minor fix in reward calculation and some minor improvements

net/net_single-inv-curr.yaml

@@ -14,6 +14,6 @@ components:
     out:
       v: [lc1.capacitor1.v, lc1.capacitor2.v, lc1.capacitor3.v]
       i: [lc1.inductor1.i, lc1.inductor2.i, lc1.inductor3.i]
-    # iref: [0,0,0]
-    # vref: [1,0,0]
+    i_ref: [15,0,0]
+    # v_ref: [1,0,0]
 

openmodelica_microgrid_gym/aux_ctl/pi_controllers.py

@@ -39,7 +39,7 @@ def step(self, error: float) -> float:
                 PI_Parameters provided during initialisation, clipped due to the defined limits
         """
 
-        self.integralSum = self.integralSum + (self._params.kI * error + self.windup_compensation) * self._ts
+        self.integralSum += (self._params.kI * error + self.windup_compensation) * self._ts
         output = self._params.kP * error + self.integralSum
         clipped = np.clip(output, *self._params.limits)
         self.windup_compensation = (output - clipped) * self._params.kB

openmodelica_microgrid_gym/env/modelica.py

@@ -301,7 +301,7 @@ def step(self, action: Sequence) -> Tuple[np.ndarray, float, bool, Mapping]:
         else:
             logger.debug("Experiment step done, experiment done.")
 
-        reward = self.reward(self.history.cols, obs)
+        reward = self.reward(self.history.cols, outputs)
         self._failed = np.isnan(reward) or np.isinf(reward) and reward < 0 or reward is None
 
         # only return the state, the agent does not need the measurement