![]() |
Fusion of Neural Networks, Fuzzy Systems and Genetic Algorithms: Industrial Applications
by Lakhmi C. Jain; N.M. Martin CRC Press, CRC Press LLC ISBN: 0849398045 Pub Date: 11/01/98 |
Previous | Table of Contents | Next |
Figure 7 shows a block diagram of the converters control strategy. The input phase voltages and output line currents are measured and used by a block named Main Controller. This block measures, at every sampling instant, the output and input errors, eo and ei, respectively. If either accepted error is exceeded, then the fuzzy controller determines which port has higher priority. Then, the selected XSVM modulator determines the next converter state. For this, it requires the output and input voltage and current space vectors, obtained by the Main Controller block. Finally, from the selected state, the gating signals to the switches are generated.
Fuzzy logic is a logical system that seeks to emulate human thinking and natural language [15][16]. Fuzzy control, which has emerged as one of the most active branches of fuzzy logic due to its intrinsic characteristics, provides a means of converting a control strategy comprised of a set of linguistic rules based on expert knowledge into an automatic control strategy. This control has proven extremely useful in various industrial applications [17-19], where, usually, control by conventional methods produces inferior results, especially when information being processed is inexact or uncertain.
Figure 7 Control algorithm block diagram for XDFC drive.
Fuzzy logic has a unique and distinct feature of allowing partial membership, that is, a given element can be a partial member of more than one fuzzy set, with various membership degrees. The degree of membership varies from 0 (nonmember) to 1 (full member). In conventional or crisp sets, an element can either be a member or not of a certain set. Figure 8 shows the differences between a fuzzy set and a crisp set of a vehicle speed control system. In Figure 8a) a vehicle doing 73 km/hr is cruising, even though the speed limit for fast is 75 km/hr. In Figure 8b), using fuzzy sets, the same vehicle is a partial member of both cruise and fast, being closer to being a full member of fast than of cruise.
Figure 8 Representation of vehicle speed using a) crisp sets, and b) fuzzy sets theory.
In this particular application, control of the XDFC, fuzzy logic is used to determine which converter port has higher priority and, thus, should be controlled. Once the decision has been made, the fuzzy controller passes the converters command to the XSVM controlling either the output or input terminals while the next converter state is being selected. To accomplish this, the fuzzy controller uses two fuzzy variables. Namely, the output line current error and the input harmonic current error. Whenever these variables trespass the corresponding accepted errors, the fuzzy controller is engaged. The fuzzified variables are then processed using the set of linguistic rules developed based on expert knowledge of the XDFC. From these rules the final decision is taken, specifying which converter port has higher priority and, thus, should be controlled in order to comply with both control objectives.
It seems clear now that the fuzzy controller is critical for the converter operation. It is basically the brains of it. This specific controller was fuzzified due to the intrinsic operation that this logic offers for controlling processes. Specifically, fuzzy control in this particular case realizes a linear interpolation between the two possible control actions it has, controlling either output or input converter ports, so the overall action is a smooth transition between both converter ports. On the contrary, in case this controller was not fuzzified, a threshold decision maker would be required to actually select the converter side with higher priority. This would produce a nonlinear transition between both possible control actions, creating a step transition instead of a linear one. As a result, the converters commutation frequency would double. The global effect produced by the fuzzy controller in the XDFC operation is that the converters commutation frequency is only lightly increased, being able to maintain it beneath 850 Hz controlling both output and input currents. This represents a significant result, as the maximum commutation frequency reaches almost 600 Hz when controlling only the output currents.
The fuzzy controller employs two fuzzy variables and one control variable [15][16]. The fuzzy variables are the fuzzy output or current error eo, and the fuzzy input or harmonic error ei. The output and input errors are defined as
Where Iref is the reference current space vector, Il is the load current space vector, and Ih is the filtered input line current space vector, or the harmonics current space vector.
As shown in Figure 9a), the universe of discourse of the fuzzy variables is divided into three fuzzy sets, namely, null error (N), small error (S), and big error (B). A triangular form membership distribution was chosen for linear interpolation.
The control variable c is the converter port. As the ports are crisp, c does not require a membership distribution.
Figure 9 a) Membership functions for fuzzy variables used; namely, input error ei, and output error eo. b) Set of fuzzy rules for ei and eo, where Out and In refer to the converter port to be controlled, output and input, respectively.
For both fuzzy variables the number of fuzzy segments was chosen to have maximum control with a minimum number of rules (Figure 9b). Each rule can be stated as
Rj: if ei is Aj and eo is Bj then c is Cj,
where Aj and Bj represent fuzzy segments N, S, or B associated to fuzzy variables ei and eo, respectively. As an example, consider the following values for ei and eo:
ei = 1.2
eo = 0.7
Previous | Table of Contents | Next |