Battery

Follow a grid dispatch schedule, one setpoint per 5-minute market block, with a 1 MW / 2 MWh lithium-ion battery. Every MWh moved wears the pack, and the running cost prices that wear in dollars at every step, so tracking is a trade against capacity fade. Leaving the 5 to 95 % charge window trips the pack, though exact tracking stays inside it over a scored episode.

import target_gym as tg

env, params = tg.make("battery")

Action space	Box((1,)), all actions in [-1, 1]
Observation space	Box((5,))
Tracked variable(s)	delivered power (MW)
Episode length	360 steps of 5 s (30 min)
Cite as	battery-v2
Physics	contract

Action space

Actions are normalised to [-1, 1] and mapped onto the plant's real actuator range inside the environment.

#	meaning	min	max
0	power	-1	1

Observation space

5 values in the plant's own units. The space declares no bounds (it spans the float32 range), so scale them before training. The reward scores obs[3] against obs[4].

#	meaning
0	soc, state of charge (0 to 1)
1	cell voltage (V)
2	pack temperature (°C)
3	delivered power (MW), positive when discharging
4	target power (MW)

Rewards

See the environment's compute_reward.

Every environment in this suite scores on one contract. The reward is minus the sum of three non-negative costs, -(tracking_cost + running_cost + failure_cost). The running cost charges consumption and the failure cost charges trips out of the operating envelope. The reward is never positive, and its scale differs by orders of magnitude between plants. Some plants are priced in dollars or euros and the rest are dimensionless; the per-plant summary says which. See Reward shaping for how each cost is built.

Starting state

reset samples the initial condition and the target; state has 9 fields.

Episode end

Termination. None by default. Leaving the operating envelope trips the plant: the step that leaves it is charged the trip cost and sets info["tripped"] and info["trip_cost"], and the plant restarts as reset_env would while the episode clock keeps running. Episodic use is optional: with continuing = False the plant is lost and terminated is raised on that step instead. See Reward shaping.

Truncation. After 360 steps.

Baselines

Recorded over 10 seeds of the 360-step episode (see Baseline results). The reward is a cost, so a return closer to zero is better.

controller	mean return	cost per step
PID	-1.276	0.003544
MPC	-0.3413	0.000948

The MPC beats the PID on 10 of 10 seeds and does not trip the plant on any of them.

Arguments

The scored configuration, which tg.make returns. A keyword argument changes one field, as in tg.make("battery", max_steps_in_episode=...).

parameter	value	unit
delta_t	5	s
max_steps_in_episode	360	steps
energy_nominal	7.2e+09	J
power_max	1e+06	W
n_series	192	-
capacity_As	9e+06	A s
R0	0.02	ohm
R1	0.01	ohm
C1	20000	F
ocv_a	3	V
ocv_b	1.15	V
ocv_c	0.3	V
ocv_d	12	-
ocv_e	0.05	V
ocv_f	8	-
ocv_g	0.85	-
C_thermal	9e+06	J/K
UA_thermal	3000	W/K
T_ambient	25	degC
k_calendar	3e-09	1/s
E_activation	20000	J/mol
T_ref_aging	298.15	K
k_cycle	1.5e-09	1/(kA s)
soc_min	0.05	-
soc_max	0.95	-
T_max	60	degC
V_cell_min	2.7	V
V_cell_max	4.25	V
power_band	150000	W
precision_floor	1000	W
max_step_cost	0.03	reward
cost_weight	0.1	-
degradation_weight	200000	reward
soc_comfort_weight	0.1	reward
dispatch_block_seconds	300	s
dispatch_range	800000	W
dispatch_noise_std	2000	W
reward_version	2	-
e_floor	1596	W
e_tol	0	W
tracking_exponent	1	-
imbalance_price	100	$/MWh
c_hold	1.92e-08	1/step
fade_price	300	$/kWh
pack_kWh	1692	kWh
failure_cost	0.5	reward
restart_steps	720	steps
rho_floor_tracking	0.000221667	reward
rho_floor	0.000221667	reward