Wind Turbine¶

Hold the electrical power of the NREL 5 MW reference turbine at a curtailed setpoint in turbulent, above-rated wind, with blade pitch and generator torque. The rotor-effective wind is hidden and has to be inferred from how the rotor behaves, while rotor speed stays regulated. Costs are in dollars.

import target_gym as tg

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


Action space	`Box((2,))`, all actions in [-1, 1]
Observation space	`Box((5,))`
Tracked variable(s)	electrical power (MW)
Episode length	400 steps of 0.25 s (100 s)
Cite as	`wind_turbine-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	pitch_raw	-1	1
1	torque_raw	-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	rotor speed (rpm)
1	blade pitch (deg)
2	generator torque, percent of maximum
3	electrical power (MW)
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 400 steps.

Baselines¶

Recorded over 10 seeds of the 400-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	-0.05611	0.0001403
MPC	-0.005915	1.479e-05

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("wind_turbine", max_steps_in_episode=...).

parameter	value	unit
`delta_t`	0.25	s
`max_steps_in_episode`	400	steps
`R`	63	m
`rho`	1.225	kg/m^3
`J_rotor`	3.87592e+07	kg m^2
`J_gen`	534.116	kg m^2
`N_gear`	97	-
`eta_gen`	0.944	-
`P_rated`	5e+06	W
`v_rated`	11.4	m/s
`omega_rated_rpm`	12.1	rpm
`v_cut_out`	25	m/s
`cp_c1`	0.5176	-
`cp_c2`	116	-
`cp_c3`	0.4	1/deg
`cp_c4`	5	-
`cp_c5`	21	-
`cp_c6`	0.0068	-
`pitch_min`	0	deg
`pitch_max`	40	deg
`pitch_rate_max`	8	deg/s
`pitch_tau`	0.2	s
`torque_max`	47400	N m
`torque_tau`	0.1	s
`turbulence_std`	1.2	m/s
`v_min`	4	m/s
`v_max`	28	m/s
`overspeed_factor`	1.25	-
`underspeed_factor`	0.4	-
`power_band`	500000	W
`precision_floor`	1000	W
`power_envelope`	5e+06	W
`pitch_activity_weight`	0.02	-
`reward_version`	2	-
`e_floor`	1680	W
`e_tol`	0	W
`tracking_exponent`	1	-
`imbalance_price`	100	$/MWh
`c_hold`	0.0013	-
`fatigue_weight`	1	-
`failure_cost`	0.0972222	reward
`restart_steps`	2400	steps
`rho_floor_tracking`	1.00972e-05	reward
`rho_floor`	1.00972e-05	reward