Skip to content

Latest commit

 

History

History
230 lines (219 loc) · 13 KB

MATPOWER-Release-Notes-7.1.md

File metadata and controls

230 lines (219 loc) · 13 KB

What's New in MATPOWER 7.1

Released Oct 8, 2020

Below is a summary of the changes since version 7.0 of MATPOWER. See the CHANGES.md file for all the gory details. For release notes for previous versions, see Appendix H of the MATPOWER User's Manual.

New Features:

  • MP-Opt-Model 3.0 brings many new enhancements to the opt_model class and the various optimization solver interfaces, including:

    • New unified interface nlps_master() for nonlinear programming problems, including implementations for MIPS, fmincon, IPOPT and Artelys Knitro.
    • New unified interface nleqs_master() for solving nonlinear equations, including implementations for fsolve, Newton's method, fast-decoupled Newton and Gauss-Seidel.
    • Automatic determination of explicit model type from characteristics of variables, costs and constraints.
    • New solve() method for opt_model to call appropriate solver for the respective problem type, including for linear and nonlinear equations.
    • Performance improvements.
    • Support for new solvers (OSQP, fsolve, Newton, fast-decoupled Newton, Gauss-Seidel) and new versions of CPLEX and Artelys Knitro.
    • Support for modifying parameters of an existing model.
    • Support for extracting specific variables, costs, constraint values and shadow prices, etc. from a solved model.

    For more details on improvements related to MP-Opt-Model, see the release notes (also in Appendix C in the MP-Opt-Model User's Manual) for MP-Opt-Model 0.8, 1.0, 2.0, 2.1, and 3.0.

  • MP-Test 7.1, with new modular, extensible have_feature() function for detecting optional functionality. For more details, see the MP-Test 7.1 release notes.

  • MIPS 1.4. For details, see the MIPS 1.4 release notes (also in Appendix C in the MIPS User's Manual).

  • Support for OSQP to solve LP and QP problems. Set option opf.dc.solver to 'OSQP' to use OSQP to solve the DC OPF. Requires the MATLAB interface to OSQP, available from https://osqp.org. See help mpoption for more OSQP options.

  • Option for makePTDF() to compute shift factors very efficiently for specific transfers or specific buses only.

  • New options:

    • pf.alg now accepts 'NR-SP' as a shortcut for Newton's method with the default power/polar formulation, and 'NR-SH' and 'NR-IH' for those with the hybrid voltage updates.
    • pf.v_cartesian now accepts 2 as a valid value to select the Newton power flow with hybrid voltage update.
    • opf.dc.solver now accepts 'OSQP' as a valid value if OSQP is installed, to select OSQP to solve the DC OPF.
    • osqp.opts overrides default OSQP options.

  • New functions/methods:

    • MP-Test 7.1
      • have_feature() detects availability and version information for optional functionality. This is a modular, extensible replacement for have_fcn() where the detection of a feature named <tag> is implemented by the function have_feature_<tag>().
    • MP-Opt-Model 3.0
      • mpopt2nleqopt() creates or modifies an options struct for nleqs_master() from a MATPOWER options struct.
      • mpopt2nlpopt() creates or modifies an options struct for nlps_master() from a MATPOWER options struct.
      • nleqs_master() provides a single wrapper function for calling any of MP-Opt-Model's nonlinear equation solvers.
      • nlps_master() provides a single wrapper function for calling any of MP-Opt-Model's nonlinear programming solvers.
      • opt_model.eval_lin_constraint() method computes the constraint values for the full set or an individual named subset of linear constraints.
      • opt_model.get_soln() method extracts solved results for a given named set of variables, constraints or costs.
      • opt_model.params_nln_constraint() method returns parameters for a named (and optionally indexed) set of nonlinear constraints.
      • opt_model.params_nln_cost() method returns parameters for a named (and optionally indexed) set of general nonlinear costs.
      • opt_model.parse_soln() method returns a complete set of solution vector and shadow price values for a solved model.
      • opt_model.problem_type() determines the type of problem based on the characteristics of the variables, costs and constraints in the model.
      • opt_model.set_params() method modifies parameters for a given named set of existing variables, costs, or constraints of an MP-Opt-Model object.
      • opt_model.solve() solves the model using mplinsolve(), qps_master(), miqps_master(), nleqs_master() or nlps_master(), depending on the problem type.
      • osqp_options() initializes options for OSQP solver.
      • osqpver() returns/displays version information for OSQP.
      • ... plus nleqs_core(), nleqs_fd_newton(), nleqs_fsolve(), nleqs_gauss_seidel(), nleqs_newton(), nlps_fmincon(), nlps_ipopt(), nlps_knitro(), opt_model.copy(), opt_model.is_mixed_integer(), and qps_osqp().

  • nlpopf_solver() implements the AC OPF solver based on opt_model's new solve() method, and replaces the individual solver-specific functions fmincopf_solver(), ipoptopf_solver(), ktropf_solver(), and mipsopf_solver().

  • ... plus 45 individual feature detection functions for have_feature() (10 in MATPOWER proper, plus 2 in MP-Test, 4 in MIPS, and 29 in MP-Opt-Model (see Table A-7 in the MP-Opt-Model User's Manual for details)..

New Case Files:

  • Nineteen new distribution system cases [ref 1, ref 2]. Thanks to Houssem Bouchekara, et. al.
    • case10ba -- 10-bus radial system from Baghzouz and Ertem
    • case12da -- 12-bus radial system from Das, Nagi, and Kothari
    • case15da -- 15-bus radial system from Das, Kothari, and Kalam
    • case15nbr -- 15-bus radial system from Battu, Abhyankar, Senroy
    • case16am -- 16-bus radial system from Das, Kothari, and Kalam
    • case16ci -- 16-bus system from Civanlar, Grainger, Yin, and Lee
    • case17me -- 17-bus radial system from Mendoza, Morales, Lopez, et. al.
    • case18nbr -- 18-bus radial system from Battu, Abhyankar, Senroy
    • case28da -- 28-bus radial system from Das, Nagi, and Kothari
    • case33mg -- 33-bus radial system from Kashem, et. al.
    • case34sa -- 34-bus radial system from Salama and Chikhani
    • case38si -- 38-bus radial system from Singh and Misra
    • case51ga -- 51-bus radial system from Gampa and Das
    • case51he -- 51-bus radial system from Hengsritawat, et. al.
    • case70da -- 70-bus system from Das
    • case74ds -- 74-bus radial system from Myint and Naing
    • case94pi -- 94-bus radial system from Pires, Antunes and Martins
    • case118zh -- 118-bus radial system from Zhang, Fu and Zhang
    • case136ma -- 136-bus radial system from Mantovani, Casari and Romero

New Documentation:

Other Improvements:

  • Refactor opt_model class to inherit from new abstract base class mp_idx_manager which can be used to manage the indexing of other sets of parameters, etc. in other contexts.
  • Add to opt_model.eval_nln_constraint() method the ability to compute constraints for a single named set.
  • Significant performance improvement for some problems when constructing sparse matrices for linear constraints or quadratic costs (e.g. during problem setup in MOST). Thanks to Daniel Muldrew.
  • Move original implementation of Newton power flow for cartesian voltage representations to newtonpf_S_hybrid() and newtonpf_I_hybrid(), accessible by setting the pf.v_cartesian option to 2. The pf.alg option can also be set to 'NR-SH' or 'NR-IH', respectively, as shortcuts to select these formulations.
  • Improve robustness of these hybrid Newton power flow formulations to avoid ill-conditioning when the real part of the voltage approaches zero.
  • Redesign implementation of Newton power flow for cartesian voltage representations to use standard Newton method with explicit voltage magnitude constraints at PV buses. This method is slower (larger number of equations and unknowns) but appears to be more reliable in some cases than the original implementation, which was a hybrid approach by Sereeter that used a modified Newton method to compute polar voltage updates using a modified cartesian Jacobian.
  • Modify voltage limit constraints for cartesian AC OPF formulation to use voltage squared, resulting in simpler derivatives.
  • Performance improvement for makePTDF() for large cases, e.g. more than 70% improvement for case9241pegase.
  • Significant performance improvement for CPLEX on small problems by eliminating call to cplexoptimset(), which was a huge bottleneck.
  • Convert dcopf_solver() to use the new solver() method of opt_model instead of calling qps_matpower() manually.
  • Convert opf_execute() to use nlpopf_solver(), based on the new solver() method of opt_model, for AC OPF when using fmincon, IPOPT, Artelys Knitro, or MIPS.
  • Reduce memory usage in modcost() for very large systems. Thanks to Christoph Funke.
  • Deprecated functions:
    • have_fcn() -- use have_feature() from MP-Test instead.
    • qps_matpower() -- use qps_master() from MP-Opt-Model instead.
    • miqps_matpower() -- use miqps_master() from MP-Opt-Model instead.
  • Removed functions:
    • fmincopf_solver() -- functionality now covered by nlpopf_solver().
    • ipoptopf_solver() -- functionality now covered by nlpopf_solver().
    • ktropf_solver() -- functionality now covered by nlpopf_solver().
    • mipsopf_solver() -- functionality now covered by nlpopf_solver().

Bugs Fixed:

  • For opt_model, incorrect evaluation of constant term has been fixed for vector valued quadratic costs with constant term supplied as a vector.
  • Calling opt_model.params_var() method with empty idx no longer results in fatal error.
  • Fix bug in scale_load() where only one of multiple dispatchable loads at a bus would be scaled. Thanks to Christoph Funke.
  • Fix bug #77 where incorrect indexing could cause fatal error in OPF with additional nonlinear constraints. Thanks to Sergio Garcia.
  • Fix OPF issue #71 for IPOPT and Artelys Knitro where structure of Jacobian and/or Hessian could change from the structure provided (i.e. elements with value of zero were being eliminated from the sparsity structure). Thanks to Drosos Kourounis.
  • Artelys Knitro 12.1 compatibility fix.
  • Fix CPLEX 12.10 compatibility issue #90.
  • Fix bug #89 where running a power flow with pf.enforce_q_lims enabled and voltage dependent ZIP loads produced incorrect results. Thanks to Florian.
  • Fix issue with missing objective function value from miqps_mosek() and qps_mosek() when return status is "Stalled at or near optimal solution."
  • Fix bug orginally in ktropf_solver() (code now moved to nlps_knitro()) where Artelys Knitro was still using fmincon options.

Incompatible Changes:

  • Modify order of default output arguments of opt_model.get_idx() (again), removing the one related to legacy costs.
  • MP-Opt-Model has renamed the following functions and modified the order of their input args so that the MP-Opt-Model object appears first. Ideally, these would be defined as methods of the opt_model class, but Octave 4.2 and earlier is not able to find them via a function handle (as used in the solve() method) if they are inherited by a subclass.
    • opf_consfcn() ––> nlp_consfcn()
    • opf_costfcn() ––> nlp_costfcn()
    • opf_hessfcn() ––> nlp_hessfcn()
  • Update case18, case22, case69, case85 and case141 to more closely match data from original papers, thanks in part to case files contributed by Houssem Bouchekara, et al. Solutions for updated cases may not match exactly. See help text in case files for details.