This page highlights key changes in NEMO since its initial public release. For a full history of NEMO releases, including the code for each version, see the Releases page on NEMO's GitHub site.
HiGHS solver: Added HiGHS as an officially supported NEMO solver. HiGHS is a high-performance, open-source solver that works with linear programming, mixed-integer programming, and quadratic programming models. This enhancement includes incorporating HiGHS in the Julia system image that's distributed with the NEMO installer program.
Microarchitecture-specific optimizations: Introduced microarchitecture-specific optimizations in the Julia system image delivered by the NEMO installer. This enhancement can reduce NEMO’s start-up and calculation time on many computers. Currently, optimizations for Skylake, Cannon Lake, Ice Lake, Rocket Lake, Tiger Lake, and Alder Lake microarchitectures are included in the installer.
Bug fix - renewable energy targets with transmission: Fixed a bug in how NEMO handles renewable energy minimum production targets for fuels for which transmission modeling is enabled.
Controlling JuMP direct mode and bridging from configuration files: Added two configuration file options that control whether NEMO uses JuMP's direct mode and bridging features when calculating a scenario. Both options are in the
jumpdirectmodeenables/disables direct mode, while
jumpbridgesenables/disables bridging. These switches override the selection of direct mode and bridging specified in calls to
calculatescenario- i.e., in the
calculatescenario. If you enable direct mode or turn off bridging, NEMO bypasses default behavior of JuMP that enhances solver compatibility at the expense of performance. This can substantially reduce memory use and solve time, but it may lead to solver errors. See the documentation on configuration files, Performance tips, and JuMP's documentation for more details.
Region groups: Added a new dimension for groups of regions and parameters to assign regions to groups (
RRGroup) and to enforce renewable energy production targets for groups (
REMinProductionTargetRG). Each region can belong to zero or more groups. When renewable energy targets are applied at the group level, NEMO finds the least-cost way to distribute renewable production across the grouped regions.
JuMP upgrade: Updated NEMO to use JuMP 1.2.0, MathOptInterface 1.7.0, and the latest versions of the Julia interfaces (packages) for the Cbc, Gurobi, and Mosek solvers.
Other enhancements: Implemented miscellaneous robustness improvements for multithreading in NEMO. Updated NEMO to use the latest versions of some other Julia packages on which it depends, including SQLite and DataFrames.
To use NEMO 1.8 with LEAP, please ensure you are running LEAP version 2020.1.0.65 or higher.
- Julia platform upgrade: Updated NEMO to run on Julia 1.7.2, JuMP 1.0.0, and new versions of the Julia interfaces (packages) for the Cbc, CPLEX, GLPK, Gurobi, Mosek, and Xpress solvers. With this enhancement, NEMO can now be used with Gurobi 9.5, CPLEX 22.1, and Xpress 8.10-8.13, among other solver versions. See Solver compatibility for full details on compatible solvers.
To use NEMO 1.7 with LEAP, please ensure you are running LEAP version 2020.1.0.57 or higher.
Multi-threading: Implemented a new paradigm for parallel processing in NEMO - replaced distributed Julia processes with multi-threading. In addition to streamlining operations that were already parallelized, this change allowed parallelization of constraint creation and results saving in
calculatescenario. It should reduce run-time and memory use for most models. Deprecated the
writescenariomodel; they were used to select distributed Julia processes and are no longer needed. When parallel processing, NEMO uses as many threads as are available in the Julia session. This in turn is controlled by a command line argument or environment variable provided when invoking Julia (see Julia's documentation for details).
Improved functionality for renewable energy targets: Increased the flexibility and simplicity of NEMO's renewable energy target calculations. Added fuel as a dimension for
REMinProductionTarget, so this parameter now describes the fraction of a fuel's production in a region and year that must be by renewable technologies. As in earlier versions of NEMO, the
RETagTechnologyparameter determines the renewability of technologies. Modified the logic for verifying compliance with
REMinProductionTargetto exclude production from storage. Removed the
vretotalproductionoftargetfuelannualoutput variables since the new calculation method made them superfluous.
Variables for generation and renewable generation: Added output variables for annual generation (i.e., production excluding production from storage) and annual generation from renewable technologies. See
Minimum production share: Added
MinShareProduction, a parameter that specifies a technology's minimum share of production of a fuel in a given region and year (excluding production from storage).
Warm starts: Added two arguments to
writescenariomodelto enable warm starting optimization. NEMO's approach to warm starting is to take starting values for model output (decision) variables from a previously calculated scenario database. You specify the path to the previously calculated database with the
startvalsdbpathargument; and you can optionally control which variables get starting values with the
startvalsvarsargument. These arguments can be provided either on the command line or through a NEMO configuration file. See the documentation for
calculatescenariofor more details.
Bypassing optimization when results have already been calculated: Added an argument to
precalcresultspath, that allows you to identify a previously calculated scenario database that should be used in lieu of optimizing the main scenario database passed to the function (i.e., the database named in the
dbpathargument). If you specify a previously calculated database with
precalcresultspath, NEMO copies it over the database at
dbpath. This feature is intended for users who are calling NEMO through LEAP and want to compile results from multiple scenarios (which may have been calculated on different machines) in one LEAP area. You can use
precalcresultspatheither on the command line or through a NEMO configuration file. See the documentation for
calculatescenariofor more details.
Bug fix - transmission line efficiency: Corrected an error in how NEMO calculates transmission energy losses and variable costs when the transmission modeling type is 3 (pipeline flow; see
TransmissionModelingEnabled) and line efficiency is less than 1. The calculations now properly account for bidirectional flow through lines. As part of this fix, removed the
vtransmissionbylineannualoutput variable and added two variables for transmission variable costs:
Windows installer enhancements for multi-user environments: Revised the NEMO Windows installer program to facilitate installations in multi-user environments (i.e., when the installer is run by one Windows user, but NEMO will be run by different Windows users). Among other changes, the new installer program installs Julia for all users. As usual, it is not necessary to uninstall older versions of NEMO before running the installer.
Other improvements: Revised NEMO's documentation to note that nodal storage modeling requires transmission modeling to be enabled for a storage's input and output fuels. See
Interest rates for technologies, storage, and transmission lines: Added parameters for interest rates for technologies (
InterestRateTechnology), storage (
InterestRateStorage), and transmission lines (
TransmissionLine). If you specify an interest rate, NEMO uses it to calculate financing costs for new endogenous capacity, assuming the capital costs are financed at the interest rate and repaid in equal installments over the life of the capacity. The financing costs are reported in three new output variables (
vfinancecosttransmission) and enter into NEMO's cost minimization objective. They are also considered when calculating salvage values for technologies, storage, and transmission lines.
Negative emissions: Revised NEMO to accommodate negative emissions from technologies. You can activate this feature by specifying a negative emission factor (
EmissionActivityRatio). If a technology has a negative emission factor for a pollutant with an externality cost (
EmissionsPenalty), it can generate negative emission penalties (
vannualtechnologyemissionpenaltybyemission), lowering total system costs. In this case, you may need to constrain the technology's operation to avoid an unbounded (infeasible) optimization problem. For example, if a technology can generate negative emissions of a pollutant with an externality cost, the cost of building and running the technology is lower than the externality value, and there are no limits on the technology's deployment and use, the optimization problem will be unbounded.
Deprecation of technology, storage, and transmission-specific discount rates: Based on discussions with users, retired this functionality that was introduced in NEMO 1.5.
Other performance improvements: Improved the robustness of the
Technology, storage, and transmission-specific discount rates: Revised NEMO so users can specify a different discount rate for each technology and region, storage and region, and transmission line. Technology and storage-specific rates are set with the new
DiscountRateStorageparameters. Rates for transmission lines are defined as part of the transmission line dimension. The
DiscountRateparameter continues to provide the default discount rate for each region.
Minimum utilization: Added a parameter -
MinimumUtilization- that enforces minimum utilization rates for technology capacity.
Other performance improvements: Streamlined logic for scenario database upgrades.
Writing NEMO models: Added a function to write an output file representing the NEMO optimization problem for a scenario (
writescenariomodel). This function supports common solver file formats including MPS and LP and can compress its output with Gzip or BZip2. Results from the function can be used as an input to solver performance tuning programs, such as Gurobi's parameter tuning tool and CPLEX's tuning tool.
Calculating selected years: Added functionality to calculate selected years of a scenario. You can invoke this feature with a new
writescenariomodel). Calculating selected years can quickly provide results for large models. The results may not be identical to what you would get if you calculated all years, but NEMO uses several methods to reduce discrepancies between the two cases. See Calculating selected years for details.
New default solver - Cbc: Changed the default solver for
calculatescenarioto Cbc. Since NEMO version 1.3, Cbc generally offers better performance than GLPK for NEMO models.
New output variable - vtransmissionbylineannual: Added a variable that reports annual transmission through a transmission line in energy terms.
Bug fix - exogenous emissions: Revised
vannualemissionsso it includes any exogenously specified annual emissions (
vmodelperiodemissionsso it includes any exogenously specified annual and model period emissions (
Other performance improvements: Implemented various enhancements to improve NEMO's performance for large models, particularly those with multiple regions.
Solver parameters in NEMO configuration files: Added support for setting solver parameters via a NEMO configuration file. Users can activate this feature by assigning a comma-delimited list of parameter name-value pairs to the
parameterskey in a configuration file's
solverblock. The pairs should be in this form: parameter1=value1, parameter2=value2, .... See the documentation for configuration files for more information.
Forcing mixed-integer optimization problems: Added an option that forces NEMO to formulate a mixed-integer optimization problem when calculating a scenario. This can improve performance with some solvers. The option can be invoked as an argument passed to
forcemip) or in a NEMO configuration file (
calculatescenarioargsblock). See the documentation for
calculatescenarioand configuration files for more information.
- Julia and JuMP upgrade: Updated NEMO to run on Julia 1.5.3 and JuMP 0.21.6. The new version of JuMP includes support for the most recent versions of key solvers, among them Cbc (2.10), CPLEX (12.10 and 20.1), and Gurobi (9.0 and 9.1). Note that this version of JuMP also uses a new solver abstraction layer,
MathOptInterface, which changes how solvers are referenced when creating a JuMP model. See the documentation for
calculatescenariofor more information and examples.
Parallel processing upgrades: Revised
calculatescenarioso users can take advantage of parallelization without having to invoke Julia's
Distributedpackage and add processes manually. Introduced the
numprocsargument, which lets users specify the number of processes to use for parallelized operations. When
numprocsis set, NEMO initializes new processes as needed. Refactored the queries in
calculatescenarioto parallelize as many of them as possible.
Xpress solver: Added Xpress as an officially supported NEMO solver. This includes incorporating Xpress in the Julia system image that's distributed with the NEMO installer program.
Installer program enhancements: Upgraded the installer program to facilitate installation when the executing user isn't an operating system administrator. Also improved the integration of the installer program with LEAP.
General error handling in
calculatescenarioso exceptions are trapped and presented along with information on how to report problems to the NEMO team.
Other changes: Streamlined NEMO's logic for upgrading legacy database versions in
calculatescenario. Now the functions that perform upgrades are only called when needed. Removed the
createnemodb_leapfunction since LEAP isn't using it.