FLC System Design NNC System Design

ISSN: 1693-6930 TELKOMNIKA Vol. 10, No. 2, June 2012 : 291 – 302 296

2.8. Proposed Control System

The AI control is utilized to provide the system with the required control action. The control signal is used to control the operation of thermostatic valve to control the mass flow rate of input gases to fuel cell as depicted in Figure 4.

2.8.1. FLC System Design

Fuzzy logic control offers a way of dealing with modeling problems by implementing linguistic. Table 3 shows possible control rule base which are used. The rows represent the rate of the error change ce and the columns represent the error e. Each pair e, ce determines the output level from NL to PL corresponding to output. Here NL is negative large, NM is negative medium, NS is negative small, ZE is zero, PS is positive small, PM is positive medium and PL is positive large. The triangular type membership function is chosen because of its linearity. The collections of the reference fuzzy set for the error, the change of error, and the control input are the same, but their scale factors are different, as shown in Figure 5, seven fuzzy subsets. Figure 4. Block diagram of AI control for PEM fuel cell Table 3. Rule base of fuzzy logic controller CHANGE OF ERROR CE NL NM NS ZE PS PM PL E rr or e NL NL NL NL NL NS NS ZE NM NL NL NL NM NS ZE PS NS NL NL NL NS ZE PS PS ZE NL NM NS ZE PS PM PL PS NS NS ZE PS PL PL PL PM NS ZE PS PM PL PL PL PL ZE PS PS PL PL PL PL Figure 5. Membership functions of error change of error and output signal Figure 6. FLC architecture + - AI Controller unit Thermostatic valve P e P ref Fuel cell Blower -1 -0.8 -0.6 -0.4 -0.2 0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 Output D eg re e o f m em b er sh ip NB NM NS ZE PS PM PB TELKOMNIKA A New Control and The continuity of input membe the continuity of the mapping the center of gravity defuzzifi used in many literatures. Figur The goal of tuning i controller. This nonlinearity c controller parameters. The co in Figure 7 Gain1 = 0.2, Gain plotted against its two inputs. FLC. Figure 7. Control surface Figure 9. Re

2.8.2. NNC System Design

On the other hand the layers which are the input lay neurons, hidden layer consi activation function used in thi The NNC is trained using a b propagation is a form of super back propagated to earlier on most often used as training a the mean square error betw analysis is depicted in Figure which means that the output tr ISSN: 1693-6930 nd Design of PEM Fuel Cell Powered Air Diffused … bership functions, reasoning method, and defuzzifi ng u fuzzy e, ce are necessary. The max-min reaso zification method are used, as those methods are gure 6 shows the MATLAB Simulink of FLC system. is to shape the nonlinearity that is implement called the control surface which is affected by a control surface for the FLC which is suggested in th in2 = 0.2, and Gain3 = 5, where the output of the ts. The surface represents in a compact way all t ce of FLC Figure 8. Mean square Regression between the network output and target the proposed NN control after many trials eventuall ayer, hidden layer, and output layer. The input laye nsists of three neurons, and output layer of on this work is logsig for hidden layer, and purelin a back propagation with Levenberg–Marquardt alg ervised learning for multi-layer nets. Error data at t ones, allowing incoming weights to these layers to algorithm in current neural network applications. F tween the network output and the target. The n re 9. As shown in this figure the regression R is a t tracks the target in a correct way. …. Doaa. M. Atia 297 zification method for soning method and are most frequently m. ented by the fuzzy all the main fuzzy this work is shown e fuzzy controller is ll the information in re error. et ally employed three yer consists of one one neurons. The lin for output layer. algorithm. The back t the output layer is to be updated. It is . Figure 8 presents network response s approximately one ISSN: 1693-6930 TELKOMNIKA Vol. 10, No. 2, June 2012 : 291 – 302 298

3. System Simulation Using MATLAB

In this section the MATLAB SIMULINK model of PEM fuel cell system using for diffused aeration system is introduced. Figure 10, and Figure 11 show the system component using FLC, and NNC respectively. The system consists of PEM fuel cell subsystem, control subsystem to control gases flow rate, and diffuser subsystem. Figure 12 introduces the PEM fuel cell subsystem, the input signals are gases flow rate, and the outputs are current, voltage, and power. Figure 10. Electrical system of diffused aeration system using FLC control Figure 11. Electrical system of diffused aeration system using NNC valve input output fuel cell subsystem Ifc PFC VFC IFC blower motor subsystem tl va w ia To Workspace9 input 1 To Workspace8 t1 To Workspace7 vo1 To Workspace6 ifc To Workspace5 w To Workspace4 ia To Workspace2 t To Workspace12 output To Workspace11 input To Workspace10 va To Workspace1 pfc Scope 9 Scope 4 Scope 3 Scope 2 Scope 11 Scope 10 Scope 1 Input power 2 Input power Gain 1 -K- Fuel cell Control Subsystem e Vc .05u2 Clock1 Control Subsystem e Vc valve input output fuel cell subsystem Ifc PFC VFC IFC blower motor subsystem tl va w ia To Workspace9 input 1 To Workspace8 t1 To Workspace7 vo To Workspace6 ifc To Workspace5 w To Workspace4 ia To Workspace3 output 1 To Workspace2 t To Workspace13 input 2 To Workspace12 output To Workspace11 input To Workspace1 pfc Scope 9 Scope 4 Scope 3 Scope 2 Scope 11 Scope 10 Scope 1 Neural Network x{1} y{1} Mux Input power 2 Input power Gain 1 -K- Fuel cell Control Subsystem e Vc .05u2 Clock1