name: bess-rl
version: "1.0"
description: >
  Battery Energy Storage System (BESS) multi-objective reinforcement learning
  environment. Co-optimizes Energy Arbitrage, Frequency Regulation, and Peak
  Shaving on real PJM market data using a Soft Actor-Critic (SAC) agent.

server:
  host: "0.0.0.0"
  port: 8000

tasks:
  - id: easy
    description: "Energy Arbitrage only - learn to buy low and sell high on the PJM energy market."
    max_steps: 720
  - id: medium
    description: "Energy Arbitrage + Frequency Regulation - co-optimize market profit with grid stability."
    max_steps: 720
  - id: hard
    description: "Energy Arbitrage + Frequency Regulation + Peak Shaving - full multi-objective BESS co-optimization."
    max_steps: 720

action:
  type: array
  description: "Continuous actions [a_PS, a_EA, a_FR] each in range [-1.0, 1.0]"
  items:
    type: number
    minimum: -1.0
    maximum: 1.0
  minItems: 3
  maxItems: 3

observation:
  type: object
  properties:
    hour_of_day:
      type: number
      description: "Hour of day (0-23)"
    soc:
      type: number
      description: "State of Charge normalized (0.0-1.0)"
    price_lmp:
      type: number
      description: "Locational Marginal Price in $/MWh"
    p_avg:
      type: number
      description: "24-hour rolling average LMP in $/MWh"
    freq_regd:
      type: number
      description: "Frequency regulation signal (-1.0 to 1.0)"
    load_mw:
      type: number
      description: "Grid load in MW"