectively single regional power system frequency is adjusted, and is realized adaptive frequency control, is eliminated
Frequency departure, ensure that the safety of electric system, is the application demand and development trend for meeting intelligent power grid technology.
Although each step is described in the way of above-mentioned precedence in above-described embodiment, this field
Technical staff is appreciated that the effect in order to realize the present embodiment, may not necessarily be according to such order between different steps
It executes, (parallel) simultaneously can execute or is executed with reverse order, these simple variations are all in protection model of the invention
Within enclosing.
So far, it has been combined preferred embodiment shown in the drawings and describes technical solution of the present invention, still, this field
Technical staff is it is easily understood that protection scope of the present invention is expressly not limited to these specific embodiments.It is all in the present invention
Spirit and principle within made any modifications, equivalent replacements, and improvements etc., should be included in protection scope of the present invention it
It is interior.
Claims (6)
1. a kind of single regional power system frequency multiplexed control method based on adaptive Dynamic Programming, which is characterized in that described
Method includes:
Obtain following measuring signal: single regional power system governor time constant, generator time constant, region duration of load application
Constant, region load gain, single regional power system frequency departure;
Proportion of utilization integral controller and adaptive dynamics programming control device generate proportional plus integral control according to the measuring signal
Signal and adaptive dynamics programming control signal, and the two is superimposed, obtain composite signals;
The composite signals are applied in single regional power system, frequency control is carried out;
Wherein, the generation proportional plus integral control signal specifically includes:
Determine the transmission function of governor equivalent model, generator equivalent model and local load equivalent model respectively according to the following formula
Model:
Wherein, the Gg(s) transfer function model of the governor equivalent model is indicated;The s indicates that Laplce becomes
Complex variable in changing;The TgIndicate governor time constant;The Gt(s) the described of the generator equivalent model is indicated
Transfer function model;The TtIndicate generator time constant;The Gp(s) the described of the local load equivalent model is indicated
Transfer function model;The kpIndicate the gain of region load;The TpIndicate region duration of load application constant;
The equivalent model based on the governor, the generator and the local load, establishes single regional power system
Simulation model;
According to single regional power system simulation model, the parameter of pi controller is obtained using examination method is gathered;
Parameter based on single regional power system simulation model and the pi controller, with single region electric power
System frequency deviation is regulated variable, calculates the proportional plus integral control signal according to the following formula:
Wherein, the u1(t) the proportional plus integral control signal is indicated;The KpIndicate proportionality coefficient;The Δ f (t) indicates institute
State single regional power system frequency departure;The KiIndicate integral coefficient;The τ indicates integration variable;The t indicates that the time becomes
Amount;
The adaptive dynamics programming control signal specifically includes:
The input signal of behavior network is determined according to the following formula:
Wherein, the xa(t) input signal of the behavior network is indicated;The t indicates current time;The t- Δ t expression is worked as
The One-step delay at preceding moment;The mfIndicate normalization coefficient;It is described | Δ f (t) | indicate the absolute value of the Δ f (t);It is described
Δ f (t) indicates the corresponding single regional power system frequency departure of t moment;It is described | and Δ f (t- Δ t) | indicate Δ f (the t- Δ
T) absolute value;(t- Δ t) indicates the corresponding single regional power system frequency departure of the t- time Δt to the Δ f;
The adaptive dynamics programming control signal is calculated according to the following formula:
Wherein, the paj(t) input of described j-th of hidden neuron of behavior network is indicated;The xai(t) behavior is indicated
The input of i-th of network input neuron, i=1, the L,The wA1, ij(t) i-th of the behavior network input is indicated
Weight of the neuron to j-th of hidden neuron, j=1, the L,It is describedIndicate hidden neuron number;It is describedExpression behavior network input layer neuron number;The qaj(t) output of j-th of hidden neuron of behavior network is indicated;Institute
State σaThe activation primitive of expression behavior network;The vak(t) input of k-th of output neuron of behavior network is indicated, it is describedIt is describedIndicate output layer neuron number;The wa2,jk(t) j-th of hidden layer nerve of behavior network is indicated
Weight of the member to k-th of output neuron;The u2k(t) output of k-th of output neuron of behavior network is indicated.
2. frequency multiplexed control method according to claim 1, which is characterized in that the adaptive dynamics programming control letter
It is number specific further include:
The minimum cost of the adaptive dynamics programming control device is determined using evaluation network;
Update the weight of the behavior network;
Update the weight of the evaluation network.
3. frequency multiplexed control method according to claim 2, which is characterized in that described in the use evaluation network determines
The minimum cost of adaptive dynamics programming control device, specifically includes:
The input signal of the evaluation network is determined according to the following formula:
Wherein, the ci(t) input signal of the evaluation network is indicated;It is describedIndicate the behavior network inputs
The transposition of signal;It is describedIndicate the transposition of the adaptive dynamics programming control signal;
The minimum cost is calculated according to the input signal of the evaluation network:
Wherein, the pcj(t) input of described evaluation j-th of hidden neuron of network is indicated, it is describedIt is described
Indicate hidden neuron number;The xai(t) input of i-th of the behavior network input neuron is indicated, it is describedIt is describedIt indicates to judge network input layer neuron number,Institute
State wc1,ij(t) weight of i-th of the network input neuron of the expression evaluation to j-th of hidden neuron;The qcj(t) table
Show the output of described evaluation j-th of hidden neuron of network;The σcIndicate the activation primitive of the evaluation network;It is described
Indicate the minimum cost of t moment;The wc2,j(t) indicate j-th of hidden neuron of the evaluation network to output neuron
Weight.
4. frequency multiplexed control method according to claim 2, which is characterized in that the power for updating the behavior network
Value specifically includes:
Behavior network error is determined according to the following formula:
Wherein, the Eat(l) the behavior network error is indicated;The l indicates the behavior network weight in the inside of t moment
The number of iterations;It is describedIndicate the minimum cost of the t moment;
Weight Training permissible error and maximum number of iterations are set, and carry out inner iterative according to the following formula, updates the behavior net
Network weight:
Wherein, describedIt is describedIt is describedWith it is describedWhen respectively indicating t
The weight gradient of the l times iteration of quarter behavior network hidden neuron and input layer;It is describedWith it is describedRespectively indicate the weight of the l+1 times iteration of t moment behavior network hidden neuron and input layer;The λa
Indicate the learning rate of the behavior network;
When the behavior network error meets first error threshold value or the behavior network internal the number of iterations reaches first number
When threshold value, stops the behavior network internal iteration, obtain behavior network hidden neuron and input layer weight.
5. frequency multiplexed control method according to claim 2, which is characterized in that the power for updating the evaluation network
Value specifically includes:
Utility function is calculated according to the following formula:
Wherein, the R (t) indicates the utility function;The Q=diag (1,0.5);The xa(t) the behavior network is indicated
The input signal;Indicate the transposition of the behavior network input signal;
Based on the utility function, setting evaluation network weight training permissible error and maximum number of iterations determine according to the following formula
Evaluate network error:
Wherein, the Ect(l) the evaluation network error to the l times iteration is indicated;The l indicates the evaluation network weight in t
The inner iterative number at moment;It is describedIndicate the minimum cost of the t moment;It is describedIndicate t- time Δt
Minimum cost;The γ indicates discount factor;The Δ t indicates the sampling time;
Using gradient descent method, inner iterative is carried out according to the following formula, updates the evaluation network weight:
Wherein, describedIt is describedIt is describedWith it is describedRespectively indicate t
The weight gradient of moment evaluation the l times iteration of network hidden neuron and input layer;It is describedWith it is describedIndicate the weight of t moment evaluation the l+1 times iteration of network hidden neuron and input layer;The λcTable
Show the learning rate of the evaluation network;
Reach the second frequency threshold value when the evaluation network error meets the second error threshold or evaluates network internal the number of iterations
When, stop the evaluation network internal iteration, obtains evaluation network hidden neuron and input layer weight.
6. frequency multiplexed control method according to claim 1, which is characterized in that described to answer the composite signals
It uses in single regional power system, carries out frequency control, specifically include:
Step 1: carving t at the beginning0, single regional power system manages and controls cell S MMC and receives frequency deviation f (t0), meter
Calculate the t0The proportional plus integral control signal at moment, and determine the input of behavior networkIts
In, the mf=1.2 | Δ f (t0) |, calculate the t0The adaptive dynamics programming control signal at moment, then will be described
Proportional plus integral control signal and the adaptive dynamics programming control signal are sent to each ginseng in single regional power system
With unit;
Step 2: the random initializtion behavior network weight w on [0,1] sectiona2,jk(t0) and wa1,ij(t0) and evaluation network weight
Value wc2,jk(t0) and wc1,ij(t0);
Step 3: initialize adaptive Dynamic Programming parameter: cost function target value, the input layer number of behavior network,
Hidden neuron number, output layer neuron number judge the input layer number of network, hidden neuron number, defeated
Layer neuron number out, behavior e-learning rate judge e-learning rate, behavior network weight training permissible error, weight instruction
Practice maximum number of iterations, judges network weight training permissible error and maximum number of iterations;
Step 4: receiving the frequency deviation f (t) in moment t, the SMMC, calculate the proportional plus integral control signal;
After data prediction, the input x of the behavior network is obtaineda(t), using the behavior network weight wa2,jk(t) and
wa1,ij(t), the adaptive dynamics programming control signal is calculated;Use the evaluation network weight wc2,jk(t) and wc1,ij
(t) approximated cost function is exported;
Step 5: calculating behavior network error, utility function and evaluation network error, update the behavior network weight respectively
wa2,jk(t+ Δ t) and wa1,ij(t+ Δ t) and the evaluation network weight wc2,jk(t+ Δ t) and wc1,ij(t+ Δ t), and under
New weight is used when one time step;
Step 6: the proportional plus integral control signal and the adaptive dynamics programming control signal being overlapped, obtained compound
When controlling each participation unit that signal is simultaneously sent in single regional power system by the SMMC, and entering next
Between step-length t+ Δ t, repeat step 4 to step 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611104714.2A CN106532691B (en) | 2016-12-05 | 2016-12-05 | Single regional power system frequency multiplexed control method based on adaptive Dynamic Programming |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201611104714.2A CN106532691B (en) | 2016-12-05 | 2016-12-05 | Single regional power system frequency multiplexed control method based on adaptive Dynamic Programming |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106532691A CN106532691A (en) | 2017-03-22 |
CN106532691B true CN106532691B (en) | 2019-05-10 |
Family
ID=58354898
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201611104714.2A Active CN106532691B (en) | 2016-12-05 | 2016-12-05 | Single regional power system frequency multiplexed control method based on adaptive Dynamic Programming |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106532691B (en) |
Families Citing this family (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108519957A (en) * | 2018-02-10 | 2018-09-11 | 大连智慧海洋软件有限公司 | A kind of data coordinating method based on acceleration broad sense reduced gradient |
CN108448594B (en) * | 2018-03-05 | 2020-04-07 | 中国科学院自动化研究所 | Frequency self-adaptive learning control method for micro-grid |
CN109066645B (en) * | 2018-08-03 | 2020-07-10 | 国网甘肃省电力公司电力科学研究院 | Load control method and device for direct current power supply network of facility agriculture |
CN109274106A (en) * | 2018-09-16 | 2019-01-25 | 天津大学 | A kind of smart grid control method for frequency based on electric car cluster |
CN109144016A (en) * | 2018-10-11 | 2019-01-04 | 常州信息职业技术学院 | Industry mixes the sensing control System and method for of Sensor Network under a kind of data compression |
CN109591659B (en) * | 2019-01-14 | 2020-04-07 | 吉林大学 | Intelligent learning pure electric vehicle energy management control method |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102545228A (en) * | 2012-02-02 | 2012-07-04 | 上海电力学院 | Design method of dispersing slip form load frequency controller of uncertainty power system |
CN103094928A (en) * | 2013-02-18 | 2013-05-08 | 天津市电力公司 | Identification and control method of power grid impact load disturbance |
-
2016
- 2016-12-05 CN CN201611104714.2A patent/CN106532691B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102545228A (en) * | 2012-02-02 | 2012-07-04 | 上海电力学院 | Design method of dispersing slip form load frequency controller of uncertainty power system |
CN103094928A (en) * | 2013-02-18 | 2013-05-08 | 天津市电力公司 | Identification and control method of power grid impact load disturbance |
Non-Patent Citations (1)
Title |
---|
《基于滑模控制的单域电力系统负荷频率控制》;米阳等;《控制与决策》;20121231;第27卷(第12期);第1881-1884页 * |
Also Published As
Publication number | Publication date |
---|---|
CN106532691A (en) | 2017-03-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106532691B (en) | Single regional power system frequency multiplexed control method based on adaptive Dynamic Programming | |
CN109270842B (en) | Bayesian network-based regional heat supply model prediction control system and method | |
Nguang et al. | H/sub/spl infin//fuzzy output feedback control design for nonlinear systems: an LMI approach | |
CN106786677B (en) | A kind of interconnected electric power system distributed dynamic matrix frequency control method | |
CN110429652B (en) | Intelligent power generation control method capable of expanding deep width self-adaptive dynamic planning | |
CN108153259A (en) | A kind of multi-controller optimal State Estimation control strategy design method based on Kalman filtering | |
CN104199302B (en) | Molding system and method of pump storage group speed regulating system | |
CN103226348B (en) | A kind of edible fungi mushroom house set of circumstances medium-long range controls system and method | |
CN104037776B (en) | The electric network reactive-load capacity collocation method of random inertial factor particle swarm optimization algorithm | |
CN109217306A (en) | A kind of intelligent power generation control method based on the deeply study with movement from optimizing ability | |
CN104533701A (en) | Automatic setting method for control parameters of water turbine speed regulating system | |
CN108181947A (en) | A kind of user side load responding method based on the regulation and control of load cluster temperature | |
CN103971160A (en) | Particle swarm optimization method based on complex network | |
CN103455716A (en) | Ultra-short-term wind power prediction-based voltage stability margin calculation method for power system | |
Wang et al. | Developing nonlinear adaptive optimal regulators through an improved neural learning mechanism | |
CN105974795A (en) | Model prediction control method for inhibiting low-frequency oscillation of electric power system based on controllable reactor | |
CN106816877A (en) | A kind of distribution network voltage containing photovoltaic falls detection compensation method | |
CN107834593A (en) | A kind of droop control type isolated island micro-capacitance sensor static voltage stability probability evaluation method of failure | |
Burns et al. | Optimal sensor design for estimation and optimization of PDE systems | |
CN104527637A (en) | Method and system for controlling hybrid electric vehicle | |
CN101635553A (en) | Integrated intelligent decoupling control method for generator excitation system and steam valve system | |
CN109103946A (en) | Wind-powered electricity generation is through flexible direct current power grid transmitting system capacitor group switching scheduling method | |
CN110474323B (en) | Method for measuring inertia time constant of power system | |
CN114039366B (en) | Power grid secondary frequency modulation control method and device based on peacock optimization algorithm | |
Shuang et al. | Taboo search algorithm based ANN model for wind speed prediction |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | 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 |