BART_operations

Project Repository for BART Make and Break problem. An optimization solution to schedule the daily make and break operations of BART.

Notebooks

• steadyStateModel_baseline.ipynb:

  • Large linear program. Loads and structures pre-processed data. Optimizes for cost while adhering to a complex set of inventory, demand, and manpower constraints across the BART system. Determines optimal number of makes/breaks, the total costs, operating costs per mile, yard team costs, and the make/break costs. Exports results for easier analysis. Runs sensitivity analysis on model paramaters.

• steadyStateModel_all10.ipynb:

  • Large linear program. Performs same functions as above notebook. Runs the model assuming all consists are 10 cars in length.

• steadyStateModel_green.ipynb:

  • Large linear program. Performs same functions as above notebook. Runs the model with a focus on minimizing operating hours for consists in the system.

Assumptions

• Consists will contain only 5 or 10 cars.
• The 1,081 train cars will all be either D or E cars.
• The system layout is the same as it currently is (i.e. no Berryessa extension and no yard openings/closures).
• There will be a 2:3 ratio of D to E cars
• Fleet consists of 5-car modules that can move independently, and two such modules can be coupled and decoupled (DEEEDDEEED → DEEED + DEEED).
• Power and maintenance costs are both linear functions of the distance a train car travels.
• Makes and breaks only occur at yards or tail tracks at the end of a route (i.e. a train traveling from
• Warm Springs to Richmond will not stop at the Hayward yard to be shortened or lengthened).
• There will always be enough yard/tail track space for a make/break to be carried out.
• Inventory fulfillment of train-cars are carried out at each yard/tail track. Rather than breaking a 10-car consist arriving to the yard to send out a 5-car consist, the 10-car consist goes to inventory while a 5-car consist is sent out from inventory, if available. This avoids unnecessary making and breaking costs.

Data

• alpha.pkl

  • Train schedule departure times.

• beta.pkl

  • Train schedule arrival times.

• intervals - intervals.csv

  • Train line data. Time to travel, miles between stations.

• all_yards_indexes all_yards_indexes.csv

  • Yard inventory data

• passenger_flow - passenger.csv

  • Passenger demand data. Week day by hour. Number of passengers.

Pre-Processing NoteBooks

• PreProcessData_v3.ipynb
  • Generate departure and arrival times from raw BART schedule data. PreProcess inventory yard event table. Generate data structures.

Presentation

See presentation directory.

Report

See report directory.