next_state, Reward, Done, Info = env.step(
control_input,
t,
setpoint=[set_point1, set_point2],
noise=False,
economics='mixed',
w_y1=1,
w_y2=0)
# RL Feedback
if t == rl.eval_feedback and t > 150:
rl.matrix_update(
action_index, Reward, state,
[env.y[t, 0] - set_point1, env.y[t, 1] - set_point2], 5)
tot_reward = tot_reward + Reward
rlist.append(tot_reward)
# Autosave Q, T, and NT matrices
rl.autosave(episode, 100)
if episode % 10 == 0:
print("Episode {} | Current Reward {}".format(episode, tot_reward))
env.plots(timestart=50, timestop=6000)
# plt.scatter(PID1.u[40:env.y.shape[0]], env.y[40:, 0])
# plt.show()
# plt.scatter(PID2.u[40:env.y.shape[0]], env.y[40:, 1])
# plt.show()
# Generate input tuple
control_input = np.array([[input_1, input_2]])
# Simulate next time
next_state, Reward, Done, Info = env.step(
control_input,
t,
setpoint=[set_point1, set_point2],
noise=False,
economics='distillate')
# RL Feedback
if t == rl.eval_feedback:
rl.matrix_update(
action_index, Reward, state,
env.y[t, :] - np.array([set_point1, set_point2]), 5)
tot_reward = tot_reward + Reward
rlist.append(tot_reward)
# Autosave Q, T, and NT matrices
rl.autosave(iteration, 250)
env.plots(timestart=50, timestop=5950)
# plt.scatter(PID1.u[40:env.y.shape[0]], env.y[40:, 0])
# plt.show()
# plt.scatter(PID2.u[40:env.y.shape[0]], env.y[40:, 1])
# plt.show()