Grid model ideas/wishes

[Bachibouzouk]

What input data are available to users?

• blackout average frequency + std(need default, because seldom available) usually available (for companies: https://data.worldbank.org/indicator/IC.ELC.OUTG)
• blackout average duration + std(need default, because seldom available) usually available (https://prosperitydata360.worldbank.org/en/indicator/EBRD+BEEPS+c8)
• coarse time period (morning (6-11), midday (11-14), afternoon (14-17), evening (17-20), night (20-6)) (), sometimes available , suggested format for optional input {(6,10):10, (10,11):13, (11,14):8}

What do we do with this data?

• generate number of blackouts and distribute them over the total time (usually one year just take the monthly frequency and extend to one year), respecting the statistical weights for month and day periods
• compute total number of blackouts and return number of blackout per month
• total blackout duration
• computed average deviation (and check it is similar to what the user gave)
• graph to see if black_outs happens when high demand stacked timeseries with demand, grid availability * capacity, all other production of electricity

in case of multiple scenarii, each scenario MUST use the same grid_availability

How do we model this in oemof-solph?

grid_availablity can be used in a converter component (oemof-solph) for efficiency, or directly in the source as fix/max argument (set flow[t,blackout]=0). One of the variante can cost more calculation

grid (max setzen) -- bus -- minigrid

grid -- bus --blackout(max setzen) --bus--minigrid

Other

Miscellaneous notes

• Random implies it is not worst or best-case blackout distribution for the system but just ANY distribution. System composition could vary largely between different runs
• Necessary to save blackout timeseries and reuse it to keep comparability
• Multiple runs of a site with DIFFERENT blackout timeseries could enhance certainty regarding system sizing
• Sometimes there is some knowledge about when blackouts tend to happen (rather mornings, rather during the day...), and that they are not completely random
  -- optional : Select time frames that are more likely with tickboxes [0/1] for a whole day
  -- Define weights/likelikoods in a timeseries of outage
  -- Define likelyhoods of outages during day [0... 100%]
• Use historical data? Identical, or define weights?

How it is modelled in offgridders

get_blackouts() from module E contains all info to simulate blackouts, this in turns calls oemof_extension_for_blackouts() and returns

    blackout_result = {
        GRID_RELIABILITY: grid_reliability,
        GRID_TOTAL_BLACKOUT_DURATION: total_grid_blackout_duration,
        GRID_NUMBER_OF_BLACKOUTS: number_of_blackouts,
    }

The actual series are calculated in another subcall of get_blackouts(): availability(), which also calls sub_routines :

• get_number_of_blackouts()
  The essential piece of code here is

  # Calculation of expected blackouts per analysed timeframe
  blackout_events_per_month = np.random.normal(
      loc=experiment[BLACKOUT_FREQUENCY],  # median value: blackout duration
      scale=experiment[BLACKOUT_FREQUENCY_STD_DEVIATION]
      * experiment[BLACKOUT_FREQUENCY],  # Standard deviation
      size=12,
  )  # random values for number of blackouts
  

• get_time_of_blackout_events()

  The essential piece of code here is

      #Choosing blackout event starts randomly from whole duration
      # (probability set by data_time_index and blackout_events_per_timeframe)
      time_of_blackout_events = pd.Series(
          [1 for i in range(0, len(date_time_index))], index=date_time_index
      ).sample(
          n=blackout_events_per_timeframe, replace=False  # number of events
      )  # no replacements
  

• get_blackout_event_durations()

  The essential piece of code here is

  blackout_event_durations = np.random.normal(
  
          loc=experiment[BLACKOUT_DURATION],  # median value: blackout duration
          scale=experiment[BLACKOUT_FREQUENCY_STD_DEVIATION]
          * experiment[BLACKOUT_DURATION],  # Standard deviation
          size=number_of_blackouts,
      )  # random values for number of blackouts
  

• availability_series()

availability_series() is where the grid_availability (pd.Series of 0 and 1 for each hour) sets 0 during blackouts and manages the blackout events overlapping

Potential useful papers

• https://ieeexplore.ieee.org/document/993975