Skip to content

Latest commit

 

History

History
244 lines (223 loc) · 14.2 KB

CHANGELOG.md

File metadata and controls

244 lines (223 loc) · 14.2 KB

Switch 2.0.5

This release standardizes all inputs and outputs as .csv files.

As usual, when you first solve an older model, Switch will prompt to backup and upgrade the inputs directory. If you accept, it will convert the existing tab-delimited *.tab files and most ampl-format *.dat files to comma-delimited *.csv files. It is recommended that you update your model data preparation scripts to create .csv files directly. Note that non-indexed parameters should now be stored in .csv files with a header row listing the parameter names and a single data row showing their values.

All multi-value outputs from Switch are also now in comma-delimited .csv files, instead of a mix of .csv, .tab and .txt files. (total_cost.txt is unchanged)

This release also includes includes the following minor updates:

  • Updated installation instructions
  • Switch version number and website are now shown in the startup banner when running with --verbose flag; solve messages have also been improved slightly
  • Some parsing errors for *.tab files have been fixed in the upgrade scripts; this may cause errors during the upgrade process for input files that use spaces instead of tabs and were previously upgraded by Switch, producing malformed files.
  • Fixed several bugs in the documentation and execution of the stochastic examples that use the PySP module of the Pyomo package

Switch 2.0.4

This release introduces compatibility with Python 3. As of version 2.0.4, Switch can now be run with either Python 2.7 or Python 3 (likely to work with 2.7.10+; has been tested on 2.7.16 and 3.7.3).

This release will prompt to upgrade your model inputs directory, but the only change it makes is to update switch_inputs_version.txt to 2.0.4.

This release includes the following updates:

  • Code has been updated in many places to achieve Python 2/3 cross-compatibility. Future contributors should ensure that their code is compatible with both Python 2 and 3 (e.g., use switch_model.utilities.iteritems(dict) instead of dict.iteritems(), be prepared for results from dict.keys(), dict.vars(), map(), range(), zip(), etc., to be either generators or lists, and use from __future__ import division whenever doing division).
  • Installation instructions in INSTALL have been updated. We now recommend that users install dependencies using the conda command first, then install Switch using pip. This follows practices recommended in https://www.anaconda.com/using-pip-in-a-conda-environment/ and should minimize problems caused by incompatibilities between conda and pip.
  • Output files (.csv, .tab, .tsv, and .txt) are now consistently written using the local system's line endings (LF on Mac or Linux, CRLF on Windows). Previously, most of these were written with only LF line endings on Windows.
  • A bug was fixed in switch_model.transmission.local_td that prevented the carrying cost of Legacy local T&D capacity from being included in the objective function. As a result, users of this module will find that Switch now reports higher total costs than previously. However, this should not affect any of the decisions that Switch makes.
  • To make switch_model.transmission.local_td module compatible with Python 3, "Legacy" was removed from the list of build years for local T&D capacity (Pyomo sorts index keys when solving the model, and Python 3 cannot sort lists that mix strings and numbers). Legacy capacity is now read directly from the existing_local_td[z] parameter when needed. This does not change the behavior of Switch, but "Legacy" rows are no longer written to the BuildLocalTD.tab output file. The LOCAL_TD_BLD_YRS set has also been removed. LOAD_ZONES * PERIODS can be used instead.
  • A new indexed set, CURRENT_AND_PRIOR_PERIODS_FOR_PERIOD[p] has been added. This is useful for simple online capacity calculations for assets that cannot be retired during the study (e.g., AssetCapacity[p] = sum(BuildCapacity[v] for v in CURRENT_AND_PRIOR_PERIODS_FOR_PERIOD[p]))
  • Code has been cleaned up a bit internally (e.g., removed trailing whitespace, changed "SWITCH" or "SWITCH-Pyomo" to "Switch")

Switch 2.0.3

  • Users can now provide data in variable_capacity_factors.tab and hydro_timeseries.tab for times before projects are built or after they are retired without raising an error. However, the extra datapoints will be ignored.
  • Various parts of the code have better formatting, documentation and performance.
  • switch_model.hawaii.smooth_dispatch is now compatible with Pyomo 5.6 and later.
  • A new '--exact' option in switch_model.hawaii.rps forces the system to exactly meet the RPS target and no more. This is useful for studying the cost of adopting various levels of renewable power, including levels below the least-cost system design (i.e., cases where low shares of renewable power cause higher system costs).
  • A bug was fixed when calculating the cost of water spillage in switch_model.generators.extensions.hydro_system.
  • Final reservoir level in switch_model.generators.extensions.hydro_system is now stored in a varaible called ReservoirFinalVol. The ReservoirSurplus variable has been eliminated.
  • Bounds on a number of inputs have been relaxed to allow unusual or edge cases. In particular, a number of variables can now be zero instead of strictly positive. This allows zero costs, zero capacity limits, zero-based year counting, etc.
  • The gen_is_baseload parameter is now optional, with a default value of False (0).
  • NEW_TRANS_BLD_YRS has been renamed to TRANS_BLD_YRS.
  • setup.py now lists an optional dependency on rpy2<3.9 instead of rpy2, because later versions of rpy2 require Python 3, which Switch doesn't support yet. This only affects the iterative demand response module.
  • A new GENS_BY_ENERGY_SOURCE set can be used to identify all the generators that use any energy source, either a fuel or a non-fuel energy source. GENS_BY_FUEL and GENS_BY_NON_FUEL_ENERGY_SOURCE also still exist.
  • We have begun migrating toward using initialize instead of rule when initializing Pyomo components, and recommend that users do the same in their custom modules. This matches the current Pyomo API documentation. rule also works for now, but initialize should be more future proof.
  • The discrete-build requirement is now enforced on generators with predetermined build quantities, in addition to optimized generators.
  • The optional psycopg2 dependency has been changed to psycopg2-binary.
  • The --debug option now uses the ipdb debugger if available; otherwise it falls back to pdb.

Switch 2.0.2

  • General

    • Added --assign-current-version argument to switch upgrade. This is useful for updating version number in example directories to match current version of Switch, even if the data files don't need an upgrade.

  • Hawaii regional package

    • Fixed bug in hawaii.rps that would crash switch solve --help.

Switch 2.0.1

  • General

    • Switch is now compatible with Pyomo 5.6+ (in addition to earlier versions).
    • A new --no-post-solve option prevents all post-solve actions (e.g., saving variable results).
    • If the user specifies --reload-prior-solution, Switch now behaves as if it had just solved the model, i.e., after loading the solution, it runs post- solve code unless --no-post-solve is specified (useful for re-running reporting code), and it only drops to an interactive Python prompt if the user also specifies --interact.
    • A new --no-save-solution disables automatic solution-saving. This saves time and disk space for models that won't need to be reloaded later.
    • New --quiet and --no-stream-solver arguments cancel --verbose and --stream-solver.
    • A new "--save-expression[s] ..." argument can be used to save values for any Pyomo Expression object to a .tab file after the model is solved (similar to the automatic saving of variable values). This also works for Param objects.
    • The --include-module(s), --exclude-module(s), --save-expression(s), --suffix(es) and --scenario(s) flags can now be used repeatedly on the command line, in options.txt or in scenarios.txt. The include and exclude options will be applied in the order they are encountered, in options.txt, then scenarios.txt, then the command line.
    • A new --retrieve-cplex-mip-duals flag can be used to support retrieving duals for a MIP program from the cplex solver (users must also turn on the "duals") suffix. This flag patches the Pyomo-generated cplex command script to pass the "change problem fix" command to the solver and then solve a second time. This fixes integer variables at their final values, then re-solves to obtain duals. This flag is not needed with the cplexamp solver.
    • A new balancing.demand_response.iterative module has been added. This was formerly in the Hawaii regional package. This module performs iterative solutions with any convex demand system, based on a bid-response process.
    • New indexed sets have been added to allow efficient selection of groups of generators that use a particular technology, fuel or non-fuel energy source: GENS_BY_TECHNOLOGY, GENS_BY_FUEL, GENS_BY_NON_FUEL_ENERGY_SOURCE.
    • Generator capacity data is saved to gen_cap.tab instead of gen_cap.txt and rows are sorted if user specifies --sorted-output.
    • Even if a model has solver warnings, results will be reported and post-solve will be performed if a valid solution is available.
    • A more descriptive warning is given when switch_model.reporting finds an uninitialized variable.
    • A warning is given about potential errors parsing arguments in the form "--flag=value". Python's argument parsing module can make mistakes with these, so "--flag value" is a safer choice.
    • Switch now monkeypatches Pyomo to accelerate reloading prior solutions. Previously Pyomo 5.1.1 (and maybe others) took longer to load prior solutions than solving the model.
    • At startup, "switch solve-scenarios" will restart jobs that were previously interrupted after being started by the same worker (same --job-id argument or SWITCH_JOB_ID environment variable). Note that workers automatically pull scenarios from the scenario list file until there are none left to solve, and avoid solving scenarios that have been pulled by other workers. Each worker should be given a unique job ID, and this ID should be reused if the worker is terminated and restarted. The new behavior ensures that jobs are not abandoned if a worker is restarted.

  • Upgrade scripts

    • The upgrade scripts now report changes in module behavior or renamed modules while upgrading an inputs directory. This only reports changes to modules used in the current model.
    • The hawaii.reserves module is automatically replaced by balancing.operating_reserves.areas and balancing.operating_reserves.spinning_reserves in the module list.
    • The hawaii.demand_response module is replaced by balancing.demand_response.iterative and hawaii.r_demand_system is replaced by balancing.demand_response.iterative.r_demand_system in the module list.
    • "switch_mod" will not be changed to "switch_modelel" if a module file is upgraded from 2.0.0b1 to 2.0.0b2 twice.

  • Hawaii regional package

    • The hawaii.reserves module has been deprecated and the hawaii.demand_response module has been moved (see upgrade scripts)
    • Switch now places limits on down-reserves from pumped-storage hydro.
    • A new --rps-prefer-dist-pv option for hawaii.rps will prevent construction of new large PV until 90% of distributed PV potential has been developed.
    • Limits on load-shifting between hours in hawaii.demand_response_simple have been formalized.
    • The Hawaii storage energy cost calculation has been fixed.
    • Total production by energy source is reported by hawaii.save_results, ad-hoc technologies are added to production_by_technology.tsv, and hourly dispatch is disaggregated by non-fuel technologies.
    • Bugs have been fixed in reserve calculation for EVs and in hawaii.smooth_dispatch and hawaii.scenario_data.
    • hawaii.smooth_dispatch minimizes total inter-hour change instead of square of levels. The old quadratic smoothing method has been moved to hawaii.smooth_dispatch.quadratic (slow and possibly buggy).
    • The must-run requirement in hawaii.kalaeloa is turned off when RPS or EV share is above 75% (can be overridden by --run-kalaeloa-even-with-high-rps)
    • Support for nominal-dollar fuel price forecasts has been dropped from hawaii.scenario_data
    • A new --no-hydrogen flag can be used to deactivate the hydrogen module.
    • The hawaii.ev_advanced module now calculates vehicle fleet emissions.

Switch 2.0.0

First public release of Switch 2. This uses a similar framework to Switch 1, but with numerous improvements. The most significant are:

  • Written in Python instead of AMPL language
  • Modular approach, so components can be easily added or removed from model
  • Modeling of unit commitment and part load heat rates (optional)
  • Generalization of sample timeseries to have arbitrary length instead of single days
  • Standardized reporting, e.g., automatic export of all variable values