Source code for RCAIDE.Framework.Optimization.Packages.pyopt.pyopt_setup

# pyopt_setup.py
#
# Created:  Jul 2015, E. Botero
# Modified: Feb 2016, M. Vegh
#           May 2021, E. Botero 

# ----------------------------------------------------------------------
#  Imports
# ----------------------------------------------------------------------

# RCAIDE imports
import numpy as np
from RCAIDE.Framework.Optimization.Common import helper_functions as help_fun


# ----------------------------------------------------------------------
#  Pyopt_Solve
# ---------------------------------------------------------------------- 


def Pyopt_Solve(problem,solver='SNOPT',FD='single', sense_step=1.0E-6,  nonderivative_line_search=False):
    """ This converts your RCAIDE Nexus problem into a PyOpt optimization problem and solves it
        PyOpt has many algorithms, they can be switched out by using the solver input. 

        Assumptions:
        None

        Source:
        N/A

        Inputs:
        problem                   [nexus()]
        solver                    [str]
        FD (parallel or single)   [str]
        sense_step                [float]
        nonderivative_line_search [bool]

        Outputs:
        outputs                   [list]

        Properties Used:
        None
    """      
   
    # Have the optimizer call the wrapper
    mywrap = lambda x:PyOpt_Problem(problem,x)
   
    inp = problem.optimization_problem.inputs
    obj = problem.optimization_problem.objective
    con = problem.optimization_problem.constraints
   
    if FD == 'parallel':
        from mpi4py import MPI
        comm = MPI.COMM_WORLD
        myrank = comm.Get_rank()      
   
    # Instantiate the problem and set objective
    import pyOpt
    opt_prob = pyOpt.Optimization('RCAIDE',mywrap)
    for ii in range(len(obj)):
        opt_prob.addObj(obj[ii,0])    
       
    # Set inputs
    nam  = inp[:,0] # Names
    ini  = inp[:,1] # Initials
    bndl = inp[:,2] # Bounds
    bndu = inp[:,3] # Bounds
    scl  = inp[:,4] # Scale
    typ  = inp[:,5] # Type
   
    # Pull out the constraints and scale them
    bnd_constraints = help_fun.scale_const_bnds(con)
    scaled_constraints = help_fun.scale_const_values(con,bnd_constraints)
    x   = ini/scl
   
    for ii in range(0,len(inp)):
        lbd = (bndl[ii]/scl[ii])
        ubd = (bndu[ii]/scl[ii])
        #if typ[ii] == 'continuous':
        vartype = 'c'
        #if typ[ii] == 'integer':
            #vartype = 'i'
        opt_prob.addVar(nam[ii],vartype,lower=lbd,upper=ubd,value=x[ii])
       
    # Setup constraints  
    for ii in range(0,len(con)):
        name = con[ii][0]
        edge = scaled_constraints[ii]
       
        if con[ii][1]=='<':
            opt_prob.addCon(name, type='i', upper=edge)
        elif con[ii][1]=='>':
            opt_prob.addCon(name, type='i', lower=edge,upper=np.inf)
        elif con[ii][1]=='=':
            opt_prob.addCon(name, type='e', equal=edge)

    # Finalize problem statement and run  
    print(opt_prob)
   
    if solver == 'SNOPT':
        import pyOpt.pySNOPT
        opt = pyOpt.pySNOPT.SNOPT()
        CD_step = (sense_step**2.)**(1./3.)  #based on SNOPT Manual Recommendations
        opt.setOption('Function precision', sense_step**2.)
        opt.setOption('Difference interval', sense_step)
        opt.setOption('Central difference interval', CD_step)

    elif solver == 'COBYLA':
        import pyOpt.pyCOBYLA
        opt = pyOpt.pyCOBYLA.COBYLA() 
        
    elif solver == 'SLSQP':
        import pyOpt.pySLSQP
        opt = pyOpt.pySLSQP.SLSQP()
        opt.setOption('MAXIT', 200)
    elif solver == 'KSOPT':
        import pyOpt.pyKSOPT
        opt = pyOpt.pyKSOPT.KSOPT()
    elif solver == 'ALHSO':
        import pyOpt.pyALHSO
        opt = pyOpt.pyALHSO.ALHSO()   
    elif solver == 'FSQP':
        import pyOpt.pyFSQP
        opt = pyOpt.pyFSQP.FSQP()
    elif solver == 'PSQP':
        import pyOpt.pyPSQP
        opt = pyOpt.pyPSQP.PSQP()    
    elif solver == 'NLPQL':
        import pyOpt.pyNLPQL
        opt = pyOpt.pyNLPQL.NLPQL()    
    elif solver == 'NSGA2':
        import pyOpt.pyNSGA2
        opt = pyOpt.pyNSGA2.NSGA2(pll_type='POA') 
    elif solver == 'MIDACO':
        import pyOpt.pyMIDACO
        opt = pyOpt.pyMIDACO.MIDACO(pll_type='POA')     
    elif solver == 'ALPSO':
        import pyOpt.pyALPSO
        #opt = pyOpt.pyALPSO.ALPSO(pll_type='DPM') #this requires DPM, which is a parallel implementation
        opt = pyOpt.pyALPSO.ALPSO()
    if nonderivative_line_search==True:
        opt.setOption('Nonderivative linesearch')
    if FD == 'parallel':
        outputs = opt(opt_prob, sens_type='FD',sens_mode='pgc')
        
    elif solver == 'SNOPT' or solver == 'SLSQP':
        outputs = opt(opt_prob, sens_type='FD', sens_step = sense_step)
  
    else:
        outputs = opt(opt_prob)        
   
    return outputs



# ----------------------------------------------------------------------
#  Problem Wrapper
# ---------------------------------------------------------------------- 


def PyOpt_Problem(problem,x):
    """ This wrapper runs the RCAIDE problem and is called by the PyOpt solver.
        Prints the inputs (x) as well as the objective values and constraints.
        If any values produce NaN then a fail flag is thrown.

        Assumptions:
        None

        Source:
        N/A

        Inputs:
        problem   [nexus()]
        x         [array]

        Outputs:
        obj       [float]
        cons      [array]
        fail      [bool]

        Properties Used:
        None
    """      
   
    obj   = problem.objective(x)
    const = problem.all_constraints(x).tolist()
    fail  = np.array(np.isnan(obj.tolist()) or np.isnan(np.array(const).any())).astype(int)

       
    print('Inputs')
    print(x)
    print('Obj')
    print(obj)
    print('Con')
    print(const)
   
    return obj,const,fail