This avoids the issue, where the efficiency stayed constant until the first non-zero 'rb' defined in the efficiency data.

Fixing an issue, where the "v_online" simulation starting boundaries were overwritten by later definitions in "3c_setVariableLimits.gms".

Also some fine tuning to the indexing of the ramp variables, as well as the way "uft_online" is defined.

Created 3 new variables and 2 new equations to permit imposing ramp constraints and costs on units. Also generalized the "resdirection" set as an "up_down" set, also used for ramp related variables.

IMPORTANT! The dummy data in "3a_modelsInit.gms" regarding the reserve requirements needs to be changed in order for these changes to function!

The new "v_genRamp" records the change in the "v_gen" over an interval, as required in the "q_genRamp" equality constraint. The "v_genRamp" variable is generated for units with ramp restrictions, and is used for enforcing those in the model. However, in order to apply larger costs to steeper ramps, another variable called "v_genRampChange" is required, that essentially tracks the changes in the ramp speed of the units according to the "q_genRampChange" equality constraint. Since this variable is used to apply the costs, it requires the "up_down" dimension to keep track of positive and negative changes separately.

In order to make the ramping constraints as general and specific as possible, they can be imposed for each unit output separately. The old ramp related parameters were thus moved from "param_unit" under "param_gnu".

Adding a new model parameter called "intervalInHours", that defines the length of a single time step (index) in hours.

Major rewriting and cleanup of the efficiency approximations (directOff, directOn, lambda), and some of the related sets.

Previously, the efficiency approximations required specific efficiency data, and no more than 2 points of data were used. Now the approximations use all the relevant input data they can find (up to the defined maximum indeces) when determining the parameters, and tests have been implemented in "1e_inputs.gms" to abort Backbone execution and warn the user of incompatible input data with the desired approximations.

However, without minimum load, there are still issues with the "directOn" and "lambda" approximations that could perhaps be solved.

Previously "ts_absolute" and "ts_energyDemand" were kind of redundant, and the name "ts_absolute" wasn't all that intuitive. The two time series are now replaced by a single time series called "ts_influx", that hopefully gets its point accross even slightly better.

The SOS1 based efficiency approximation would require further variables in order to correctly restrain its behavior. This would most likely render it just as computationally demanding as the SOS2 based "lambda" approximation, but less accurate.

Implemented a new set called "uft_online", that oversees that the "v_online" and "v_startup" variables are only used when necessary.

Beforehand, the "v_online" and "v_startup" variables were defined based on the "unit_online" set, which was not dependent on "sufts", resulting in some erronous behaviour.

Renamed the "param_nu" to "param_unit" as it is currently used in "p_unit", which is not dependent on the node. Also commented out some unused/redundant assets in the "1a_definitions.gms", and made appropriate changes to "1e_inputs.gms" (since "p_gnu" efficiency and conversion related parameters don't affect anything anymore).

Since the last timesteps of the individual optimization cycles are now defined separately to ensure state variable boundaries, it is unnecessary to account for it in the loops as well.

Fixed a small issue, where there would be divisions by zero if the optimizaiton horizon was set to an end of a defined interval, which wasn't the last defined interval.

This is nothing crucial, as intervals past the optimization horizon probably shouldn't be defined at all. However, makes it a little more convenient to experiment with the model settings. As the last step is defined separately in order to correctly set the end point state boundaries anyway, there is no need to account for it in the interval loop.

Fixed an issue where the unit aggregation horizon would not move along with the moving optimization.

Added "rampSched" conditionals to the end step parameters that need to be defined only if "rampSched" is on.

Changed all energy time-series to power time-series, as well as made "vq_gen" dummy variables into power terms in "q_balance".

Also made some clarifications to the time-series desctriptions, as well as removed "ts_import" from the model, since it was no longer used for anything.

Fixing an index issue that caused the data reading loop to fail if the interval started with known data but ended with reused data.

Added definitions for the data at the last simulation step, otherwise e.g. temperature boundaries reset to 0 during the last time step.

CAUTION! Currently, doesn't average or sum anything, and uses only the raw data. This is apparent from the energy demand summation, and hopefully doesn't cause any weird problems. The generation and reserve variables aren't generated for the very last time index, so that at least should work as intended.

Added aliases for the other aggregated time-seres data, as it seems to be the way these are handled in Backbone.

Had forgotten to define "ts_nodeState" and "ts_unit" for when the interval length is one. Didn't seem to hinder the simulations though, but added them just to be safe. Also changed "gn" to "gn_state" for "ts_nodeState" definitions.

Added "ts_nodeState" and "ts_unit" to the averaging over calculation intervals. Time-dependent efficiency parameters seem to be working as intended.

Intermediate commit for implementing time series form unit parameters. HAS NOT BEEN TESTED WITH ACTUAL TIME SERIES DATA  AS OF YET!

First steps for implementing the possibility of time series form input data for units, required for e.g. heat pumps. The idea is to use a flag "useTimeseries" in p_unit to easily disable the use of time series when necessary. However, the flag is not parameter dependent, meaning that if the flag is set all available time series data for each unit will be used. The code should use time series data only when able, and supplement missing values with constant parameter data, which should make it possible (if not advised) to feed incomplete time series as input.

Created a new p_gn flag for cyclic bounds, forcing the states to have the same value on their first and last timesteps. CAUTION! I'm not very familiar with the temporal structure of Backbone, and cannot guarantee that my changes don't cause unintended side-effects.

Fixing problems in the new equations and their inputs. Works, but there are remaining bugs in the temporal structure of the model.

Changing the way constant and timeseries bounds on state variables are defined. Also updating the equations on state slack variables and reserve use with state variables.

1) Units and nodes can be aggregated in later time steps
2) Piecewise linear implementations (SOS1 and SOS2) for part-load conversion efficiency
3) Piecewise linear implementation can change in later time periods

Method to aggregate nodes or units for later time periods in the model. Data driven using unitUnit_aggregate. Aggregation point defined by mSettings(m, 't_aggregate').

Big name convention clean-up. Most parameters now start with p_... Underscore in sets is used to draw the line between dimensions and explanatory text. Dimensions with one letter shorthands are always first in the name, even when they are not first dimensions.

New calculation for averaging time series (much faster) and circular time for the models and time series

Corrected ts_energyDemand_ and ts_import_ calculation so that they will equal to the accrued energy over the time step.