RedispReward returns reward greater than the reward_max

[mlanden]

Environment

• Grid2op version: 1.4.0
• System: Red Hat Enterprise Linux Server 7.9 (Maipo)"
• Additional system information

Bug description

The redispReward is return rewards that are greater than the maximum reward calculated.

How to reproduce

Create environment with default reward and compare the rewards returned to the max reward computed in the initiative function.

### Code snippet
<!--Expose the python code you want us to test-->
```python
import grid2op

env = grid2op.make()
_, reward, _, _ = env.step(env.action_space({}))

The output of the code snippet above

The reward returned is greater than the reward_max computed in initialize. 

[BDonnot]

Hello,
Thanks for noticing, l will try to update it asap.

It should not be a problem, maybe in the mean time you can have a look at other environments, that should behave more normally i hope.

The other env. close to the default one is

import grid2op
env_name="l2rpn_case14_realistic"
env = grid2op.make(en_name) 

[mlanden]

Unfortunately, that does not seem to be the case. with the ieee 14 bus, I still see

Max: 706.4000244140625, Reward: 1085.8973388671875 redispach 1085.8973388671875, overflow 0.0

where the overflow reward is what I'm adding, but it is not relevant here because it is 0. I am applying interpolation to this reward so the max should be accurate to get correct results from np.interp.

[BDonnot]

For me, for later resuse:

import grid2op
env_name="l2rpn_case14_sandbox"
env = grid2op.make(env_name)
obs = env.reset()
obs, reward, done, info = env.step(env.action_space())
print(f"obtained reward: {reward:.2f}")
print(f"max reward: {env.reward_range[1]:.2f}")

So indeed there is problem with the way the reward max is computed for this specific case.