CN107292764A - Alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value - Google Patents
Alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value Download PDFInfo
- Publication number
- CN107292764A CN107292764A CN201710333433.2A CN201710333433A CN107292764A CN 107292764 A CN107292764 A CN 107292764A CN 201710333433 A CN201710333433 A CN 201710333433A CN 107292764 A CN107292764 A CN 107292764A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msup
- alternating current
- current filter
- mtr
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 36
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 19
- 239000002245 particle Substances 0.000 claims abstract description 75
- 238000003062 neural network model Methods 0.000 claims abstract description 47
- 238000005457 optimization Methods 0.000 claims abstract description 25
- 238000012549 training Methods 0.000 claims abstract description 25
- 230000001373 regressive effect Effects 0.000 claims abstract description 10
- 230000009467 reduction Effects 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 30
- 210000002569 neuron Anatomy 0.000 claims description 12
- 238000013528 artificial neural network Methods 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 9
- 230000006978 adaptation Effects 0.000 claims description 8
- 230000005540 biological transmission Effects 0.000 claims description 8
- 238000012886 linear function Methods 0.000 claims description 4
- 230000001537 neural effect Effects 0.000 claims description 4
- 230000003111 delayed effect Effects 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims 1
- 230000001052 transient effect Effects 0.000 abstract description 5
- 238000006243 chemical reaction Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 230000007246 mechanism Effects 0.000 description 4
- 230000007935 neutral effect Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000035882 stress Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 206010016256 fatigue Diseases 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000032683 aging Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000000653 nervous system Anatomy 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 230000003313 weakening effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The adaptive choosing method of phase selecting switching-on apparatus definite value is switched the invention discloses a kind of alternating current filter, is comprised the following steps：Obtain phase selecting switching-on apparatus definite value initial value；The actual closing time of history group's alternating current filter switch is closed a floodgate next time before constitutes closing time sequence with X(0)Once tire out and operation generation is tired and sequence is used as training sample, set up GNNM (1,1) Grey Neural Network Model, it is trained using training sample and with particle swarm optimization algorithm, corresponding time series number that group's alternating current filter is closed a floodgate next time is used as input, network is exported and carries out the closing time predicted value that a regressive reduction is closed a floodgate next time, and is set to new definite value.The setting of present invention switch phase selecting switching-on apparatus definite value adaptively changing with the variation tendency of actual closing time, ensure that alternating current filter threephase switch can close a floodgate near voltage over zero, reduce transient state impact, with step is simple, fast network convergence speed and the characteristics of high precision of prediction.
Description
Technical field
Determine the present invention relates to HVDC main equipment field, more particularly to a kind of alternating current filter switch phase selecting switching-on apparatus
It is worth adaptive choosing method.
Background technology
With the expansion of power network scale, the construction and development of power system, power equipment quantity is continuously increased, various new
Equipment puts into operation, and the reliability of direct current main equipment turns into the guarantee of safe operation of power system.
Group's alternating current filter of current conversion station not only undertakes the filtering operation of system, also provides certain reactive capability, category
In one of visual plant of current conversion station.The switching operation and failure of wave filter can produce transient state stress on its element, if exceeding
The transient state rated value of design may then damage element.Wave filter minor group switches installing phase selecting switching-on apparatus can be prevented effectively greatly
The overvoltage stress that angle combined floodgate is produced, reduces the impact to element.
The principle of phase selecting switching-on apparatus is by setting combined floodgate definite value so that alternating current filter threephase switch is respectively in voltage
Near zero-crossing point closes a floodgate, to suppress the overvoltage in transient process and shove.Switch is from combined floodgate order is connected to mechanism
Dynamic/static contact contact, the time of loop conducting are actual closing time, and the actual closing time has certain dispersiveness, and
As the increase for switching the time limit that puts into operation has certain drift.But from the principle of phase selection it is recognised that phase selection
Technology requires higher to the stable action time of breaker mechanism, if phase selecting switching-on apparatus can not Accurate Prediction subsequent operation
The closing time of mechanism, then accurately controlling switch can not close a floodgate in desired phase.
The setting generic way of current phase selecting switching-on apparatus closing time definite value is set for definite value one, is run in the future
During do not make an amendment.But under the influence of mechanism aging, the factor such as closing coil change in resistance, actual closing time is with fortune
The row time limit has a certain degree of skew.Guangdong current conversion station whole station alternating current filter switch practical operation situation, real according to statistics
Border closing time all occurs in that drift compared with the phase selecting switching-on apparatus definite value set before putting into operation, and amplitude is between 1-5ms.
This will required value set must with the variation tendency of actual closing time adaptively changing.
The content of the invention
In view of the shortcomings of the prior art, it is an object of the invention to propose a kind of alternating current filter switch phase selecting switching-on apparatus
The adaptive choosing method of definite value, solving phase selecting switching-on apparatus in the prior art can not the next alternating current filter switch of Accurate Prediction
The problem of closing time of closing operation, the setting of present invention switch phase selecting switching-on apparatus definite value is with the change of actual closing time
Change trend and adaptively changing, can accurate controlling switch closed a floodgate in desired phase.
The technical scheme is that：A kind of alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value,
Comprise the following steps：
Step 1, acquisition phase selecting switching-on apparatus definite value initial value；
The step 1 comprises the following steps：
Step 11, group alternating current filter switch is subjected to offline mechanical characteristic test several times, obtains group exchange
The intrinsic closing time average value of filter switch；
Step 12, consideration are expected that wearing time and secondary circuit is delayed, and obtains phase selecting switching-on apparatus definite value initial value, the phase selection
Closing device definite value initial value is the intrinsic closing time average value, is expected to wear time and secondary circuit delay three's sum；
Step 2, put into operation group's alternating current filter, and history group's alternating current filter switch closes a floodgate next time before is actual
Closing time constitutes closing time sequence X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) }, wherein x(0)(k) represent
The corresponding closing time of kth time combined floodgate, definition k is x(0)(k) in closing time sequence X(0)In corresponding time series number, k=
1,2,3...n, with X(0)Once tire out and operation generation is tired and sequence is used as training sample；
The step 2 includes：Put into operation group's alternating current filter, offline several times mechanical with group's alternating current filter switch
The closing time sequence of attribute testing generates the initial value of training sample, group's alternating current filter switch closing time sequence X(0)With
It is updated with the closing operation of this group of alternating current filter switch, update method is：Group alternating current filter switch often closes a floodgate 1
After secondary, when the actual closing time of this combined floodgate of group's alternating current filter is updated to the combined floodgate of group's alternating current filter switch
Between sequence X(0), new closing time sequence is formed, to weaken the discreteness of sample, combined floodgate time series is tired out and operated
The tired training sample predicted with sequence as closing time next time of generation, be specially：
Original closing time sequence is defined for X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) }, then once tire out
And sequence X(1)={ x(1)(1),x(1)(2),x(1)(3),...,x(1)(n) }, whereinK=1,2,3...n；
Step 3, network receptance function and network structure are determined, set up GNNM (1,1) Grey Neural Network Model, it is described small
Group alternating current filter switch closes a floodgate next time before, using the training sample and with GNNM of the particle swarm optimization algorithm to foundation
(1,1) grey neural network is trained, and trains obtained result to be used as GNNM (1,1) gray neural particle swarm optimization algorithm
The weights and threshold value of network, the GNNM trained (1,1) Grey Neural Network Model；
Input layer, the output layer neuron number of the network structure of GNNM (1,1) Grey Neural Network Model be
1, hidden layers numbers are 2, and 2 layers of hidden neuron number is respectively 1 and 2, and the input of network is time series k, net
Network is output as the corresponding tired and sequence analog values of time series k；
The network parameter of GNNM (1,1) Grey Neural Network Model is：
Wherein：W11,W21,W22,W31,W32For each layer weights of neuron；θ is threshold value, and k is that network inputs are closing time sequence
Row number, X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) } the original time to be made up of the actual closing time of history
Sequence, X(1)Represent X(0)Tired and vector beA, b are grey parameter,
The hidden layer of the network structure of GNNM (1, the 1) Grey Neural Network Model includes the first hidden layer and the second hidden layer,
The transmission function of first hidden layer is set to Sigmoid functions, and the transmission function of input layer, output layer and the second hidden layer is linear
Function g (x)=x.
The method optimized with particle swarm optimization algorithm to GNNM (1,1) Grey Neural Network Model is：
Step 31：Using the output error of network as the fitness of each particle, the dimension of each particle is set as 2 dimensions, initially
Change population, set the initial position Z of i-th of particlei=(ai,bi), wherein, ai,biFor the grey parameter of i-th group of network, and
Choose suitable particle fitness function；
The particle fitness function is chosen for：
Wherein：N is the dimension of training sample, and O (k) exports for network, and d (k) exports for target.
Step 32：Compare the fitness of each particle and the optimal adaptation degree stored, by the two smaller value as current
The local extremum of particle, while it is the optimal adaptation degree stored to update the two smaller value, will be minimum in all particle fitness
Value, be used as global extremum；Particle rapidity and particle position are updated simultaneously；
Step 33：If the output error meets default error precision, or reaches maximum iteration, then terminate
Optimization of the particle swarm optimization algorithm to weights and threshold value, trains obtained result to be used as GNNM the particle swarm optimization algorithm
(1,1) weights and threshold value of grey neural network；Otherwise described step 32 is returned, continues to optimize；
The equation of the particle renewal speed and position is：
vij(m+1)=ω vij(m)+c1r1[Qij(m)-zij(m)]+c2r2[Qgj(m)-zij(m)]
zij(m+1)=zij(m)+vij(m+1)
Wherein, ω is inertial factor；c1,c2For accelerated factor；r1,r2For two random numbers, interval is [0 1]；vij∈
[-vmax,vmax] for the speed of i-th particle jth dimension space, vmaxTo allow mobile maximal rate, vij∈[-vmax,vmax] be
The position of i-th of particle jth dimension space, xmaxFor the maximum space position of permission；QijFor the part of i-th of particle jth dimension space
Extreme value, QgjFor the global extremum that jth is space.
The choosing method of the network parameter of GNNM (1,1) Grey Neural Network Model is：
Network receptance function is determined first, the network receptance function determined according to GNNM (1,1) Grey Neural Network Model
For
IfSo as to derive
Further derive
O=f (W11k)W21W31+f(W11k)W22W32-θ
IfThe network parameter for obtaining GNNM (1,1) Grey Neural Network Model is：
The GNNM that step 4, corresponding time series number input that group's alternating current filter closes a floodgate next time are trained
(1,1) Grey Neural Network Model, a regressive reduction is carried out to network output and obtains group's alternating current filter switch next time
The closing time predicted value of combined floodgate；
The step 4 includes：Train after GNNM (1,1) Grey Neural Network Model, group's alternating current filter is next
GNNM (1,1) Grey Neural Network Model that secondary closing time sequence number input is trained, a regressive is carried out to network output
Reduction obtains closing time predicted value；
The regressive restoring method is：
x(0)(k+1)=x(1)(k+1)-x(1)(k), k=1,2 ... n；
So as to obtain the closing time predicted value that group's alternating current filter switch closes a floodgate next time.
The described closing time predicted value of step 5, output, and the closing time predicted value is set to group's alternating current filter
The new definite value of phase selecting switching-on apparatus switch closes a floodgate next time before.
Compared with prior art, the beneficial effects of the present invention are：The present invention is excellent based on grey neural network and population
Change algorithm, it is contemplated that the dispersiveness and drift of the actual closing time of switch, instructed with all previous actual closing time sequence generation
Practice sample to be trained GNNM (1,1) Grey Neural Network Model of foundation, utilize GNNM (1, the 1) gray neural trained
Closing time predicted value that network model is closed a floodgate next time is as the new definite value of phase closing device, and phase selecting switching-on apparatus determines
Value keeps the variation tendency as actual closing time, it is ensured that alternating current filter with closing operation continuous renewal each time
Threephase switch can close a floodgate near voltage over zero, reduce transient state impact, with step is simple, network convergence speed is fast and pre-
Survey the characteristics of precision is high.
Brief description of the drawings
Fig. 1 is the flow that alternating current filter of the present invention switchs the adaptive choosing method embodiment 1 of phase selecting switching-on apparatus definite value
Schematic diagram；
Fig. 2 is that alternating current filter of the present invention switchs adaptive choosing method GNNM (1,1) grey of phase selecting switching-on apparatus definite value
Neutral net topology schematic diagram.
Embodiment
Present disclosure is described in further details with reference to the accompanying drawings and detailed description.
A kind of alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, comprises the following steps：
Step 1, acquisition phase selecting switching-on apparatus definite value initial value；
The step 1 comprises the following steps：
Step 11, group alternating current filter switch is subjected to offline mechanical characteristic test several times, obtains group exchange
The intrinsic closing time average value of filter switch；
Step 12, consideration are expected that wearing time and secondary circuit is delayed, and obtains phase selecting switching-on apparatus definite value initial value, the phase selection
Closing device definite value initial value is the intrinsic closing time average value, is expected to wear time and secondary circuit delay three's sum.
Step 2, put into operation group's alternating current filter, and history group's alternating current filter switch closes a floodgate next time before is actual
Closing time constitutes closing time sequence X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) }, wherein x(0)(k) represent
The corresponding closing time of kth time combined floodgate, definition k is x(0)(k) in closing time sequence X(0)In corresponding time series number, k=
1,2,3...n, with X(0)Once tire out and operation generation is tired and sequence is used as training sample；
The step 2 includes：Put into operation group's alternating current filter, offline several times mechanical with group's alternating current filter switch
The closing time sequence of attribute testing generates the initial value of training sample, group's alternating current filter switch closing time sequence X(0)With
It is updated with the closing operation of this group of alternating current filter switch, update method is：Group alternating current filter switch often closes a floodgate 1
After secondary, when the actual closing time of this combined floodgate of group's alternating current filter is updated to the combined floodgate of group's alternating current filter switch
Between sequence X(0), new closing time sequence is formed, to weaken the discreteness of sample, combined floodgate time series is tired out and operated
The tired training sample predicted with sequence as closing time next time of generation, be specially：
Original closing time sequence is defined for X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) }, then once tire out
And sequence X(1)={ x(1)(1),x(1)(2),x(1)(3),...,x(1)(n) }, whereinK=1,2,3...n；
Step 3, network receptance function and network structure are determined, set up GNNM (1,1) Grey Neural Network Model, it is described small
Group alternating current filter switch closes a floodgate next time before, using the training sample and with GNNM of the particle swarm optimization algorithm to foundation
(1,1) grey neural network is trained, and trains obtained result to be used as GNNM (1,1) gray neural particle swarm optimization algorithm
The weights and threshold value of network, the GNNM trained (1,1) Grey Neural Network Model；
Input layer, the output layer neuron number of the network structure of GNNM (1,1) Grey Neural Network Model be
1, hidden layers numbers are 2, and 2 layers of hidden neuron number is respectively 1 and 2, and the input of network is time series k, net
Network is output as the corresponding tired and sequence analog values of time series k；
The network parameter of GNNM (1,1) Grey Neural Network Model is：
Wherein：W11,W21,W22,W31,W32For each layer weights of neuron；θ is threshold value, and k is that network inputs are closing time sequence
Row number, X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) } the original time to be made up of the actual closing time of history
Sequence, X(1)Represent X(0)Tired and vector beA, b are grey parameter,
The hidden layer of the network structure of GNNM (1, the 1) Grey Neural Network Model includes the first hidden layer and the second hidden layer,
The transmission function of first hidden layer is set to Sigmoid functions, and the transmission function of input layer, output layer and the second hidden layer is linear
Function g (x)=x.
The method optimized with particle swarm optimization algorithm to GNNM (1,1) Grey Neural Network Model is：
Step 31：Using the output error of network as the fitness of each particle, the dimension of each particle is set as 2 dimensions, initially
Change population, set the initial position Z of i-th of particlei=(ai,bi), wherein, ai,biFor the grey parameter of i-th group of network, and
Choose suitable particle fitness function；
The particle fitness function is chosen for：
Wherein：N is the dimension of training sample, and O (k) exports for network, and d (k) exports for target.
Step 32：Compare the fitness of each particle and the optimal adaptation degree stored, by the two smaller value as current
The local extremum of particle, while it is the optimal adaptation degree stored to update the two smaller value, will be minimum in all particle fitness
Value, be used as global extremum；Particle rapidity and particle position are updated simultaneously；
Step 33：If the output error meets default error precision, or reaches maximum iteration, then terminate
Optimization of the particle swarm optimization algorithm to weights and threshold value, trains obtained result to be used as GNNM the particle swarm optimization algorithm
(1,1) weights and threshold value of grey neural network；Otherwise described step 32 is returned, continues to optimize；
The equation of the particle renewal speed and position is：
vij(m+1)=ω vij(m)+c1r1[Qij(m)-zij(m)]+c2r2[Qgj(m)-zij(m)]
zij(m+1)=zij(m)+vij(m+1)
Wherein, ω is inertial factor；c1,c2For accelerated factor；r1,r2For two random numbers, interval is [0 1]；vij∈
[-vmax,vmax] for the speed of i-th particle jth dimension space, vmaxTo allow mobile maximal rate, vij∈[-vmax,vmax] be
The position of i-th of particle jth dimension space, xmaxFor the maximum space position of permission；QijFor the part of i-th of particle jth dimension space
Extreme value, QgjFor the global extremum that jth is space.
The choosing method of the network parameter of GNNM (1,1) Grey Neural Network Model is：
Network receptance function is determined first, the network receptance function determined according to GNNM (1,1) Grey Neural Network Model
For
IfSo as to derive
Further derive
O=f (W11k)W21W31+f(W11k)W22W32-θ
IfThe network parameter for obtaining GNNM (1,1) Grey Neural Network Model is：
The GNNM that step 4, corresponding time series number input that group's alternating current filter closes a floodgate next time are trained
(1,1) Grey Neural Network Model, a regressive reduction is carried out to network output and obtains group's alternating current filter switch next time
The closing time predicted value of combined floodgate；
The step 4 includes：Train after GNNM (1,1) Grey Neural Network Model, group's alternating current filter is next
GNNM (1,1) Grey Neural Network Model that secondary closing time sequence number input is trained, a regressive is carried out to network output
Reduction obtains closing time predicted value；
The regressive restoring method is：
x(0)(k+1)=x(1)(k+1)-x(1)(k), k=1,2 ... n；
So as to obtain the closing time predicted value that group's alternating current filter switch closes a floodgate next time.
The described closing time predicted value of step 5, output, and the closing time predicted value is set to group's alternating current filter
The new definite value of phase selecting switching-on apparatus switch closes a floodgate next time before.
Illustrate with reference to specific embodiment.
Embodiment 1：, it is application example that this, which is sentenced from the western AC filter and breaker of current conversion station 551, as shown in figure 1, this hair
A kind of bright alternating current filter switch adaptive choosing method schematic flow sheet of phase selecting switching-on apparatus definite value comprises the following steps：
S11, to 551 AC filter and breakers carry out mechanical characteristic test, obtain 5 closing times.When taking estimated wear
Between be 0.4ms, secondary circuit delay be 0.1ms, obtain phase selecting switching-on apparatus three-phase definite value initial value be respectively 60.5ms,
60.5ms, 60.8ms；
S12, from western current conversion station alternating current filter switch select state-run intelligent 3YL types phase selecting switching-on apparatus, to 551 switch select
Phase closing device carries out definite value setting, and definite value is the definite value initial value that S11 steps are obtained；
Put into operation 551 group's wave filters after S13, device initialization, and the time definite value that each wave filter closes a floodgate is predicted with this time
The history closing time of first five time combined floodgate be used as training sample, constitute one group of 5 dimension closing time sequence X(0)。
S14 and S15, set up and train GNNM (1,1) Grey Neural Network Model.
Artificial neural network is the class functionally simulated from microstructure and to people's cerebral nervous system and set up
Computation model, the part image thinking ability with simulation people is adapted to nonlinear prediction and reasoning, and gray prediction method can
The general trend of data variation is showed well, and it can be very using the means that mainly generate of accumulated generating sequence as sample ordered series of numbers
The interference of the former data randomness of good weakening, the sequence after one or many Accumulating generations shows the spy of monotonic increase
It is more suitable for being trained as the sample of neutral net for property, the data strong compared to original randomness.Traditional Man
Neutral net carries out network convergence training using BP algorithm, but BP algorithm speed is slow, and is easily trapped into local extremum.Use grain
Weights and threshold value of the subgroup optimized algorithm (PSO, Particle Swarm Optimization) to GNNM (1,1) neutral net
It is trained, can reaches that computing is simple, the purpose of Fast Convergent is limited in disposal ability to adapt to, the phase selection of small internal memory is closed
Run in brake gear.
It is foregoing to set up GNNM (1,1) Grey Neural Network Models and utilize particle swarm optimization algorithm pair
What the weights and threshold value of GNNM (1,1) Grey Neural Network Model were trained concretely comprises the following steps：
A, determine network receptance function and network structure：Network receptance function is
Wherein,
IfThe network parameter that GNNM (1,1) Grey Neural Network Model can be obtained is：
In formula：W11,W21,W22,W31,W32For each layer weights of neuron；θ is threshold value, and k is natural number, X(0)To be closed by history
The original time series of lock time composition, X(1)Represent X(0)Tired and vector beA, b are grey parameter,
It is exactly that the continuous albefactions of b obtain process to a on GNNM (1,1) learning training process nature.
Network structure is as shown in Fig. 2 training sample is the history closing time generation of first five combined floodgate of this combined floodgate
Once tire out and sequence, network inputs are time series number, network is output as the corresponding tired and sequence prediction value of sequence number, network
Input layer, output layer neuron number are 1；It is 2 that network, which implies number of layers, and every layer of neuron number is respectively 1 and 2
It is individual；
B, according to A, in addition to the transmission function of the first hidden layer is set to Sigmoid functions, the transmission function of remaining each layer
It is all linear function g (x)=x.
C, the Optimal Parameters of particle swarm optimization algorithm are set as the dimension of particle；Initialize population：
Zi=(ai,bi) (3)
Wherein, ai,biFor the grey parameter of i-th group of network inputs；
Network weight and threshold value initial value are obtained according to the network parameter equation of GNNM (1,1) Grey Neural Network Model.
D, the fitness function according to particleThe fitness of each particle is calculated, wherein n is
Training sample dimension, O (k) exports for target, and d (k) is output error；
E, the fitness for comparing current particle and the optimal adaptation degree stored, regard the two smaller value as current particle
Local extremum, while it is the optimal adaptation degree stored to update the two smaller value；
F, by value minimum in all particle fitness, be used as global extremum；
G, particle rapidity is updated as the following formula：
vij(m+1)=ω vij(m)+c1r1[Qij(m)-zij(m)]+c2r2[Qgj(m)-zij(m)] (4)
Wherein, ω is inertial factor；c1,c2For accelerated factor；r1,r2For two random numbers, interval is [0 1]；vij∈
[-vmax,vmax] for the speed of i-th particle jth dimension space, vmaxTo allow mobile maximal rate, zij∈[-zmax,zmax] be
The position of i-th of particle jth dimension space, xmaxFor the maximum space position of permission；QijFor the part of i-th of particle jth dimension space
Extreme value, QgjFor the global extremum that jth is space.
The position of H, as the following formula more new particle：
zij(m+1)=zij(m)+vij(m+1) (5)
If I, the output error meet default error precision, or reach maximum iteration, then exit described
Particle swarm optimization algorithm, otherwise return to step D；
J, the particle swarm optimization algorithm trained obtained result as the weights of GNNM (1,1) grey neural network and
Threshold value, the GNNM trained (1,1) Grey Neural Network Model.
S16, to train GNNM (1,1) Grey Neural Network Model input next time closing time sequence number input
Closing time sequence number 6, obtains the corresponding tired and sequence prediction value of the closing time sequence number, tired and sequence prediction value is carried out
Closing time predicted value is used as the new definite value of phase selecting switching-on apparatus next time for repeated subtraction output；
Embodiment 2：
State-run intelligent 3YL types phase selecting switching-on apparatus is installed from the western alternating current filter of current conversion station 551, in 551 ac filters
Device is switched before putting into operation to it has carried out divide-shut brake mechanical characteristic test, obtains 551 Switch Three-Phase closing time average values, takes estimated
The time is worn for 0.4ms, secondary circuit delay is 0.1ms, obtains phase selecting switching-on apparatus definite value initial value for 60.5ms, 60.5ms,
60.8ms.The three-phase average value is switched to the definite value initial value of phase selecting switching-on apparatus as 551 alternating current filters, and has put into operation this
551 AC filter and breakers.On April 3rd, 2017 has carried out recording reading to the phase selecting switching-on apparatus, during the divide-shut brake of reading
Between as shown in table 1.
Table 1 551 switchs 3YL device recorder datas
GNNM (1,1) Grey Neural Network Model is set up, uses (1)-(5) to organize data as training sample, (6) group number
According to being used as test samples.GNNM (1,1) Grey Neural Network Model is entered using training sample and with particle swarm optimization algorithm
Row training, error precision is set to 0.01, and maximum iteration is 10000, and model is carried out using test samples after model convergence
Examine, as a result as shown in table 2.
Table 2 predicts the outcome
Further, in order to illustrate effectiveness of the invention and applicability, filtered to being exchanged from remaining 21 group of western current conversion station
Ripple device switch carries out closing time prediction using this law, and worst error is 2.1ms, and maximum relative error is 3.5%.Application effect
Show that the present invention can be good at following the variation tendency of actual closing time, and effectively predicted.Effectively prevent exchange from filtering
Ripple device switch wide-angle is closed a floodgate, and reduces overvoltage stress damage element.With step is simple, network convergence speed is fast and prediction is smart
The characteristics of spending high
Embodiment described above only expresses the several embodiments of the present invention, and it describes more specific and detailed, but simultaneously
Therefore the limitation to the scope of the claims of the present invention can not be interpreted as.It should be pointed out that for one of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the guarantor of the present invention
Protect scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (9)
1. a kind of alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, it is characterised in that it includes following
Step：
Step 1, acquisition phase selecting switching-on apparatus definite value initial value；
Step 2, put into operation group's alternating current filter, group's alternating current filter switch close a floodgate next time before the actual combined floodgate of history
Time constitutes closing time sequence X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) }, wherein x(0)(k) kth is represented
The secondary corresponding closing time that closes a floodgate, definition k is x(0)(k) in closing time sequence X(0)In corresponding time series number, k=1,2,
3...n, to X(0)Once tire out and operation generation is tired and sequence is used as training sample；
Step 3, network receptance function and network structure are determined, set up GNNM (1,1) Grey Neural Network Model, the group hands over
Stream filter switch closes a floodgate next time before, using the training sample and with particle swarm optimization algorithm to the GNNM of foundation (1,
1) grey neural network is trained, and trains obtained result to be used as GNNM (1,1) gray neural net particle swarm optimization algorithm
The weights and threshold value of network, the GNNM trained (1,1) Grey Neural Network Model；
The GNNM (1,1) that step 4, corresponding time series number input that group's alternating current filter closes a floodgate next time are trained
Grey Neural Network Model, regressive reduction of progress is exported to network and obtains what group's alternating current filter switch closed a floodgate next time
Closing time predicted value；
The described closing time predicted value of step 5, output, and the closing time predicted value is set to group's alternating current filter switch
The new definite value of the phase selecting switching-on apparatus close a floodgate next time before.
2. alternating current filter according to claim 1 switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, its feature
It is, the step 1 comprises the following steps：
Step 11, group alternating current filter switch is subjected to offline mechanical characteristic test several times, obtains group's ac filter
The intrinsic closing time average value of device switch；
Step 12, consideration are expected that wearing time and secondary circuit is delayed, and obtains phase selecting switching-on apparatus definite value initial value, the phase selection
Device definite value initial value is the intrinsic closing time average value, is expected to wear time and secondary circuit delay three's sum.
3. alternating current filter according to claim 1 switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, its feature
It is, the step 2 includes：Put into operation group's alternating current filter, and with group's alternating current filter switch, offline machinery is special several times
Property experiment closing time sequence generate the initial value of training sample, group alternating current filter switch closing time sequence X(0)Follow
The closing operation of this group of alternating current filter switch is updated, and update method is：Group's alternating current filter switch often closes a floodgate 1 time
Afterwards, the actual closing time of this combined floodgate of group's alternating current filter is updated into the closing time to group's alternating current filter switch
Sequence X(0), new closing time sequence is formed, to weaken the discreteness of sample, combined floodgate time series is carried out tired and operates life
Into tired and sequence as training sample close a floodgate next time before, to X(0)Once tire out and tired and sequence the method for operation generation is：
Original closing time sequence is defined for X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) }, then once tire out and sequence
Arrange X(1)={ x(1)(1),x(1)(2),x(1)(3),...,x(1)(n) }, wherein
4. alternating current filter according to claim 1 switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, its feature
It is, input layer, the output layer neuron number of the network structure of GNNM (1, the 1) Grey Neural Network Model are 1, hidden
Number is 2 layer by layer, and 2 layers of hidden neuron number is respectively 1 and 2, and the input of network is time series k, network
It is output as the corresponding tired and sequence analog values of time series k；
The network parameter of GNNM (1,1) Grey Neural Network Model is：
<mrow>
<msup>
<mrow>
<mo>&lsqb;</mo>
<mi>W</mi>
<mo>,</mo>
<mi>&theta;</mi>
<mo>&rsqb;</mo>
</mrow>
<mi>T</mi>
</msup>
<mo>=</mo>
<msup>
<mrow>
<mo>&lsqb;</mo>
<msub>
<mi>W</mi>
<mn>11</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>21</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>22</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>31</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>32</mn>
</msub>
<mo>,</mo>
<mi>&theta;</mi>
<mo>&rsqb;</mo>
</mrow>
<mi>T</mi>
</msup>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>a</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>-</mo>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>0</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>m</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mrow>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mrow>
<mfrac>
<mi>m</mi>
<mn>2</mn>
</mfrac>
<mo>-</mo>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>0</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein：W11,W21,W22,W31,W32For each layer weights of neuron；θ is threshold value, and k is that network inputs are closing time sequence number,
X(0)={ x(0)(1),x(0)(2),x(0)(3),...,x(0)(n) it is } original time series being made up of the actual closing time of history,
X(1)Represent X(0)Tired and vector beA, b are grey parameter,
5. alternating current filter according to claim 4 switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, its feature
It is, the hidden layer of the network structure of GNNM (1, the 1) Grey Neural Network Model includes the first hidden layer and the second hidden layer, first
The transmission function of hidden layer is set to Sigmoid functions, and the transmission function of input layer, output layer and the second hidden layer is linear function
G (x)=x.
6. alternating current filter according to claim 4 switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, its feature
It is, the method being trained with particle swarm optimization algorithm to GNNM (1,1) Grey Neural Network Model is：
Step 31：Using the output error of network as the fitness of each particle, the dimension of each particle is set as 2 dimensions, initializes grain
Subgroup, sets the initial position Z of i-th of particlei=(ai,bi), wherein, ai,biFor the grey parameter of i-th group of network, and choose
Suitable particle fitness function；
The particle fitness function is chosen for：
<mrow>
<mi>E</mi>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mn>2</mn>
<mi>N</mi>
</mrow>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<mi>O</mi>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
<mo>-</mo>
<mi>d</mi>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
Wherein：N is the dimension of training sample, and O (k) exports for network, and d (k) exports for target；
Step 32：Compare the fitness of each particle and the optimal adaptation degree stored, regard the two smaller value as current particle
Local extremum, while it is the optimal adaptation degree stored to update the two smaller value, by value minimum in all particle fitness,
It is used as global extremum；Particle rapidity and particle position are updated simultaneously；
Step 33：If the output error meets default error precision, or reaches maximum iteration, then terminate particle
Optimization of the colony optimization algorithm to weights and threshold value, trains obtained result to be used as GNNM (1,1) particle swarm optimization algorithm
The weights and threshold value of grey neural network；Otherwise described step 32 is returned, continues to optimize.
7. alternating current filter according to claim 6 switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, its feature
It is, the equation of the particle renewal speed and position is：
vij(m+1)=ω vij(m)+c1r1[Qij(m)-zij(m)]+c2r2[Qgj(m)-zij(m)]
zij(m+1)=zij(m)+vij(m+1)；
Wherein, ω is inertial factor；c1,c2For accelerated factor；r1,r2For two random numbers, interval is [0 1]；vij∈[-
vmax,vmax] for the speed of i-th particle jth dimension space, vmaxTo allow mobile maximal rate, vij∈[-vmax,vmax] for the
The position of i particle jth dimension space, xmaxFor the maximum space position of permission；QijFor the local pole of i-th of particle jth dimension space
Value, QgjFor the global extremum that jth is space.
8. alternating current filter according to claim 4 switchs the adaptive choosing method of phase selecting switching-on apparatus definite value, its feature
It is, the choosing method of the network parameter of GNNM (1, the 1) Grey Neural Network Model is：
Network receptance function is determined first, is according to the network receptance function that GNNM (1,1) Grey Neural Network Model is determined：
<mrow>
<mi>O</mi>
<mo>=</mo>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mrow>
<mo>(</mo>
<mo>(</mo>
<mrow>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>0</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mfrac>
<mi>b</mi>
<mi>a</mi>
</mfrac>
</mrow>
<mo>)</mo>
<mo>(</mo>
<mfrac>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
</mfrac>
<mo>)</mo>
<mo>+</mo>
<mfrac>
<mi>b</mi>
<mi>a</mi>
</mfrac>
<mo>&times;</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
</mfrac>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
IfSo as to derive
<mrow>
<mi>O</mi>
<mo>=</mo>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>a</mi>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mo>-</mo>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>0</mn>
<mo>)</mo>
</mrow>
</msup>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
<mo>&CenterDot;</mo>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
<mo>)</mo>
<mo>+</mo>
<mi>f</mi>
<mo>(</mo>
<mrow>
<mi>a</mi>
<mi>k</mi>
</mrow>
<mo>)</mo>
<mo>(</mo>
<mfrac>
<mrow>
<mn>2</mn>
<mi>b</mi>
</mrow>
<mi>a</mi>
</mfrac>
<mo>)</mo>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
<mo>)</mo>
<mo>+</mo>
<mo>(</mo>
<mrow>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>0</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mfrac>
<mi>b</mi>
<mi>a</mi>
</mfrac>
</mrow>
<mo>)</mo>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
Further derive
O=f (W11k)W21W31+f(W11k)W22W32-θ；
IfThe network parameter for obtaining GNNM (1,1) Grey Neural Network Model is：
<mrow>
<msup>
<mrow>
<mo>&lsqb;</mo>
<mi>W</mi>
<mo>,</mo>
<mi>&theta;</mi>
<mo>&rsqb;</mo>
</mrow>
<mi>T</mi>
</msup>
<mo>=</mo>
<msup>
<mrow>
<mo>&lsqb;</mo>
<msub>
<mi>W</mi>
<mn>11</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>21</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>22</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>31</mn>
</msub>
<mo>,</mo>
<msub>
<mi>W</mi>
<mn>32</mn>
</msub>
<mo>,</mo>
<mi>&theta;</mi>
<mo>&rsqb;</mo>
</mrow>
<mi>T</mi>
</msup>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>a</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>-</mo>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>0</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>m</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mrow>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>+</mo>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>a</mi>
<mi>k</mi>
</mrow>
</msup>
</mrow>
<mo>)</mo>
</mrow>
<mrow>
<mo>(</mo>
<mrow>
<mfrac>
<mi>m</mi>
<mn>2</mn>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<msup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mn>0</mn>
<mo>)</mo>
</mrow>
</msup>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>.</mo>
</mrow>
9. the adaptive choosing method of phase selecting switching-on apparatus definite value according to claim 1, it is characterised in that the step 4
Including：Train after GNNM (1,1) Grey Neural Network Model, by group's alternating current filter, closing time sequence number is defeated next time
Enter GNNM (1, the 1) Grey Neural Network Model trained, a regressive reduction acquisition closing time is carried out to network output pre-
Measured value；
The regressive restoring method is：
x(0)(k+1)=x(1)(k+1)-x(1)(k), k=1,2 ... n；
So as to obtain the closing time predicted value that group's alternating current filter switch closes a floodgate next time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710333433.2A CN107292764A (en) | 2017-05-12 | 2017-05-12 | Alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710333433.2A CN107292764A (en) | 2017-05-12 | 2017-05-12 | Alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107292764A true CN107292764A (en) | 2017-10-24 |
Family
ID=60095351
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710333433.2A Pending CN107292764A (en) | 2017-05-12 | 2017-05-12 | Alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107292764A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108767821A (en) * | 2018-06-15 | 2018-11-06 | 广州供电局有限公司 | A kind of phase selection operation/cutting method and system |
| CN109921425A (en) * | 2019-03-29 | 2019-06-21 | 云南电网有限责任公司电力科学研究院 | A kind of alternating current filter phase selection control method and system based on converter station |
| CN111208418A (en) * | 2020-01-10 | 2020-05-29 | 中国南方电网有限责任公司超高压输电公司广州局 | Phase selection switching-on and switching-off state monitoring system and method for converter station alternating current filter |
| CN112103929A (en) * | 2020-09-24 | 2020-12-18 | 中国南方电网有限责任公司超高压输电公司曲靖局 | Phase selection and switching-off method for AC filter of same-tower double-circuit DC converter station |
| CN113555856A (en) * | 2020-04-24 | 2021-10-26 | 中国南方电网有限责任公司超高压输电公司南宁局 | Intelligent control method, device and system for circuit breaker |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678773A (en) * | 2013-11-15 | 2014-03-26 | 华南理工大学 | Sphere gap discharge voltage detection method |
| CN103729685A (en) * | 2013-12-02 | 2014-04-16 | 国电南瑞科技股份有限公司 | Method for photovoltaic power station group region power prediction on basis of improved RBF neural network |
| CN104156766A (en) * | 2014-07-28 | 2014-11-19 | 山东山大世纪科技有限公司 | Application of intelligent memory and self-learning system aiming at high-voltage switch divide-shut brake time |
| CN105139264A (en) * | 2015-06-06 | 2015-12-09 | 安徽工程大学 | Photovoltaic generation capacity prediction method based on particle swarm algorithm wavelet neural network |
| CN106503689A (en) * | 2016-11-14 | 2017-03-15 | 哈尔滨理工大学 | Neutral net local discharge signal mode identification method based on particle cluster algorithm |
-
2017
- 2017-05-12 CN CN201710333433.2A patent/CN107292764A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103678773A (en) * | 2013-11-15 | 2014-03-26 | 华南理工大学 | Sphere gap discharge voltage detection method |
| CN103729685A (en) * | 2013-12-02 | 2014-04-16 | 国电南瑞科技股份有限公司 | Method for photovoltaic power station group region power prediction on basis of improved RBF neural network |
| CN104156766A (en) * | 2014-07-28 | 2014-11-19 | 山东山大世纪科技有限公司 | Application of intelligent memory and self-learning system aiming at high-voltage switch divide-shut brake time |
| CN105139264A (en) * | 2015-06-06 | 2015-12-09 | 安徽工程大学 | Photovoltaic generation capacity prediction method based on particle swarm algorithm wavelet neural network |
| CN106503689A (en) * | 2016-11-14 | 2017-03-15 | 哈尔滨理工大学 | Neutral net local discharge signal mode identification method based on particle cluster algorithm |
Non-Patent Citations (11)
| Title |
|---|
| 俞秀婷等: "基于改进的PSO算法优化灰色神经网络的话务量预测", 《计算机工程与设计》 * |
| 吴丽丽: "基于BP神经网络的时序残差GM（1,1）模型", 《科技经济市场》 * |
| 张庆杰等: "配永磁机构的真空断路器同步分合闸控制系统设计与实现", 《电力自动化设备》 * |
| 杨帆等: "±800kV侨乡换流站500kV交流滤波器小组断路器选相合闸问题分析", 《南方电网技术》 * |
| 王科等: "基于自适应预测的同步断路器", 《中国电力》 * |
| 罗新等: "粒子群优化自适应小波神经网络在带电局放信号识别中的应用", 《电工技术学报》 * |
| 赵佳等: "基于分布式神经网络的同步合闸时间预测方法研究", 《上海电气技术》 * |
| 顾胜等: "基于灰色粒子群神经网络的航天器参量预测", 《数据采集与处理》 * |
| 马吉明等: "基于粒子群优化的灰色神经网络组合预测模型研究", 《计算机工程与科学》 * |
| 马晶晶等: "基于永磁操动机构的同步合闸时间预测方法研究", 《合肥工业大学学报（自然科学版）》 * |
| 龙会典等: "基于GM(1,1)和BP神经网络集成模型的我国人口数量时间序列预测研究", 《计算机光盘软件与应用》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108767821A (en) * | 2018-06-15 | 2018-11-06 | 广州供电局有限公司 | A kind of phase selection operation/cutting method and system |
| CN108767821B (en) * | 2018-06-15 | 2020-01-17 | 广州供电局有限公司 | Switching method and system for phase selection switch-on |
| CN109921425A (en) * | 2019-03-29 | 2019-06-21 | 云南电网有限责任公司电力科学研究院 | A kind of alternating current filter phase selection control method and system based on converter station |
| CN109921425B (en) * | 2019-03-29 | 2023-03-14 | 云南电网有限责任公司电力科学研究院 | AC filter phase selection and closing control method and system based on converter station |
| CN111208418A (en) * | 2020-01-10 | 2020-05-29 | 中国南方电网有限责任公司超高压输电公司广州局 | Phase selection switching-on and switching-off state monitoring system and method for converter station alternating current filter |
| CN111208418B (en) * | 2020-01-10 | 2022-05-13 | 中国南方电网有限责任公司超高压输电公司广州局 | Phase selection switching-on and switching-off state monitoring system and method for converter station alternating current filter |
| CN113555856A (en) * | 2020-04-24 | 2021-10-26 | 中国南方电网有限责任公司超高压输电公司南宁局 | Intelligent control method, device and system for circuit breaker |
| CN113555856B (en) * | 2020-04-24 | 2023-02-10 | 中国南方电网有限责任公司超高压输电公司南宁局 | Intelligent control method, device and system for circuit breaker |
| CN112103929A (en) * | 2020-09-24 | 2020-12-18 | 中国南方电网有限责任公司超高压输电公司曲靖局 | Phase selection and switching-off method for AC filter of same-tower double-circuit DC converter station |
| CN112103929B (en) * | 2020-09-24 | 2022-03-22 | 中国南方电网有限责任公司超高压输电公司曲靖局 | Phase selection and switching-off method for AC filter of same-tower double-circuit DC converter station |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107292764A (en) | Alternating current filter switchs the adaptive choosing method of phase selecting switching-on apparatus definite value | |
| Li et al. | Design of a fuzzy-PID controller for a nonlinear hydraulic turbine governing system by using a novel gravitational search algorithm based on Cauchy mutation and mass weighting | |
| Weerasooriya et al. | Towards static-security assessment of a large-scale power system using neural networks | |
| Abdechiri et al. | Neural network learning based on chaotic imperialist competitive algorithm | |
| CN114217524B (en) | Power grid real-time self-adaptive decision-making method based on deep reinforcement learning | |
| Tso et al. | An ANN-based multilevel classification approach using decomposed input space for transient stability assessment | |
| Bashir et al. | A new hybrid particle swarm optimization for optimal coordination of over current relay | |
| Park et al. | Economic load dispatch for nonsmooth cost functions using particle swarm optimization | |
| Ng et al. | Fast convergence for backpropagation network with magnified gradient function | |
| Tjahjono et al. | Overcurrent relay curve modeling and its application in the real industrial power systems using adaptive neuro fuzzy inference system | |
| CN116454926A (en) | Multi-type resource cooperative regulation and control method for three-phase unbalanced management of distribution network | |
| Mohamed et al. | Fault diagnosis system for tapped power transmission lines | |
| Tripathy et al. | Radial basis probabilistic neural network for differential protection of power transformer | |
| Biswas et al. | Rainfall forecasting by relevant attributes using artificial neural networks-a comparative study | |
| Etemadi et al. | Neural network based fault diagnosis in an HVDC system | |
| Guo et al. | BP neural network optimized with PSO algorithm and its application in forecasting | |
| Aryanezhad et al. | Voltage dip mitigation in wind farms by UPQC based on Cuckoo Search Neuro Fuzzy Controller | |
| Zhang | The application of imperialist competitive algorithm based on chaos theory in perceptron neural network | |
| Balaga et al. | Numerical differential protection of power transformer using ANN as a pattern classifier | |
| Nogay et al. | Diagnosis of chaotic ferroresonance phenomena using deep learning | |
| Niazi et al. | A hybrid approach for security evaluation and preventive control of power systems | |
| Skoundrianos et al. | Modelling and FDI of dynamic discrete time systems using a MLP with a new sigmoidal activation function | |
| CN112100767A (en) | Aero-engine service life prediction method based on singular value decomposition and GRU | |
| Boozari | State estimation and transient stability analysis in power systems using artificial neural networks | |
| Tripathy et al. | Optimal radial basis function neural network power transformer differential protection |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171024 |