CHAPTER 6 ANFIS BASED NEURO-FUZZY CONTROLLER

Transcription

1 143 CHAPTER 6 ANFIS BASED NEURO-FUZZY CONTROLLER 6.1 INTRODUCTION The quality of generated electricity in power system is dependent on the system output, which has to be of constant frequency and must maintain the scheduled power and voltage. The dynamic behaviour of the system depends on disturbances and on changes in the operating point. The unsteady nature of wind and frequent change in load demands may cause large and severe oscillation of power for the considered wind-micro hydrodiesel hybrid power system. In this Thesis, the supplementary controller for LFC of the diesel generating unit takes care of sudden load changes and maintains the system frequency, and the supplementary controller for BPC takes care of the wind input variation and maintains the wind power generation. In the proposed work, Adaptive Neuro-Fuzzy Inference system based Neuro-Fuzzy Controller (NFC) is designed individually for both governor in diesel side and blade pitch control in wind side for performance improvement of the wind-micro hydro-diesel hybrid system. This newly developed control strategy combines the advantage of Neural Network and Fuzzy Inference System and has simple structure that is easy to implement. In order to keep system performance near its optimum, it is desirable to track the operating conditions and use the updated parameters to control the system. This work investigates the performance of wind-micro hydro-diesel hybrid power system using ANFIS based NFC for LFC and BPC by simulation.

2 144 This chapter is organized as follows. Section 2 describes the introduction of Neuro-Fuzzy system. Section 3 demonstrates the Adaptive Neuro-Fuzzy Inference System architecture. Section 4 describes the design of ANFIS based Neuro-Fuzzy Controller for LFC and BPC of wind-micro hydro-diesel hybrid power system. Section 5 demonstrates the simulation results of the hybrid system with ANFIS based NFC. Section 6 shows the analysis and performance comparison of the proposed controller. Summary of the chapter is given in section NEURO-FUZZY SYSTEM The techniques of fuzzy logic and neural networks suggest the novel idea of transforming the burden of designing fuzzy logic systems to the training and learning of connectionist neural networks and vice-versa. That is, the neural networks provides connectionist structure and learning to the fuzzy logic systems and the fuzzy logic systems provide the neural networks with a structural framework with high level fuzzy IF-THEN rule thinking and reasoning. These benefits can be witnessed by the success in applying neurofuzzy systems in areas like control of power system. Although the benefits of combining fuzzy logic and neural networks are well known and have been widely demonstrated, this work investigates its application in LFC and BPC of an isolated wind-micro hydro-diesel hybrid power system, to improve the system performance. ANN have a massive parallel structure in the form of a directed graph, composed of processing units( neurons) that are linked through connections which may or may not have adjustable weights. Figure 6.1 shows a very simple structure of a Neural Network, composed of two layers (input and output), each of them composed of three and two processing units respectively.

3 145 Figure 6.1 The general structure of a Neural Network The advantages of neural networks over conventional systems are their ability to perform non-linear input-output mapping, generalisation, adaptability and fault tolerance (Lin and Lee 1996). On the other hand, the main disadvantage of neural network is the broad lack of understanding of how they actually solve a given problem. The main reason for this is that neural networks do not break a problem down into its logical elements, but rather solve it by a holistic approach, which can be hard to understand logically. The main result of neural network learning process is reflected only in a set of weights in which a full understanding of the functioning of the neural network is an almost impossible task. To overcome this, hybrid Neuro- Fuzzy system is proposed. Fuzzy logic which gives the benefit of enabling systems more easily to make human-like decisions (Zadeh 1965), was discussed in the previous chapters in detail. The advantage gained from fuzzy logic approach is the ability to express the amount of ambiguity in human thinking and subjectivity (including natural language) in a comparatively undistorted manner. So, fuzzy logic technique finds their applications in areas such as control (the most widely applied area), pattern recognition, quantitative analysis etc.

4 146 The main disadvantage of fuzzy systems, however, is that they do not have much learning capability to tune their fuzzy rules and membership functions. Normally, fuzzy rules are decided by experts or operators according to their knowledge or experience. However, when the fuzzy system model is designed, it is often too difficult (sometimes impossible) for human beings to define all the fuzzy rules or membership functions. Fuzzy Logic (Zadeh 1965) and Artificial Neural Networks (Haykin 1998) are complementary technologies in the design of intelligent systems. The combination of these two technologies into an integrated system appears to be a promising path toward the development of the intelligent systems capable of capturing qualities characterising the human brain. The neural network can improve the transparency, making them closer to fuzzy systems, while fuzzy systems can self adapt, making them closer to neural networks (Lin and Lee 1996). Neural fuzzy systems (Jang et al and Lin and Lee 1996) have attracted the growing interest of researchers in various scientific and engineering areas, especially in the area of control; hybrid Neuro-Fuzzy systems seem to be attracting increasing interest. 6.3 ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM (ANFIS) Adaptive Neuro-Fuzzy Inference System (ANFIS) is an Artificial Intelligence technique which creates a fuzzy inference system based on the input-output model data pairs of the system. ANFIS combines neural network and fuzzy system together. ANFIS can be employed in a wide variety of applications of modelling, decision making, signal processing and control. ANFIS is a class of adaptive network that is functionally equivalent to Fuzzy Inference System. Since ANFIS design starts with the pre-structured system, the membership function of input and output variables contain more

5 147 information that Neural Network has to drive from sampled data sets. Knowledge regarding the systems under design can be used right from the start. Hence, the proposed ANFIS controller is more efficient. The rules are in the linguistic forms and so intermediate results can be analyzed and interpreted easily (Ashok Kusagur et al. 2010). ANFIS is a multi layer adaptive neural network based Fuzzy Inference System. ANFIS algorithm is composed of fuzzy logic and neural networks with 5 layers to implement different node functions to learn and tune parameters in a Fuzzy Inference System (FIS) structure using a hybrid learning mode. In the forward pass of learning, with fixed premise parameters, the least squared error estimate approach is employed to update the consequent parameters and to pass the errors to the backward pass. In the backward pass of learning, the consequent parameters are fixed and the gradient descent method is applied to update the premise parameters. Premise and consequent parameters will be identified for membership function (MF) and FIS by repeating the forward and backward passes. ANFIS is fuzzy Sugeno model put in the framework of adaptive systems to facilitate learning and adaption (Jang 1993). Such framework makes Fuzzy Logic Controller more systematic and less relying on expert knowledge. To present the ANFIS architecture, let us consider two fuzzy rules based on a first order sugeno model: Rule 1 : IF (x is A1) and (y is B1) THEN (f 1 =p 1 x+ q 1 y +r 1 ) Rule 2 : IF (x is A2) and (y is B2) THEN (f 2 =p 2 x + q 2 y +r 2 ) where x and y are the inputs, Ai and Bi are the fuzzy sets, f i are the outputs within the fuzzy region specified by the fuzzy rule, p i, q i and r i are the design parameters that are determined during the training process.

6 148 Figure 6.2 illustrates the equivalent ANFIS architecture for this sugeno model (Jang et al. 2005), where nodes of the same layer have similar functions. Figure 6.2 ANFIS architecture Out of the five layers, the first and fourth layers consist of adaptive nodes while the second, third and fifth layers consist of fixed nodes. The adaptive nodes are associated with their respective parameters, get duly updated with each subsequent iteration while the fixed nodes are devoid of any parameters. Layer 1: Fuzzification layer: Every node i in this layer 1 is an adaptive node. The outputs of layer 1 are the fuzzy membership grade of the inputs, which are given by: O i 1 = Ai (x), For i= 1,2 (6.1) O i 1 = Bi-2 (y), For i= 3,4 (6.2)

7 149 where x and y is the inputs to node i, where A is a linguistic label (small, large) associated with this node. Where Ai(x), Bi-2(y) can adapt any fuzzy membership function. The membership function for A can be any appropriate parameterized membership function such as the generalised bell function. A(x) 1 1 x c a i i 2b (6.3) where { a i, b i., c i } is the parameter set. In fact, any continuous and piecewise differentiable functions, such as commonly used trapezoidal (or) triangular-shaped membership functions, are also qualified candidates for node functions in this layer. Parameters in this layer are referred to as premise parameters. Layer 2: Rule layer: Every node in this layer is a fixed node labelled M, whose output is the product of all the incoming signals. The outputs of this layer can be represented as: O i 2 = wi = Ai (x). Bi (y), i=1,2 (6.4) Each node output represents the firing strength of a rule. Layer 3: Normalization layer: Every node in this layer is a fixed node labeled N. The i th node calculates the ratio of the i th rule s firing strength to the sum of all rule s firing strength. w Oi w i 1, 2 (w w ) 3 i i 1 2 (6.5)

8 150 For convenience, outputs of this layer are called normalized firing strengths. Layer 4: De-fuzzification layer: Every node i in this layer is an adaptive node. The output of each node in this layer is simply the product of the normalized firing strength and a first order polynomial. O i 4 = w i f i = w i (p i x+ q i y +r i ) i=1, 2 (6.6) where w i is a normalized firing strength from layer 3 and { p i, q i, r i } is the parameter set of this node. Parameters in this layer are referred to as consequent parameters. Layer5: Summation neuron: This single node in this layer is a fixed node labelled computes the overall output as the summation of all incoming signals., which O w f i i 5 i i wi fi i 1 2 (w w ) (6.7) Thus we have constructed an adaptive network that is functionally equivalent to a Sugeno fuzzy model. From the ANFIS architecture shown in Figure 6.2, it is observed that when the values of the premise parameters are fixed, the overall output can be expressed as a linear combination of the consequent parameters. More precisely, the output f in Figure 6.2 can be rewritten as w w f. f.f w.f w.f w w w w 1 2 i

9 151 (w 1.x)p 1 (w 1.y)q 1 (w 1)r 1 (w 2.x)p 2 (w 2.y)q 2 (w 2)r 2 (6.8) which is linear in the consequent parameters p 1, q 1, r 1, p 2, q 2 and r 2. The ANFIS structure is tuned automatically by least square estimation and the back propagation algorithm (hybrid learning). 6.4 DESIGN OF ANFIS BAESD NEURO-FUZZY CONTROLLER The development of the control strategy to control the frequency deviation of the wind-micro hydro-diesel hybrid power system using the concept of ANFIS control scheme is presented here. The proposed neuro-fuzzy method combines the advantages of neural networks and fuzzy system to design a model that uses a fuzzy theory to represent knowledge in an interpretable manner and the learning ability of a neural network to optimize its parameters. ANFIS is a specific approach in Neuro-fuzzy development which was first introduced by Jang (1993). To start with, we design the controller using ANFIS scheme. The model considered here is based on Takagi- Sugeno fuzzy inference model. The block diagram of the proposed ANFIS based NFC for LFC and BPC of wind-micro hydro- diesel hybrid power system is shown in Figure 6.3. Figure 6.3 Block diagram of ANFIS based Neuro-Fuzzy Controller

10 152 The inputs to the ANFIS based Neuro-Fuzzy controller for LFC of the hybrid system on diesel side are error E( Fs) and change in error E Fs). The fuzzification unit converts the crisp data into linguistic variables, which is given as input to the rule based block. The set of 49 rules are written on the basis of previous knowledge/experiences in the rule based block. Hybrid learning algorithm is used to train the neural network to select the proper set of rule base. For developing the control signal, the training is very important step in the selection of the proper rule base. Once the proper rules are selected and fired, the control signal required to obtain the optimal output is generated. The output of NN unit is given as input to the de-fuzzification unit and the linguistic variables are converted back into the crisp form. This chapter proposes a systematic approach for establishing a concise ANFIS that is capable of online self-organizing and self-adapting its internal structure for learning the required control knowledge that satisfies the desired system performance. The proposed ANFIS based NFC uses a hybrid learning algorithm to identify consequent parameters of Sugeno type fuzzy inference system. This algorithm applies a combination of least square method and back propagation gradient descent method for training fuzzy inference system membership function parameters to emulate a given training data set. Steps to design the ANFIS based NFC are given below. 1. Draw the simulink model for LFC and BPC of an isolated wind-micro hydro-diesel hybrid power system with Takagi- Sugeno inference model Fuzzy Logic Controller.

11 Simulate it with seven membership functions for the two inputs, error ( Fs) and change in error ( Fs) and with rule base shown in Figure 6.5. Simulation steps are same as explained in previous chapters for. 3. Collect the training data while simulating the system with Fuzzy Logic Controller to design the ANFIS based NFC. 4. The two inputs, i.e. error ( Fs on diesel side and P GW on wind side) and change in error ( Fs, P GW ) and the output signal gives the training data. 5. Use ANFIS edit to create the FIS file (lfcanfi23.fis). 6. Load the training data collected from step2 and load the lfcanfi23.fis file. 7. Choose the hybrid learning algorithm. 8. Train the collected data with generated FIS up to particular number of Epochs. Figure 6.4 shows the FIS editor of Sugeno type Fuzzy Inference System with two inputs (error and change in error) and one output. Each input is having seven linguistic variables with triangular membership functions. 49 rules are framed and it is shown in Figure 6.5. The rules are viewed by rule viewer as shown in Figure 6.6.

12 154 Figure 6.4 FIS editor (Sugeno model) with two inputs and one output Figure 6.5 Rule editor of Fuzzy Sugeno model

13 155 Figure 6.6 Rule viewer of Fuzzy Sugeno model After running this FIS file with simulink model, the training data are collected and loaded by the ANFIS editor. The ANFIS editor window shown in Figure 6.7 has all the provisions for loading the data and FIS file from the workspace and also for training and testing the data for better control performance. Figure 6.7 ANFIS editor window

14 156 Figure 6.8 shows the loading and training of data to ANFIS structure. The ANFIS structure is trained with hybrid learning up to 50 epochs, with error tolerance of zero. Figure 6.8 Training of data with hybrid learning method The ANFIS information obtained after simulation is as follows. ANFIS info: Number of nodes: 131 Number of linear parameters: 49 Number of nonlinear parameters: 42 Total number of parameters: 91 Number of training data pairs: 101 Number of checking data pairs: 0 Number of fuzzy rules: 49

15 157 Start training ANFIS e Figure 6.9 shows the ANFIS structure for the designed NFC for LFC and BPC of wind-micro hydro-diesel hybrid power system. Figure 6.9 ANFIS architecture of Neuro-Fuzzy Controller for LFC and BPC of wind-micro hydro-diesel hybrid power system 6.5 SIMULATION RESULTS Simulations are performed with the proposed ANFIS based NFC for LFC and BPC of the hybrid power system with system parameters given in Appendix 1. The ANFIS based NFC, improves the performance of the hybrid system by the learning and training approach. For the same system parameters, simulink model of the hybrid power system is simulated with (Mamdani model) and conventional for performance comparison. All the responses such as change in frequency, change in wind power, change in diesel power and change in hydro power during various load disturbances

16 158 are observed and investigated in terms of settling time, overshoot and steady state error value to get the optimum performance of the wind-micro hydrodiesel hybrid power system. Simulation is carried out for 1%, 2%, 3%, 4% and 5% step load change ( PL=0.01 p.u., 0.02 p.u., 0.03 p.u., 0.04 p.u. and 0.05 p.u.) at t = 0 sec. The change in frequency of the system, change in wind power generation, change in diesel power generation and change in hydro power generation for 4% (0.04 p.u.) step load change is shown in Figures 6.10, 6.11, 6.12 and 6.13 respectively ANFIS Time in secs.s) Figure 6.10 Change in frequency of the hybrid system with ANFIS based NFC, and for a step load change of 4%

17 ANFIS Time in secs. Figure 6.11 Change in wind power generation of the hybrid system with ANFIS based NFC, and for a step load change of 4% ANFIS Time in secs. Figure 6.12 Change in diesel power generation of the hybrid system with ANFIS based NFC, and for a step load change of 4%

18 160 x ANFIS Time in secs. Figure 6.13 Change in hydro power generation of the hybrid system with ANFIS based NFC, and for a step load change of 4% From the simulation results, settling time for change in frequency, change in wind, diesel and hydro power generation of the hybrid power system for the proposed ANFIS based NFC, conventional and for a step load change of 1%, 2%, 3%, 4% and 5% are observed and tabulated in Table 6.1.

19 161

20 162 On analysing the performance from the Table 6.1, it is observed that the proposed ANFIS based NFC damps out the deviations in frequency, wind, diesel and hydro power with less settling time for various load disturbances (from 0.01 p.u. to 0.05 p.u.). The proposed ANFIS based NFC maintains steady response and is more reliable than the fixed parameter and, regardless of changes in load power variations. The proposed controller generates a control signal to the governor, which in turn controls the diesel power generation to maintain the system frequency and power generation of the renewable hybrid system. The amplitude of the second oscillation for dynamic responses (deviations in frequency, wind, diesel and hydro power) of the hybrid system for various load disturbances are observed and tabulated in Tables 6.2 and 6.3 for ANFIS based NFC, and. Table 6.2 Amplitude of oscillations for deviations in frequency and wind power for NFC, and against various load disturbances Load change (p.u.) Change in frequency(hz) ANFIS based NFC Change in wind power(p.u. KW) ANFIS based NFC

21 163 Table 6.3 Amplitude of oscillations for deviations in diesel and hydro power for NFC, and against various load disturbances Load change (p.u.) Change in diesel power(p.u. KW) ANFIS based NFC ANFIS based NFC Change in hydro power(p.u.kw) On analysing the performance from the Tables 6.2 and 6.3, the amplitude of oscillations compared to and. of the proposed ANFIS based NFC is less Simulation of the hybrid power system with the proposed ANFIS based NFC shows improved system performance when compared to conventional and. 6.6 ANALYSIS AND PERFORMANCE COMPARISON An investigation on the dynamic responses of frequency change and change in wind, diesel and hydro power generation have been carried out for LFC and BPC of the hybrid power system using the proposed ANFIS based NFC for different load changes and compared with and conventional PI Controller. Figure 6.14 shows the comparison of responses (frequency response, change in wind, diesel and hydro power) in terms of settling time for ANFIS based NFC, and against a step load change of 4%(0.04 p.u.).

22 Settling time in seconds ANFIS based NFC Change in frequency (Hz) Change in wind power (p.u.kw) Change in diesel power (p.u.kw) Change in hydro power (p.u.kw) Figure 6.14 Comparison of dynamic responses of the hybrid system for NFC, and in terms of settling time (seconds) against a step load change of 4% The bar chart in Figures 6.15, 6.16, 6.17 and 6.18 illustrates the performance comparison of the hybrid system for change in frequency, change in wind power, change in diesel power and change in hydro power respectively for three controllers (ANFIS based NFC, and ) in terms of settling time against various load disturbances. 4 Settling time in seconds ANFIS based NFC 1% 2% 3% 4% 5% Load disturbance in percentage Figure 6.15 Comparison of frequency response for NFC, and in terms of settling time (seconds) against various load disturbances

23 165 Settling time in seconds ANFIS based NFC 0 1% 2% 3% 4% 5% Load disturbance in percentage Figure 6.16 Comparison of change in wind power response for NFC, and in terms of settling time (seconds) against various load disturbances 4 Settling time in seconds % 2% 3% 4% 5% Load disturbance in percentage ANFIS based NFC Figure 6.17 Comparison of change in diesel power response for NFC, and in terms of settling time (seconds) against various load disturbances

24 Settling time in seconds ANFIS based NFC 0 1% 2% 3% 4% 5% Load disturbance in percentage Figure 6.18 Comparison of change in hydro power response for NFC, and in terms of settling time (seconds) against various load disturbances The bar chart in Figures 6.19, 6.20, 6.21 and 6.22 illustrates the performance comparison of three controllers (ANFIS based NFC, and ) in terms of amplitude of oscillations against various load disturbances. Amplitude of oscillation in Hz % 2% 3% 4% 5% ANFIS based NFC Load disturbance in percentage Figure 6.19 Comparison of change in frequency response of the hybrid system for NFC, and in terms of oscillations (Hz) against various load disturbances

25 167 Amplitude of oscillation in p.u.kw % 2% 3% 4% 5% ANFIS based NFC Load disturbance in percentage Figure 6.20 Comparison of change in wind power response of the hybrid system for NFC, and in terms of oscillations (p.u. KW) against various load disturbances Amplitude of oscillation in p.u.kw % 2% 3% 4% 5% ANFIS based NFC Load disturbance in percentage Figure 6.21 Comparison of change in diesel power response of the hybrid system for NFC, and in terms of oscillations (p.u. KW) against various load disturbances

26 168 Amplitude of oscillation in p.u.kw % 2% 3% 4% 5% ANFIS based NFC Load disturbance in percentage Figure 6.22 Comparison of change in hydro power response of the hybrid system for NFC, and in terms of oscillations (p.u. KW) against various load disturbances From the performance comparison, it is observed that the overshoot and settling time of the ANFIS based NFC is lower than those of and conventional. The results obtained using ANFIS based NFC proposed in this work outperform both conventional and by its hybrid learning algorithm. 6.7 SUMMARY An attempt is made in this work to develop a control strategy that combines the advantages of neural networks and fuzzy inference system for LFC and BPC of an isolated wind-micro hydro-diesel hybrid power system. The hybrid learning algorithm applied by the proposed ANFIS structure trains the Fuzzy Inference System membership function parameters for better dynamic performance of the hybrid system. The designed ANFIS based NFC for LFC and BPC of the hybrid system is investigated for various load disturbances by simulation. It is observed from the simulation results that the proposed controller is effective and provides significant improvement in

27 169 system performance by combining the benefits of fuzzy logic and neural networks. The proposed ANFIS based Neuro-Fuzzy Controller maintains the system reliable for sudden load changes and proves its superiority.