From GridLAB-D Wiki
Jump to: navigation, search

Expectation object/property

Instead of having a separate expectation_object and expectation_property, the reference could be combined into a single expectation_reference with a format like object:property. Ditto for standard deviation. --Dchassin 21:41, 13 May 2010 (UTC)

  1. It could be, but the space-savings would require more time to parse through the string, and introduce complexity where it isn't needed. It ain't broke, and it's pretty hard to break it from how it is now. Mhauer 18:38, 17 May 2010 (UTC)

Double_ramp mode

The original purpose for creating the double_controller was that an individual controller did not have the available information to work cooperatively with a second controller, due to ranking and execution order problems. Will there be extra fields to provide double_ramp mode enough information to run, or is a ternary option suggested? -Mhauer 19:09, 17 May 2010 (UTC)

mean_observation_property

Doesn't the ramp mode take the mean price as an input to determine how relatively high or low the price is? Mhauer 20:07, 11 June 2010 (UTC)

  1. No, but the expectation property is normally pointed at the mean price for TOU and fixed pricing. Mhauer 20:15, 11 June 2010 (UTC)
  1. Correct. Right now, average and standard deviation of price are the only ones specifically handled. However, by leaving it generically as expectation property, other values can be used for defining the ramp rates - such as frequency or an independent regulation signal. Of course, this would be up to the user to correctly set this up, and we will only be specifically handling and testing price deviations versus temperature. jcfuller 20:19, 11 June 2010 (UTC)

Simple mode

Should there be specs for simple mode in the passive controller? The way it's written implies that there are only simple mode options for the transactive controller. Mhauer 22:43, 11 June 2010 (UTC)

  1. That's a good point. I had envisioned only for the active controller, but it does make sense for the passive ones also. I'll add that into the spec when I get a chance jcfuller 14:47 June 14 (PST)

Probability_off mode

First, 'k_w' is not defined within the spec. Undefined variables bad. Second, should the uniform mode use (P_clear - P_avg) / P_avg, or (P_clear - P_avg) / P_stdev ? Using deviation instead of the average gives the result a much greater dynamic range for large average values with a tight distribution. Mhauer 19:07, 14 June 2010 (UTC)

  1. See comfort_level for 'k_w' (they are the same thing).
  2. Should be as written. A greater dynamic range can be achieved by the user adjusting the comfort_level ('k_w').