Paper Presentation -- AI AUTOMATES SUBSTATION CONTROL

ABSTRACT

Controlling a substation by a fuzzy controller speeds up response time diminishes up the possibility of risks normally related to human operations. The automation of electric substation is an area under constant development Our research has focused on, the Selection of the magnitude to be controlled, Definition  and  implementation  of  the  soft  techniques, Elaboration  of  a  programming  tool  to  execute  the  control  operations. it is possible to control the desired status while supervising some important magnitudes as the voltage, power factor, and harmonic distortion, as well as the present status. The status of the circuit breakers can be control by using a knowledge  base that relates some of the operation magnitudes, mixing status variables with time variables and fuzzy sets .The  number of necessary magnitudes to a supervise and to control a substation can be very high in the present research work, many magnitudes were not included .To avoid the extensive number of required rules nevertheless , controlling a substation by a fuzzy controller has the advantage that it can speed up the response time and diminish the possibility of risks normally related to human operations.  

AI AUTOMATES SUBSTATION CONTROL

Introduction

                Electric substations are facilities in charge of the voltage transformation to provide safe and effective energy to the consumers. This energy supply has to be carried out with sufficient quality and should guarantee the equipment security.  The associated cost to ensure quality and security during the supply in substations is high.  Automatic mechanisms are generally used in greater or lesser scale, although they mostly operate according to an individual control and protection logic related with the equipment itself and not with the topology of the whole substation in a given moment.

                The automation of electric substation is an area under constant development.  Nevertheless, the control of a substation is a very complex task due to the great number of related problems and, therefore, the decision variables that can influence the substation performance.  Under such circumstances, the use of learning control systems can be very useful.

                Many papers on applications of artificial intelligence  (AI) techniques to power system have been published in the last year.  The difficulties associated with the application of this technique include:

§  Selection  of  the  magnitude  to  be  controlled

§  Definition  and  implementation  of  the  soft  techniques

§  Elaboration  of  a  programming  tool  to  execute  the  control  operations

§  Selection,  acquisition  and  installation  of  the  measurement  and  control  equipment 

§  Interface  with  this  equipment  and

§  Applications of the controlling technique in existent substations.

                   Our research has focused on the first three points, and the interest of the present work is to expose the obtained result and to present them for discussion.  The objective is to show that it is possible to control the status of circuit breakers (CB) in a substation making use of a knowledge, mixing status variables with time variables and fuzzy sets.

                     Even  when all the magnitudes to be controlled cannot be included in the analysis  (mostly due to the great number of measurements and status variables of the substation and, therefore, to the rules that would be required by the controller), it is possible to control the desired status while supervising some important magnitudes as the voltage, power factor, and harmonic distortion, as well as the present status.

PLANT DESCRIPTION

                      The system under study represents a test substation with two 30KVA three-phase transformers, two CBs, two switches, three current transformers, and two potential transformers.  It also contains an auto transformer  (to regulate the input voltage) as well as impedance to simulate the existence of a transmission line.  The input voltage are the same  (220V), this characteristic was selected in order to analyze the operation of the controller in a laboratory scale in a second stage of the development of the present work.

Therefore, the first transformer increases the voltage to a value of 13.2KV, while the second lowers it again to 220V. fixed filter, an automatic filter for the control of the power factor and the regulation of the voltage, and three feeding lines with diverse type of loads of different nature  (including nonlinear loads) are connected through CBs to the output bar.  Figure.  1 shows the proposed outline 

                Since the control elements in substations are the CBs and switches, the goal is to allow the control of the five selected CBs (in the output bar) according to some configurations and measurements of the observation variables.  Initially, the computer-aided system will try to control the plant and will send alarm signals when it can not find a solution, waiting for the human intervention.  So it will learn how the human operator reacts to the inputs and will generate the corresponding behavior rules.  In this way, the system will replace gradually the human operator.

CONTROLLER DESIGN

 DEFINITION OF N THE INPUT AND OUTPUT VARIABLES

                  There are a great number of variables that can be chosen to control a substation.  Nevertheless, a limited number of variables were selected for this study.

      The following input variables have been defined:

§  Vout:  Voltage at output bus, phase A (V).

§  PF:  power  factor  at  output  bus,  phase  A

§  THDv(%):Total  voltage   harmonic  distortion  at  output  bus(%)

§  Tv, t(s):  Amount  of  time  the  voltage   Vout  is  in  range  [114.3V;119.5V](tolerance  zone)

§  Tv,nt(s): Amount  of  time  the  voltage  Vout   is  below  114.3V(alert  zone).

     In  selecting  the   variables,  several  aspects  were  kept  in  mind.  For  example,  the  voltage  influences  in  the  connection  and  disconnection  of  loads  when  its  value  leaves  some  ranges  during  certain  time.  These  ranges  are  represented  in  figure.2

             A  decrease  of  the  voltage  7.5%  below  the  nominal  value  is  allowed  during  a  certain  time  (for  example,  10  min),  while  from  10%  on,  the  maximum  allowed  time  is  much  smaller  (for  example,  20  s).  In  case  these  limits  are exceeded,  loads  will  be  disconnected  in  an  established   preference  order.  The  reconnection  of  loads  will  occur  when  the  voltage  arrives  at  values  above  the  tolerance  interval, i.e.,  and the normal interval.

     On  the  other  hand,  the  power  factor  and  the  total  voltage  harmonic  distortion  influence  the  connection  and  disconnection  of  capacitor   and  filters.  These  variables  can  be  read  by  means  of  sensors  and/or  transducers,  including  signal  conditional  accessories  and    directed  to  the  data  acquisition  card (DAC)  by  means  of  analog  lines.  In  this  example,  the  currents  and  voltages  in  each  phases  are  not  kept  in  mind  due  to  the  great  number  of  rules  that  would  be  required  by  the  controller.

       The  CBs  were  defined  as  status  variables.  The  switches  were  not  included  in  this  study,  because  their  status  is  only  important  for  switching  and  not  for  control  purposes. 

              The  following  status  variables  were  defined:

§  Df:  Status  of  the  CB  connecting  the  fixed  filter 

§  Dc:  Status  of  the  CB  connecting  the  controlled  filter

§  D11,D12,D13:  Status  of  the  CBs  connecting  the  load  feeders  1,2,  and3.

                 Each  combination  of  status  variables  defines  a  topology  and  is  an  input  for  the  controller.  Possible  values  for   each  status  variable  are  0(open)  and  1  (closed).

                Thus,  what  is  intended  to  control  is  the  moment  when  the  filters  and  the  loads  should  be  connected  or  disconnected  by  means  of  signals  that  are  sent  to  the  CBs.  The  controlled  filter,  once   connected,  will  maintain  its  functionality  as  an  automatic  filter  in  dependence  of  the  present  harmonic  distortion  over  the  time.  In  case  of  a  disconnection  of  some  filter  due  to  over currents,  the  controller  can  activate  connection  rules  after  some  time  that  can  be  freely  defined  before  the  controller  starts.

                 The  actions  to  carry  out  as  a  response  to  disturbances  in  the  measurements  (values  out  of  normal  ranges)  are  dependent  not  only  on  the  present  topology  of  the  substation,  i.e.,  on  the  values  of  the  present  status  variables.  For  example;  the  connection  of  the  controlled  filter  can  only  occur  when  the  fixed  filter  is  connected.  Similarly,  the  disconnection  of  the  fixed  filter  can  only  occur  after  the  controlled   filter  has  been  disconnected.

                  The  status  variables  are  input  and  also  output  variable.  Each  CB  will  maintain  its  standard  protection  function  against  over currents.  Since  signal  to  these  devices  are  sent  by  means  of  additional  relays  to  have  the  ability  to  be  controlled  in  parallel.

                   The  status  variables  are  read  by  sensors  and/or  transducer  and  connected  to  digital  lines  to  a  DAC.  Digital lines to the CB relays also send the outputs.

                   For  the  definition  of  sets,  triangular  and  trapezoidal  shape  functions  were  used.

                   The  status  variables  were  not  fuzzified  because  their  measurement  are  not  provided  with  uncertainty.  They  can  only  accept  two  values:0  and  1.

        The  time  the  voltage  is  in  tolerable  and  not  tolerable  ranges  is  supervised  through  the  event  counters.

RULE  SYSTEM

         The  syntax  of  each  rule  can  be  expressed  for  example  follows:

                           IF

                           (V is  Tolerable )  and  (PF is Low) and(THD is Tolerable)and                                   (Tv,t is acceptable) and (Tv,  nt  is  zero) and  (present  topology  is  00110)

                            THEN

                            (Desired  topology  is  10110)  

         The  topology  is  expressed  as  a  five-digit  binary  number   that  refers  to  the  five  CBs  in  the  following  order:

§  First  digit:  CB  of  the  fixed  filter

§  Second  digit:  CB  of  the  controller  filter

§  Third  digit:  CB  of  the  first  load  (priority  one)

§  Fourth  digit: CB of  the  second  load  (priority  two)

§  Fifth  digit:  CB  of  the  third  load(priority  three).

          In  simple  words,  the  rule  expressed  above  means:  If  the  present  topology  is  00110(i.e., both  filter  are  disconnected  and  only  load  1  and  2  are  connected)  and  the  following  situation  is  found:

§  Voltage  is  tolerable

§  Power  factor  is  low

§  Total  harmonic  distortion  is  tolerable 

§  Voltage  has  been  in  a  tolerable  zone  for  an  acceptable  time

§  Voltage  has  not  yet  entered  a  not  acceptable  zone  then  the  desirable  topology  is  10110,  which  means  that  we  must  switch  in  the  fixed  filter.

           To  establish  the  connection  and  disconnection  rules  of  the  loads,  it  was  attributed  to  load  1  the  biggest  preference  and  to  load  3  the  smallest.

           The  definition  of  the  analog  variables  was  carried  out  using  the  following  terms:

§  For  Vout,  the  fuzzy  sets:  NT(not  tolerable),  T(tolerable),  H(height),  N(normal)

§  For  PF,  the  fuzzy  sets:  L(low),  T(tolerable),  H(height)

§  For  THDv,  the  fuzzy  sets:L(low),t(tolerable),  H(height)

§  For  Tv,t, the crisp sets:Z(zero),A(acceptable),NA(not acceptable)

Figure 3 shows the defined sets for each input variable. In the  present work a preliminary calculation of the Maximum number  of rules yields that 7,776 rules are necessary in order to start the controller .However, this number could be reduced notably by keeping in mind that, among the 32 available states , not all  can

Be considered as possible. This way, only 9 topologies have been found as possible, decreasing to 2,187 the Maximum number of necessary rules so that the controller can totally replace the human operator . Nevertheless , this number was reduced with the inclusion of some initials rules .

INITIAL RULE BASE                                                                                                                                                                                                                                                                                                                                          In case there exists a knowledge base on the  plant to be controlled , some rules can  be included as a   starting point . The initial knowledge  base  can be defined in  such

a way that diminish  the  necessary number  of rules for  the controller to work properly.

V	PF	THDv	Tv,t	Tv,nt	Present state	Desired state
NT	T	U	NA	Z	Z	Z
NT	U	H	U	A	A	A
NT	U	U	U	NA	NA	NA

Fig. 4. Some of the first 144 rules as they are presented to the software

Layer	Number of Neurons	Type of Activation function
1	Number of  input sets =20	Linear(input=output)
2	2*(number of input sets)+1=41	Sigmoidal
3	Number of state variables=5	Sigmoidal

This  happens  in  most  cases   where  some  situations  are  not  possible.  In  the  presented  study,  144  initial  rules  were  included  (Figure.4),  all  in  form  of  extended  rules    (including  one  term  U,  which  indicates  that  the  attribute  can  take  any  value).

AUTOMATIC  RULE  EXTRACTION.

                 The   rule   base  represents  the  knowledge  base  of    the  controller.  The  proposed  approach  is  able   to  start the  operation  with  an  empty  or  uncompleted  rule  base.

                 During  the  inference  process,  the  membership  degree  corresponding  to  each  column  in  the  rule  is  calculated.  The  type  of  the  output  of  this  function  in  the  universe  of  discourse  depends  on  the  variable  type  and  on  the  selected  term.

                  The  rule  extraction   takes  place  each  time  the  controller  does  not  find  any  rule  in  the  rule  base  with  a fire  degree  bigger   than  zero,  and,  as  a  result, an  alarm  requiring  an  operator  action  is  sent.

INFERENCE  MODULE

                 The  controller  outputs  are  decided  by  searching  in  the  rule  base.  In  this  step,  called  inference,  the  fire  degree  of  each  rule  is  calculated.                                                    

                 Since  the consequence  part  in  each  rule  only  deals  with  status  variables  whose  values  are  crisp  numbers(0  and  1),  use  of  defuzzification  method  is  not  necessary. Therefore,  the  controller   output  in  each  case  will  be  the  consequent  part  of  the  rule  with  the  biggest  fire  degree                                                     

OPERATON  MODULES  OF  THE  CONTROLLER

               The  controller  operation  can  be  carried  out  in  two  ways.  Following  the  typical  way  the  controller  run,  it  can  be  placed  in  operation  with  a  completed  rule  base,  totally  replacing the  human  operator.  Nevertheless,  it  can  be  started  with  an  empty  or  incomplete   rule  base,  which  means  that  it  will  activate  a  leaning  mechanism.  This  way,  the  controller  will  complete  step  by  step  the  operation  rules  and  at  the  same  time  will  replace  the  human  operator.  A  representation  of  the  controller  function  can  be  viewed   in  figure 5.

               For  the  disconnection  of  loads  when  the  voltage  diminishes,  the  following  steps  can  be  processed  as  an  automatic  response  of  some  devices:

§  Automatic  tap  change  in  transformers

§  Voltage  regulation  in  the  autotransformer.

               If  all  these  possibilities  have  been  tried  and  the  voltage  continues  being    low,  then  operator  outputs  are  processed.

                To  avoid  interferance  with  this  response,  the  execution  of  controller  outputs  (i.e.,  status  changes  of  the  CBs)  can  be  postponed  each  time  for  one  sampling  interval,  provided  the  inference  in  the  next  sampling  time  yields  the  necessity  of  status  changes

               In  case  the  system  cannot  find  a  satisfactory  solution  because   no  rule  could  be  fired,  the  system  sends  an  alarm  to  the  operator.  He  will  have  some  time  to  decide  what  action  to  take  using  the  switch  components  on  the  PC  screen.

               If  the  operator  action  does  not  arrive  during  this  time  limit  (either  due  to  delays  or  to  operator  absence),  then  the  controller  will  execute  a  protection  rule  previously  defined,  which  could  be,  for  example,  the  total  disconnection   of  the  substation.  After  saturation.  Of  the  knowledge  base  or  after  a  certain  operating  time,  the  system  will  generate  and  train  an  artificial  neural  network  (ANN),  in  order  to  replace  the  rule  base.

SUBSTATION   OF  THE  INFERENCE  ENGIN  BY  AN  ARTIFICIAL  NEURAL  NETWORK 

                According  to  the  type  and  the  number  of  initial  rule  as  well  as  the  number  of  existent  linguistic  variables,  the  system  will  calculate  the  maximum  number  of  possible  rule.

                When  arriving  at  this  number  or  time  limit,  the  system  will   start  a  module,  where  a  feed-forward  ANN  with  a  hidden  layer  will  be  generated  and  trained  with  a codified  rule  base.

                This   way,  the  system  will  try  to  diminish  the  time  controller  needs  for  the  inference,  avoiding  searching  in  an extensive knowledge base. Once network has been trained, the interference will be carried out by means of network and not through the rule base.

Each training pattern is coded with binary digits(0,1). In case of analog variables , these digits are the codification of the sets in each rule.

                  The network of the present example is a feed forward network built by 20 neurons in the input layer, 41 neurons in the hidden layer, 5 neurons in the exit layer (table 1)

                                  

                    The training process is carried out via a back-propagation algorithm. Stop criteria for the training is the total network error as well as a time limit. In fact , interference by means of  the rule base will finish when the network is completely trained.

                     During the interference through the neural network, the analog measurements will have to be processed initially through a membership function module before being presented to the network. This step is necessary inn order to codify the rules  with membership degrees of each sets for all magnitudes, i.e., instead of having binary digits (0,1), the input pattern will be vectors of real numbers in the range [0,1] that represent the membership degrees of the inputs  in each set.

The results of the output layer are real numbers between zero and one; thus, the controllers will round these results to an integer-type  value in order to find the proposed status for the CBs.

Experimental results.

                    To carry out the experiment, a software for the platform windows 9x/2000 using Delphi was elaborated. Signal generators for the analog input variables were used.

                     The experiment was starting from the status 0011. During the first 400 measurments,243 actions could not be determined by the controller ,an expert gave the answers i.e., ,and as a result,243 new controller rules were extracted.

                      Figure 6 shows the status behavior of the plant during the first 400 sampling times with extreme simulated variations in the inputs. The status frequency is shown in table2.

Table 2. Frequency of topologies obtained during the 400 system operations
Topology	Frequency
0(=00000)	0.019
4(=00100)	0.009
6(=00110)	0.102
7(=00111)	0.046
20(=10100)	0.93
22(=10110)	0.056
23(=10110)	0.009

                     The following  can  be  said  about  the  comparison  of  the  plant  behavior  with  and  without  the  controller.  This   control  system  has  been  designed  for human  operator  replacement,  i.e.,

To  decide  about  actions  beyond  the  conventional  automatic  control  procedure,  i.e.,  action  for  which  a  human  operator    is  always  needed.  Since  the  implemented  controller  responds  according  to  a  knowledge  base  extracted  by  human  operations,  the  topology  change  will  always  be  the  same  (with  or  without  the  controller).  Nevertheless,  the  difference  and  advantage  due  to  the  use  of  this  controller  reside  in:

§  Speeding  up  response  time

§  Avoiding  operation  mistakes

§  Gradually  replacing  a  human  operator.

                        The  network  was  trained  with  a  group  of  1,377  rule.  It  was  found  that,  for  activation  functions  of  lineal  type  (first  layer)  and  sigmoidal  type  (hidden  and  last  layer),  an  approximate  time  of  2  hours  a  455  MHz  Pentium-based  PC  was  sufficient  to  obtain   approaches  with errors  smaller  than  0.1  for  each  output  neuron.  Since  the  final  result  of  the  approximation  by  means  of  the  network  is  based  on  whole  rounds  to  zero  or  one,  this  level  of  accuracy  was  found  acceptable.

FAST  RESPONSE  AND  DIMINISHED   RISK

                          For  the  control  of  a  substation  by  means  of  the  connection  of  devices  for  improving  its  performance,  it  is  necessary  to  keep  in  mind  not  only  the  measurement  of the  electric  magnitudes  but  also  the  status  of  some  control  devices  that  define  their  topology.

                          Control  and  protection  devices  governed  by  individual  decision  methods  that  allow  the  substation  operator  when  needed  can  be  coordinated   from  an  upper  supervisory  level.  It  is  possible  to  follow  a  preference   criterion  for  the  disconnection  or  connection   of  CBs  (in  this  case  for  load  feeders  and  filters)  when  using  a  controller  based  system.                                                                        

                          Controllers  can  be  used   for  rule  extraction  in  a  first  working  stage.  Control  systems  that  do  not  require  instantaneous  responses  in  an  initial  stage,  can  be  designed  for  learning  the  operator  action  and  constructing   a  decision  table  for  total  replacement  of  the  human  operator  in  the  future.

                            Even  when  the  number  of  rules  for  controlling  the  substation  is  very  high,  it  is  possible  to  obtain  these  rules  automatically  by  means  of  a  fuzzy  controller.  After  a  certain  operating  time  (which  depends  on  the  initial  knowledge  base  as  well  as  on  the  status  variables  and  the  fuzzification  of  the  inputs),  the  inference  process  through  a  rule  base  can  be  replaced   by  an  approximation  via  an  ANN  to  diminish  response  time.

                             The  number  of  necessary  magnitudes  to  supervise  and  to  control  a  substation  can  be  very  high.  In  the  present  research  work,  many  magnitudes   were  not  included  to  avoid  the  extensive  number  of  required  rules.

                             Nevertheless,  controlling  a  substation  by  a  fuzzy  controller  has  the  advantage  that  it  can  speed  up  the  response  time  and  diminish  the  possibility  of  risks  normally  related  to  human  operations.