Updated-new yaml with extra tuning parameters

Author: musahin33

Examples/34_AEPS_Simulation.py

@@ -4,8 +4,8 @@
 Adaptive Envelope Protection System (AEPS) for Wind Turbines
 
 This example demonstrates the Adaptive Envelope Protection System (AEPS) for a wind turbine. It presents simulation results with the AEPS system disabled (ASO = 0) and enabled (ASO = 1).
-The AEPS is designed to protect turbines from excessive loads, i.e., excessive thrust force. It is an adaptive algorithm that learns the turbine’s operating conditions in real time and ensures its safe operation within a predefined thrust limit. By monitoring wind and turbine states such as wind speed, thrust force, and rotor speed, AEPS online adapts to the turbine’s operational conditions through an adaptive neural network (LPNN) and predicts the near-future thrust response.
-When the predicted thrust response is about to exceed the predefined thrust limit, AEPS collectively changes the turbine blade pitch angles to ensure safe operation. This is achieved by calculating a theoretical wind speed, referred to as the envelope wind speed, and comparing it with the actual wind speed to determine whether the turbine is operating under excessive thrust or not. If the actual wind speed is lower than the envelope wind speed, the turbine operates safely; otherwise, it experiences excessive thrust force.
+The AEPS is designed to protect turbines from excessive loads, i.e., excessive thrust force. It is an adaptive algorithm that learns the turbine’s operating conditions in real time and ensures its safe operation within a predefined thrust limit. By monitoring wind and turbine states such as wind speed, thrust force, and rotor speed, AEPS adapts online to the turbine’s operational conditions through an adaptive neural network (LPNN) and predicts the near-future thrust response.
+When the thrust response is about to exceed the predefined thrust limit, AEPS collectively changes the turbine blade pitch angles to ensure safe operation. This is achieved by calculating a theoretical wind speed, referred to as the envelope wind speed, and comparing it with the actual wind speed to determine whether the turbine is operating under excessive thrust or not. If the actual wind speed is lower than the envelope wind speed, the turbine operates safely; otherwise, it experiences excessive thrust force.
 When the thrust is close to exceeding the predefined limit, AEPS generates an additional blade pitch reference signal through a limit avoidance design parameter based on the offset between the actual and envelope wind speeds. This additional signal is added to the blade pitch controller output, thereby collectively increasing the blade pitch angles. This avoidance action reduces thrust force and keeps the turbine within the predefined safe limit, ensuring continued safe operation of the turbine.
 More information on AEPS can be found in:
 Sahin, M., Yavrucuk, I., Adaptive envelope protection control of wind turbines under varying operational conditions, Energy, Volume 247, 2022, 123544, ISSN 0360-5442, https://doi.org/10.1016/j.energy.2022.123544.
@@ -14,20 +14,21 @@
 Current AEPS Design and Implementation
 
 In the current implementation, different from the above reference, thrust and wind speed estimates, and filtered rotor speed are used in AEPS rather than their actual values. In addition, since blade pitch response exhibits a lag when AEPS applies an avoidance signal, the system is activated earlier: when the actual wind speed approaches the calculated envelope wind speed. Also, a squashing function is used for the NN inputs.
-Finally, the predefined thrust force is calculated based on the maximum thrust force of the turbine using an eps_percent parameter. When the user selects an eps_percent value and activates the AEPS system (ASO = 1), the algorithm determines whether the turbine will operate under excessive thrust conditions and changes the blade pitch controller output, only when necessary, thereby keeping the thrust force within the selected limit.
+Finally, the predefined thrust force is calculated based on the maximum thrust force of the turbine using an eps_percent parameter. When the user selects an eps_percent value and activates the AEPS system (ASO = 1), based on the selected AEPS activation time t_act for avoidance, the algorithm determines whether the turbine will operate under excessive thrust conditions and starts changing the blade pitch controller output, only when necessary, thereby adjusting the blade pitch angles and keeping the thrust force within the selected limit. AEPS is not recommended to be activated during learning phase, i.e at initial seconds or aggressive thrust behavior possibly occurring at the beginning of simulations.
 
 AEPS Tuning Procedure
 
-The AEPS implementation is realized such that the user can tune the following three parameters:
+The AEPS implementation is realized such that the user can tune the following parameters:
 •	Observer gain (Kc)
 •	Neural network learning rate (Γ / Gamma)
-•	e-modification term gain (k)
+•	e-modification term gain (ke)
+•	wind margin (Um)
 •	Effective limit avoidance parameter (e_dp)
-AEPS tuning can be started with selecting the observer gain (Kc) and neural network (NN) learning rate (Γ) when AEPS is disabled (ASO = 0). The key purpose is to obtain an AEPS-predicted thrust force (Thrst_est) that is very close to the turbine thrust force (Thrst). The designer can compare these two parameters (Thrst_est and Thrst). For ease of comparison, the designer can also check the error between them (T_error = Thrst – Thrst_est), which should be close to zero.
+AEPS tuning can be started with selecting the observer gain (Kc) and neural network (NN) learning rate (Γ) when AEPS is disabled (ASO = 0). The key purpose is to obtain an AEPS-predicted thrust force (Thrst_est) that is very close to the estimated turbine thrust force (Thrst). The designer can compare these two parameters (Thrst_est and Thrst). For ease of comparison, the designer can also check the error between them (T_error = Thrst – Thrst_est), which should be close to zero.
 Greater emphasis can be first placed on tuning the observer gain Kc, which controls the convergence rate of the algorithm. Values such as 2, 3, 4, 5 and so on can be tested, with the goal of minimizing T_error. Once a suitable Kc is chosen, the NN learning rate Γ can be adjusted to further reduce the error. The e-modification term gain (ke) is usually a small number such as 0.01, 0.02, etc. Nevertheless, it might take higher values such as 0.5, or 1. Small numbers such as 0.02 are recommended for ke.
-The learning rate Γ determines how quickly the neural network adapts to turbine operating conditions, i.e., predicting the turbine thrust force. A higher learning rate Γ results in faster learning but high oscillations in the AEPS-predicted thrust force. Once adaptation is achieved, the AEPS-predicted thrust (Thrst_est) should closely track the turbine thrust force (Thrst). Here, it should be also noted that the adaptation/learning is satisfied for a range of these tuning values. Therefore, the designer should decide all of these values based on the satisfied error.
-When the observer gain, learning rate and e-modification term gain are tuned as desired, AEPS can be activated (ASO = 1) to tune the effective limit avoidance parameter (e_dp). This parameter is highly sensitive to the turbine actuator. If the actuator model changes, e_dp may need to be re-tuned. It should be tuned such that the turbine thrust stays within the predefined thrust limit value. It should be kept in mind that choosing a high value for e_dp might cause oscillations in simulations.
-Finally, if desired, after AEPS activation, the designer may fine-tune the observer gain and learning rate, again by comparing AEPS-predicted thrust with the turbine thrust and monitoring the resulting error and also the limit avoidance parameter, e_dp by checking the turbine thrust force whether staying within the limit or riding the turbine approximately at the limit.
+The learning rate Γ determines how quickly the neural network adapts to turbine operating conditions, i.e., predicting the turbine thrust force. A higher learning rate Γ results in faster learning but high oscillations in the AEPS-predicted thrust force. Once adaptation is achieved, the AEPS-predicted thrust (Thrst_est) should closely track the estimated turbine thrust force (Thrst). Here, it should be also noted that the adaptation/learning is satisfied for a range of these tuning values. Therefore, the designer should decide all of these values based on the satisfied error.
+When the observer gain, learning rate and e-modification term gain are tuned as desired, AEPS can be activated (ASO = 1) to tune the effective limit avoidance parameter (e_dp). This parameter is highly sensitive to the turbine actuator. If the actuator model changes, e_dp may need to be re-tuned. It should be tuned such that the turbine thrust stays within the predefined thrust limit value. It should be kept in mind that choosing a high value for e_dp might cause oscillations in simulations. Also, the lag in the blade pitch response can also be compensated by the parameter Um, which helps AEPS generate an earlier avoidance signal in order to keep the turbine thrust within the limit. Um is just a wind margin showing how much close the wind speed estimate to the envelope wind speed.
+Finally, if desired, after AEPS activation, the designer may fine-tune the observer gain and learning rate, again by comparing AEPS-predicted thrust with the estimated turbine thrust and monitoring the resulting error and also the limit avoidance parameter, e_dp , wind margin Um by checking the turbine thrust force whether staying within the limit or riding the turbine approximately at the limit. 
 
 """
 

Examples/35_BEPS_Simulation.py

@@ -2,9 +2,13 @@
 35_BEPS_Simulation
 ------------------
 Basic Envelope Protection System (BEPS) for Wind Turbines
-This example demonstrates a basic Envelope Protection System for wind turbines. Compared to the AEPS system in Example 34, this system does not include any Neural Network or Observer and can be considered a simplified version.
-As in AEPS, an eps_percent parameter is defined in this algorithm, and the predefined thrust limit is calculated based on the turbine’s maximum thrust using this parameter. The offset between the predefined thrust limit (PreDf_Thrst) and the turbine thrust (Thrst) is used to calculate an avoidance signal through a limit avoidance design parameter, e_dp. This avoidance signal is then applied to modify the blade pitch controller output, i.e., the blade pitch reference, collectively increasing the blade pitch angles when the turbine approaches the limit. This avoidance action allows the turbine to operate safely within the defined thrust limit.
-The BEPS system generates an avoidance signal when the turbine thrust approaches within five percent of the predefined limit, accounting for the delay in blade pitch response after the avoidance signal is applied. In this system, the user only needs to tune the limit avoidance design parameter, e_dp. A value around 0.5 or less might be choosen for this parameter. A high value might cause oscillation in simulations.
+This example demonstrates a Basic Envelope Protection System for wind turbines. It presents simulation results with the BEPS system disabled (ASO = 0) and enabled (ASO = 2). Compared to the AEPS system in Example 34, this system does not include any Neural Network or Observer components and can be considered a simplified version.
+As in AEPS, an eps_percent parameter is defined in this algorithm, and the predefined thrust limit is calculated based on the turbine’s maximum thrust using this parameter. The offset between the predefined thrust limit (PreDf_Thrst) and the estimated turbine thrust (Thrst) is used to calculate an avoidance signal through a limit avoidance design parameter, e_dp. This avoidance signal is then applied to modify the blade pitch controller output, i.e., the blade pitch reference, collectively increasing the blade pitch angles when the turbine approaches the thrust limit. This avoidance action allows the turbine to operate safely within the defined thrust limit.
+The BEPS system starts generating an avoidance signal when the estimated turbine thrust approaches within a specified percentage (Tm) of the predefined thrust limit, accounting for the lag in blade pitch response when the avoidance signal is applied. 
+
+BEPS Tuning Procedure
+
+In this system, the user needs to tune the limit avoidance design parameter, e_dp. A value around 0.5 or less might be choosen for this parameter as higher values might cause oscillation in simulations. Another paramater to tune might be thrust margin parameter Tm, which is defined in percent and indicates how close the estimated turbine thrust is to the predefined thrust limit. The parameter Tm just decides when BEPS starts generating the avoidance signal and helps compensate for the lag in blade pitch response by applying the avoidance signal earlier to keep the turbine within the limit.
 
 """
 

Examples/Tune_Cases/IEA15MW_AEPS.yaml

@@ -61,6 +61,7 @@ controller_params:
   Kc:                 2                     # AEPS Observer Gain 
   gamma:              2400                  # AEPS Learning Rate
   ke:                 0.02                  # AEPS e-modification term gain
+  Um:                 1.0                   # Wind margin for AEPS avoidance
   e_dp:               0.032                 # AEPS design parameter for effective avoidance
-
+  t_act:              80                    # AEPS activation time for avoidance
 

Examples/Tune_Cases/IEA15MW_BEPS.yaml

@@ -58,6 +58,7 @@ controller_params:
   #Adaptive Safe Operation 
   #BEPS Parameters 
   eps_percent:        0.8                   # Percent Peak Reduction  [%, <= 1 ], {default = 80%}
+  Tm:                 0.05                  # Thrust margin for BEBS avoidance [%]
   e_dp:               0.4                   # BEPS design parameter for effective avoidance
 
 

docs/source/toolbox_input.rst

@@ -424,22 +424,42 @@ controller_params
 :code:`Kc` : Float
     AEPS Observer Gain
 
-    *Default* = 50    *Maximum* = 70 
+    *Default* = 2    *Minimum* = 1
 
 :code:`gamma` : Float
     AEPS Learning Rate
 
-    *Default* = 20    *Maximum* = 150 
+    *Default* = 2400  
+
+:code:`ke` : Float
+    AEPS e-modification term gain
+
+    *Default* = 0.02 
+
+:code:`Um` : Float
+    Wind margin for AEPS avoidance
+
+    *Default* = 1.0 
+
+:code:`Tm` : Float
+    Thrust margin for BEBS avoidance [%]
+
+    *Default* = 0.05
 
 :code:`e_dp` : Float
+    AEPS/BEPS design parameter for effective avoidance
+
+    *Default* = 0.032
+
+:code:`t_act` : Float
     AEPS design parameter for effective avoidance
 
-    *Default* = 0.1    *Maximum* = 5 
+    *Default* = 80 *Minimum* = 5
 
 :code:`PreDf_Thrst` : Float
     AEPS Pre-Defined Thrust Limit [MN]
 
-    *Default* = 0.5    *Maximum* = 5 
+    *Default* = 2.04356174e+00    
 
 :code:`sd_maxpit` : Float, rad
     Maximum blade pitch angle to initiate shutdown [rad], {default =

rosco/controller/rosco_registry/rosco_types.yaml

@@ -495,9 +495,18 @@ ControlParameters:
     ke:
       <<: *real
       description: AEPS e-modification term gain
+    Um:
+      <<: *real
+      description: Wind margin for AEPS avoidance
+    Tm:
+      <<: *real
+      description: Thrust margin for BEBS avoidance [%]
     e_dp:
       <<: *real
-      description: AEPS design parameter for effective avoidance 
+      description: AEPS/BEPS design parameter for effective avoidance     
+    t_act:
+      <<: *real
+      description: AEPS activation time for avoidance
     PreDf_Thrst:
       <<: *real
       description: AEPS Pre-Defined Thrust Limit [MN] 

rosco/controller/src/ControllerBlocks.f90

@@ -229,24 +229,22 @@ SUBROUTINE AdaptiveEnvelopeProtectionSystem(LocalVar, CntrPar, objInst, PerfData
                     if (CntrPar%ASO_Mode == 0) then
                        LocalVar%Del_Beta = 0
 
-                    !else if (CntrPar%ASO_Mode == 1) then
-                    !else if ((CntrPar%ASO_Mode == 1) and (LocalVar%time >= 80)) then
-                    else if (CntrPar%ASO_Mode == 1 .and. LocalVar%time >= 80) then
+                    else if (CntrPar%ASO_Mode == 1 .and. LocalVar%time >= CntrPar%t_act) then
 
                        ! Detecting the Excessive Thrust Force and Generating Extra Blade Pitch Output
                        !if (LocalVar%We_Vw - LocalVar%Uenv >= 0) then
-                       if (LocalVar%We_Vw - LocalVar%Uenv >= -1.0) then  
+                       if (LocalVar%We_Vw - LocalVar%Uenv >= -CntrPar%Um) then  
                            !LocalVar%Del_Beta = CntrPar%e_dp * (LocalVar%We_Vw - LocalVar%Uenv)                                                                                            
-                           LocalVar%Del_Beta = CntrPar%e_dp * abs(LocalVar%We_Vw - (LocalVar%Uenv-1.0))                                                  
+                           LocalVar%Del_Beta = CntrPar%e_dp * abs(LocalVar%We_Vw - (LocalVar%Uenv-CntrPar%Um))                                                  
 
                        else 
                            LocalVar%Del_Beta = 0
                        end if   
 
                     else if (CntrPar%ASO_Mode == 2) then                                       
 
-                       if (LocalVar%Thrst - CntrPar%PreDf_Thrst >= -0.05*CntrPar%PreDf_Thrst) then
-                            LocalVar%Del_Beta = CntrPar%e_dp * abs(LocalVar%Thrst - (CntrPar%PreDf_Thrst-0.05*CntrPar%PreDf_Thrst))                       
+                       if (LocalVar%Thrst - CntrPar%PreDf_Thrst >= -CntrPar%Tm*CntrPar%PreDf_Thrst) then
+                            LocalVar%Del_Beta = CntrPar%e_dp * abs(LocalVar%Thrst - (CntrPar%PreDf_Thrst-CntrPar%Tm*CntrPar%PreDf_Thrst))                       
                        else 
                            LocalVar%Del_Beta = 0
                        end if

rosco/controller/src/ROSCO_Types.f90

@@ -128,7 +128,10 @@ MODULE ROSCO_Types
     REAL(DbKi)                    :: Kc                          ! AEPS Observer Gain
     REAL(DbKi)                    :: gamma                       ! AEPS Learning Rate
     REAL(DbKi)                    :: ke                          ! AEPS e-modification term gain
-    REAL(DbKi)                    :: e_dp                        ! AEPS design parameter for effective avoidance
+    REAL(DbKi)                    :: Um                          ! Wind margin for AEPS avoidance
+    REAL(DbKi)                    :: Tm                          ! Thrust margin for BEBS avoidance [%]
+    REAL(DbKi)                    :: e_dp                        ! AEPS/BEPS design parameter for effective avoidance
+    REAL(DbKi)                    :: t_act                       ! AEPS activation time for avoidance
     REAL(DbKi)                    :: PreDf_Thrst                 ! AEPS Pre-Defined Thrust Limit [MN]
     CHARACTER(1024)               :: OL_Filename                 ! Input file with open loop timeseries
     INTEGER(IntKi)                :: OL_Mode                     ! Open loop control mode {0 - no open loop control, 1 - open loop control vs. time, 2 - open loop control vs. wind speed}