Assessing the proficiency of adaptive neuro-fuzzy system to estimate wind power density: Case study of Aligoodarz, Iran

https://doi.org/10.1016/j.rser.2015.12.269Get rights and content

Abstract

The prime aim of this study is appraising the suitability of adaptive neuro-fuzzy inference framework (ANFIS) to compute the monthly wind power density. On this account, the extracted wind power from Weibull functions are utilized for training and testing the developed ANFIS model. The proficiency of the ANFIS model is certified by providing thorough statistical comparisons with artificial neural network (ANN) and genetic programming (GP) techniques. The computed wind power by all models are compared with those obtained using measured data. The study results clearly indicate that the proposed ANFIS model enjoys high capability and reliability to estimate wind power density so that it presents high superiority over the developed ANN and GP models. Based upon relative percentage error (RPE) values, all estimated wind power values via ANFIS model are within the acceptable range of −10% to 10%. Additionally, relative root mean square error (RRMSE) analysis shows that ANFIS model has an excellent performance for estimation of wind power density.

Introduction

Nowadays, owing to the detrimental influence of fossil fuels exploitation as well as their increasing prices, it is essential to utilize non-conventional sources for energy generation. As a result of the compulsions imposed by Kyoto protocol to lessen a series of environmental issues such as climate change, air pollution and water contamination, exploitation of the renewable and sustainable energies sources such as wind has come to the fore to conquer the mentioned problems [1], [2]. Since the wind energy is free, environmental friendly and inexhaustible, it is considered to be appropriate source to supply a part of energy demands [3]. In fact, harnessing the wind energy potential is being continuously growing so that it is a major competitor of traditional energy sources, especially as a result of fast installation of wind turbines in many locations across the globe. In this regard, prediction of wind speed and power plays a significant role for the purpose of wind energy development [4]. As a matter of fact, accurate and reliable predictions would be so profitable for operators and investors to provide a secure situation with low level of economic risks. Prior to any attempt towards wind energy harnessing, having information regarding the estimates of wind speed and wind power density is so critical. Nonetheless, as the frequency distribution of wind speed may offer different wind power densities for the similar wind speed, wind power can be considered as a more reliable parameter.

Although several mathematical functions have been suggested to model wind power density, they have some disadvantages such as higher required calculation time. On this account, soft computing methodologies can be utilized alternatively because they offer advantages including no required knowledge of internal system parameters, compact solution for multi-variable problems and fast calculation.

Soft computing methodologies have been successfully employed in the wind energy area in recent years.

Mabel and Fernandez [5] used an artificial neural network (ANN) model to predict the generated wind energy of wind farms in Muppandal, India. They considered average wind speed, relative humidity and generation hours as input elements for the ANN model. They found favorable agreements between the predicted and measured energy output of wind farms. Bilgili et al. [6] employed ANN approach for predicting the monthly mean wind speed at eight locations in the eastern Mediterranean area of Turkey. They utilized the wind speed data of neighboring stations as input elements of the ANN model for predicting the wind speed in the target location. Meharrar et al. [7] determined the optimum speed rotation of a variable wind speed generator using adaptive neuro-fuzzy inferences system (ANFIS) methodology. For this aim, they considered wind speed variation as the input. Their results showed that using the developed system it is possible to achieve the highest power tracking for the wind. Fadare [8] employed ANN technique to model and predict wind speed profile in Nigeria. The monthly wind speed data for 28 stations distributed in different parts of Nigeria were utilized and the elements of latitude, longitude and altitude as well as month of the year were considered as inputs to predict monthly wind speed. Mohandes et al. [9] applied adaptive neuro-fuzzy inferences system (ANFIS) to estimate wind speed profile up to 100 m height. For this aim, they used measured wind speed data of 10, 20, 30 and 40 m heights and demonstrated the proficiency of the developed ANFIS-based model for estimating wind speed profile. Rahmani et al. [10] used particle swarm optimization (PSO) technique to identify the optimal location of wind turbines in a wind farm. The output power and total cost required to develop the wind farm were considered as benchmark. Their results demonstrated that application of PSO is effective for optimizing the placement of wind turbines. Chen et al. [11] proposed a process using ANFIS methodology to diagnose the wind turbines pitch faults. They used the faults data of 6 wind turbines for training the system and the data related to a wind farm including 26 wind turbines. They concluded that the developed approach enjoys high capability to predict the pitch fault of wind turbines. Meharrar et al. [12] used ANFIS technique to design a controller for tracking a maximum power of the wind turbines with a variable-speed wind-generator. The capability of the developed ANFIS controller was checked for fast variation of wind speed. Based on the achieved results, they found that the developed system is robust and has a higher performance than fuzzy logic method. Petković et al. [13] developed an ANFIS-based model to estimate the wind turbine power coefficient in a wind farm. The achieved results clearly indicated the proposed ANFIS model is efficient to provide accurate estimations. Nikolic et al. [14] employed the ANFIS technique to predict the wake power and wind speed deficit in a wind farm. They showed that wind direction is a significant element for wake effect. Their results demonstrated that application of ANFIS is promising and effective for predicting the wake power and wind speed deficit. Petković et al. [15] employed ANFIS to make optimal performances of the wind generator system with continuously variable transmission (CVT). By regulating the CVT ration, it was found that the ANFIS model could determine the highest wind turbine power generation efficiency with extracting maximum wind energy. Salcedo-Sanz et al. [16] proposed a hybrid method based upon Coral Reefs Optimization (CRO) algorithm and the Harmony Search (HS) technique for wind speed prediction. They used wind speed data of two locations from USA and Spain and found that further precision can be attained by the developed method than HS and CRO approaches. Petković and Shamshirband [17] applied ANFIS methodology to determine the influence of different parameters on the energy output of the wind turbines. For this aim, they evaluated the influence of four parameters and found that the blade pitch angle is the most significant element to accurately predict the wind turbine power coefficient. Ramasamy et al. [18] developed an ANN-based model for predicting wind speed in 11 locations of western Himalayan State of India. As the required inputs for the ANN model, air temperature and pressure, solar radiation and altitude were considered to predict daily wind speed. They verified the validity of the model by predicting wind speed for another place with available measured data. The achieved results showed the high precision of the ANN model. Adaptive neuro-fuzzy inferences system (ANFIS), as a soft computing methodology, is a hybrid intelligent system that merges technique of the learning power of the ANNs with the knowledge representation of fuzzy logic. Over the past years, along with wind energy area ANFIS technique has been employed in many other scientific and engineering studies [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29]. ANFIS is an adaptable and efficient method in computational process.

In this study, therefore, an application of adaptive neuro-fuzzy inferences system (ANFIS) is proposed to develop a soft computing model for estimating the wind power density. The prime aim of this research work is to evaluate the sufficiency of ANFIS scheme to estimate the wind power density on a monthly scale. To achieve this, the extracted wind power density values from two Weibull function methods named standard deviation and power density are used and defined as learning and testing data for the developed ANFIS model. The performance evaluation is conducted thoroughly, via several widely-utilized statistical indicators, by making comparisons between the estimated wind power by ANFIS and the wind power computed based upon measured wind data. The validity of the ANFIS model is confirmed by comparing its performance with artificial neural network (ANN) and genetic programming (GP) techniques.

Section snippets

Wind data and wind power density

In this paper, the measured wind speed data at 10 m height for city of Aligoodarz located in the west part of Iran (at 33°37′N and 49°67′E) were used. The utilized wind speed data were measured every three hour recorded for the period of January 2005 to December 2009. For data analysis, firstly, the three hourly wind speed data was averaged to obtain daily data. Afterwards utilizing the daily data of each month, the monthly calculations were performed over each specific month using standard

Performance assessment criteria

The robustness of proposed ANFIS model to estimate the monthly wind power is evaluated via different statistical indicators, which are reviewed briefly in the following.

Results and discussion

In this research work, the ANFIS methodology is employed to develop a model to estimate monthly wind power density. The proficiency of the developed ANFIS model is assessed by providing comparisons with ANN and GP approaches. The descriptions of the ANN and GP approaches can be widely found in the literature [44], [45], [46], [47], [48], [49], [50].

In the first step of analysis, the extracted wind power data from the standard deviation and power density methods were used as input parameters.

Conclusions

The precise and reliable wind resource evaluation plays a substantial role in effective wind energy harnessing. In this regard, the knowledge of wind power density is a vital parameter. In this research work, a soft computing model based upon the ANFIS approach was proposed to estimate the monthly wind power density. The extracted wind power data from the standard deviation and power density methods were used as training and testing data for the developed ANFIS model. In fact, the ANFIS model

Acknowledgements

The authors acknowledge the support provided by University of Malaya Research Grant; RP006 B14HNE, "Quantum Computing for Designing and Validation Procedure".

References (50)

  • D. Petković et al.

    Adaptive neuro-fuzzy approach for wind turbine power coefficient estimation

    Renew Sustain Energy Rev

    (2013)
  • V. Nikolic et al.

    Wind wake influence estimation on energy production of wind farm by adaptive neuro-fuzzy methodology

    Energy

    (2015)
  • D. Petković et al.

    Adaptive neuro-fuzzy maximal power extraction of wind turbine with continuously variable transmission

    Energy

    (2014)
  • S. Salcedo-Sanz et al.

    A Coral Reefs Optimization algorithm with Harmony Search operators for accurate wind speed prediction

    Renew Energy

    (2015)
  • D. Petkovic et al.

    Soft methodology selection of wind turbine parameters to large affect wind energy conversion

    Electr Power Energy Syst

    (2015)
  • P. Ramasamy et al.

    Wind speed prediction in the mountainous region of India using an artificial neural network model

    Renew Energy

    (2015)
  • S. Shamshirband et al.

    Survey of four models of probability density functions of wind speed and directions by adaptive neuro-fuzzy methodology

    Adv Eng Softw

    (2014)
  • S. Shamshirband et al.

    Adaptive neuro-fuzzy generalization of wind turbine wake added turbulence models

    Renew Sustain Energy Rev

    (2014)
  • D. Petković et al.

    Adaptive neuro-fuzzy prediction of modulation transfer function of optical lens system

    Infrared Phys Technol

    (2014)
  • D. Petković et al.

    Determining the joints most strained in an underactuated robotic finger by adaptive neuro-fuzzy methodology

    Adv Eng Softw

    (2014)
  • B. Khoshnevisan et al.

    Prediction of potato yield based on energy inputs using multi-layer adaptive neuro-fuzzy inference system

    Measurement

    (2014)
  • D. Petković et al.

    Adaptive neuro-fuzzy prediction of grasping object weight for passively compliant gripper

    Appl Soft Comput

    (2014)
  • J. Piri et al.

    Modelling solar radiation reached to the Earth using ANFIS, NNARX, and empirical models (Case studies: Zahedan and Bojnurd Stations)

    J Atmos Sol-Terr Phys

    (2015)
  • A. Moghaddamnia et al.

    Comparison of LLR, MLP, Elman, NNARX and ANFIS models—with a case study in solar radiation estimation

    J Atmos Sol-Terr Phys

    (2009)
  • M.S. Pourtahmasb et al.

    Resilient modulus prediction of asphalt mixtures containing Recycled Concrete Aggregate using an adaptive neuro-fuzzy methodology

    Constr Build Mater

    (2015)
  • Cited by (15)

    • Determination of the most effective control methods of SO<inf>2</inf> Pollution in Tehran based on adaptive neuro-fuzzy inference system

      2021, Chemosphere
      Citation Excerpt :

      In previous researches, different percentages of total data has been used as test data. For examples; Shahaboddin Shamshirband (Shamshirband et al., 2016), Ling Zou (Zou et al., 2017), Shadi Ausati (Ausati and Amanollahi, 2016), Bijan Yeganeh (Yeganeh et al., 2018), Qiuju Xie (Xie et al., 2017), Kanchan Prasad (Prasad et al., 2016) and Osman Taylan (2017) used 30, 30, 8.2, 15, 6.7, 20 and 20% of the data as test data, respectively. Therefore, in this study, in order to increase the accuracy of the model, the highest value was selected for test data (30%).

    • Evaluation of neural network-based methodologies for wind speed forecasting

      2019, Computers and Electrical Engineering
      Citation Excerpt :

      During the past decades, fuzzy logic had pervasive applications in sustainable energy systems. The efficiency of fuzzy logic utilization for wind speed prediction is scrutinized in [17]. The fifth class is Wavelet transformation methodologies.

    • A numerical model based on prior distribution fuzzy inference and neural networks

      2017, Renewable Energy
      Citation Excerpt :

      To decide which parameters to use one must take into account the alignment of the particularities of the problem at hand with the characteristics of the techniques [84]. For instance, when selecting parameters of a routine item with many potential values [86–88], techniques that do not limit analysis to only a few alternatives are more adequate than others. Table 5 shows the Parameters of ANFIS set in this paper, which are selected followed reference [85].

    View all citing articles on Scopus
    View full text