Comparing Mamdani Sugeno fuzzy logic and RBF ANN network for PV fault detection

TL;DR: A new fault detection algorithm for photovoltaic (PV) systems based on artificial neural networks (ANN) and fuzzy logic system interface and both Mamdani, Sugeno fuzzy logic systems interface is proposed.

About: This article is published in Renewable Energy.The article was published on 2018-03-01 and is currently open access. It has received 167 citations till now. The article focuses on the topics: Neuro-fuzzy & Fuzzy logic.

Summary

INTRODUCTION

• The monitoring and regular performance supervision on the functioning of grid-connected photovoltaic (GCPV) systems is necessary to ensure an optimal energy harvesting and reliable power production.
• There are existing techniques which were developed for possible fault detection in grid-connected PV systems.
• Finally, B. Amrouche & X. Pivert [30] offered an ANN network based daily local forecasting for global solar radiation (GHI).

M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT

• The proposed algorithm is comparing between two different approaches for detecting failure conditions which can be described as the following: 1. Artificial Neural Network (ANN) Approach: Four different ANN networks have been compared using a logged data of several faulty conditions affecting the examined PV plant.
• The implementation of the ANN network has been developed using MATLAB/Simulink software.
• Moreover, the minimum Mean Square Errors (MSE) achieved during the training and test processes are 0.005 and 0.007 respectively.
• Ten fuzzy logic rules were decided according to a sensitivity analysis made by varying the number and type of the rule.
• A satisfactory level of performance was obtained after a tuning process, i.e. starting from faulty PV module only and progressively modifying the fuzzy system to detect all possible faults the may occur in the PV plant according to the faults types listed in Table 1 .

3.2. Photovoltaic Theoretical Modelling

• The Current-Voltage (I-V) and Power-Voltage (P-V) curves of the examined PV module is shown in Fig. 3 (A) and Fig. 3 (B) respectively.
• The simulation temperature remains at STC (25 °C).

3.3 Overall PV Fault Detection Algorithm

• The fuzzy logic systems are explained in section 3.5.
• Moreover, the type of the fault which can be detected using the machine learning techniques are shown in Table 1 .

3.4 ANN Model Implementation

• 10% of samples are used to validate the ANN network.
• This test is not used in the training process.
• 20% of samples are used to test the actual ANN network detection accuracy.

Sugeno-type: Mamdani-type:

• As can be noticed, ten different regions have been selected, where region 1 is the low partial shading (PS) condition.
• The minimum and maximum limits for each region of the VR and PR is also shown in Table 3 , the defuzzification process for the input rules is the centroid type.
• After identifying the input variables VR and PR regions, it is required to set the rulers for the fuzzy logic system interface.

RESULTS AND DISCUSSION

• This section reports the results of the developed fault detection algorithm.
• Furthermore, a comparison between the developed machine learning techniques with some ANN and fuzzy logic systems obtained by various researchers is briefly explained in section 4.4 (discussion section).

4.1 Experimental Data

• As can be noticed, the data samples for both sleep and normal operation modes are not included in the evaluation process of the machine learning techniques, since both scenarios can be detecte3d using the mathematical regions explained in Fig. 5 .
• Furthermore, scenarios 3~5 and 7~11 are evaluated by the ANN network and the fuzzy logic system, were the total number of sample for the faulty conditions is equal to four hundred and eighty.
• Moreover, a comparison between the theoretical output power vs. the real time 351 long term measured data of the PV system during the tested faulty conditions are is shown in Fig. 10 .

4.2 Performance Evaluation of the proposed ANN Networks

• As can be noticed that all examined faulty conditions are accurately detected by Mamdani fuzzy logic system.
• This situation is occurring in the fuzzy system due to the high number of faulty regions identified by the fuzzy system, additionally, the VR and PR ratios are strongly depends on the performance of the voltage and current sensors used to detect the change in the PV parameters (voltage, current and power).
• Therefore, the fuzzy logic system might need some extra few seconds to start detecting the exact faulty occurring in the PV installation.

4.4 Discussion

• PV fault detection based on multi-resolution signal decomposition [36 & 37].
• The overall detection accuracy for both machine learning techniques are high if they have been built accurately.
• Finally, Table 6 shows some of the recent applications for ANN networks and the fuzzy logic systems developed nowadays in PV plants.

5. CONCLUSION

• Additionally, two different fuzzy logic systems have been examined.
• Mamdani fuzzy logic system interface and Sugeno type fuzzy system.
• Both examined fuzzy logic systems show approximately the same output during the experiments.
• There are slightly difference in developing each type of the fuzzy systems such as the output membership functions and the rules applied for detecting the type of the fault occurring in the PV plant.

Figures

Fig. 4. DC side Numerical Calculations at Maximum and Minimum Operating Points

Fig. 13(B). Similarly, case 10 and 11 presents a three faulty PV modules with 20% and 0% PS 436 respectively. Both cases are within the same membership function region due to the low PS condition 437 affecting the PV modules, this situation is labeled as case 10 and case 11 on both Figs. 13(A) and 13(B). 438

Fig. 6. The Adopted ANN Network. (A) 2 Inputs, 5 Outputs using 1 Hidden Layer, (B) 2 Inputs, 5 Outputs using 2 Hidden Layers, (C) 2 Inputs, 9 Outputs using 1 Hidden Layer, (D) 2 Inputs, 9 Outputs using 2 Hidden Layers

TABLE 2 ELECTRICAL CHARACTERISTICS OF SMT6 (60) P PV MODULE

TABLE 6 COMPARISON BETWEEN ANN AND FUZZY LOGIC SYSTEMS

TABLE 4 MULTIPLE FAULTS OCCURRING IN THE EXAMINED PV SYSTEM

Fig. 10. Theoretical Output Power vs. Measured Output Power for All Tested Scenarios Applied on the Examined PV system, Each Case is Perturbed as Shown in Appendix B

Fig. 1. Overall System Architecture Design for theExamined PV Plant

Frequently asked questions

Q1. What contributions have the authors mentioned in the paper "Comparing mamdani sugeno fuzzy logic and rbf ann network for pv fault detection" ?

8 This work proposes a new fault detection algorithm for photovoltaic ( PV ) systems based on artificial 9 neural networks ( ANN ) and fuzzy logic system interface. There are few instances of machine learning 10 techniques deployed in fault detection algorithms in PV systems, therefore, the main focus of this paper is 11 to create a system capable to detect possible faults in PV systems using radial basis function ( RBF ) ANN 12 network and both Mamdani, Sugeno fuzzy logic systems interface. 1 %. Furthermore, both examined fuzzy logic 17 systems show approximately the same output during the experiments. 

Q2. What is the effect of a high number of fuzzy rules on the PV system?

a high number of fuzzy rules may lead to an over 326 parameterized system, thus reducing generalization capability and accuracy of detection the type of the 327 fault accruing in the examined PV system. 

Q3. What is the importance of fault detection in PV systems?

The 25 development of diagnostic methods for fault detection in the PV systems behaviour is particularly 26 important due to the expansion degree of GCPV systems nowadays and the need to optimize their 27 reliability and performance. 

Q4. What is the simplest way to detect a fault in a PV system?

A computer 98 interface has two options, a PV fault detection algorithms which use MATLAB/Simulink software which 99 contains the ANN and the fuzzy logic interface system. 

Q5. What is the way to improve the reliability of PV systems?

53PV systems reliability improvement by real-time field programmable gate array (FPGA) based on switch 54 failures diagnosis and fault tolerant DC-DC converters is presented by [16]. 

Q6. what is the simplest way to improve the reliability of a photovoltaic system?

Photovoltaic systems reliability improvement by real-554 time FPGA-based switch failure diagnosis and fault-tolerant DC–DC converter. 

Q7. What is the second machine learning technique used to detect faults in the PV system?

In this study, the second machine learning technique used to detect faults in the PV system is the fuzzy 282 logic system interface. 

Q8. what is the type of fault detected in the ANUS CRIP TAC?

further investigation of the proposed fault 503 detection algorithm is intended to be used with field programmable gate array (FPGA) platforms which 504 accelerate the speed of detecting possible faults occurring in PV systems. 

Q9. What is the proposed approach for a system with a two-loop control scheme?

The developed 56 approach is based on a novel model-based, two-loop control scheme for a particular MIPC system, where 57 bidirectional Cuk DC-DC converters are used as the bypass converters and a terminal Cuk boost 58 functioning as a while system power conditioner. 

Q10. How many samples were used as inputs for each ANN network?

In order to verify the performance of the proposed ANN networks, the VR and PR ratios of 480 samples 354 illustrated in Table 4 have been used as an input for each ANN network shown previously in Fig. 

Q11. What is the way to detect faults in a PV system?

Whereas [25] proposed 72 an ANN network that detects faults in the DC side of PV systems which includes faulty bypass diodes 73 and faulty PV modules in a PV string.