Property-based biomass feedstock grading using k-Nearest Neighbour technique

Highlights

• We applied an intelligent classification method to biomass resources.

• The approach for the classification is based on k-NN model.

• The best classification result was obtained with Mahalanobis distance function at K = 3.

• This method offers a high prospect in strategic decision-making.

Abstract

Energy generation from biomass requires a nexus of different sources irrespective of origin. A detailed and scientific understanding of the class to which a biomass resource belongs is therefore highly essential for energy generation. An intelligent classification of biomass resources based on properties offers a high prospect in analytical, operational and strategic decision-making. This study proposes the $k$-Nearest Neighbour ($k$-NN) classification model to classify biomass based on their properties. The study scientifically classified 214 biomass dataset obtained from several articles published in reputable journals. Four different values of $k$ ($k=\mathrm{1,2,3,4}$) were experimented for various self normalizing distance functions and their results compared for effectiveness and efficiency in order to determine the optimal model. The $k$–NN model based on Mahalanobis distance function revealed a great accuracy at $k=3$ with Root Mean Squared Error (RMSE), Accuracy, Error, Sensitivity, Specificity, False positive rate, Kappa statistics and Computation time (in seconds) of 1.42, 0.703, 0.297, 0.580, 0.953, 0.047, 0.622, and 4.7 respectively. The authors concluded that $k$–NN based classification model is feasible and reliable for biomass classification. The implementation of this classification models shows that $k$–NN can serve as a handy tool for biomass resources classification irrespective of the sources and origins.

Introduction

Having come to the end of Millennium Development Goals (MDGs), the United Nations (UN) general assembly adopted the 2030 Sustainable Development Goals (SGDs) in September 2015. Focussing on inclusive development, the new agenda emphasize a holistic approach to achieving sustainable development for all [1]. Incidentally, many of the proposed SDGs are dependent on biomass [2]. Goal 7, 9, 12 and 13 of the SDGs substantially speak to the need for alternative energy [1]. By insight into the SDGs, biomass stands as the main avenue through which massive renewable energy can be sustainably achieved in order to ensure access to affordable, reliable, and modern energy for all. This would inevitably culminate in action to combat climate change and its impacts by elevating the sustainable use of biomass resources, since the power generation from fossil fuel continues to raise environmental concerns. Biomass has been identified as the only alternative naturally occurring fossil fuel substitute both regarding carbon content and the available quantity [3]. In 2013, 462 TWh of electricity was produced globally from biomass [4]. Ever since then, the global consumption of biofuel sourced from biomass has been on the increase, and this trend promises to continue as long as major issues surrounding the exploration of bioenergy is frontally addressed. In the future, biomass energy has the potential to provide cost-effective and sustainable energy supply to the population in the developing countries [5,6]. There is an abundant supply of biomass feedstocks which can be converted to biofuels. These feedstocks include; agricultural crops, residues, forest products, energy crops and algae biomass. Municipal Solid Waste (MSW) is also gaining attention as a viable source of biofuel [7,8].

There is a consensus that biomass fuel is a renewable energy resource, but lack of widely accepted terminology, classification metrics, and global standard lead to some serious misconception during the investigations. The main concerns surrounding biomass exploration are related to how to extend and improve the basic understanding of the composition and properties of biomass and also to profitably apply this knowledge in the interest of environmental safety [9]. The knowledge of the proximate properties of biomass and their classification are essential in their selection as an energy feedstock [10]. Among several classification criteria which have been adopted in the classification of biomass feedstock, the most prominent is based on origin and properties [11,12]. There is no known scientific classification method based on artificial intelligence to the best of the authors’ knowledge. Since energy generation from biomass often requires a nexus of different sources, a detailed and scientific understanding of the class to which a biomass resource belongs is highly essential. An intelligent classification of biomass resources based on proximate properties of biomass feedstock offers a high prospect in analytical, operational and strategic decision-making.

Artificial intelligence has opened a new page in the field of data analysis. Machine learning has been applied to several processes, which include data mining and data analysis [[13], [14], [15], [16], [17], [18]] related to supply chain management [19,20], biomass elemental composition [21], municipal solid waste management [[22], [23], [24]], energy consumption prediction [25] and so on. One of the primary objectives of data mining is classification. It is a form of predictive modelling which defines groups within the entire population. The process of classification focusses on finding a model which describes a data class. The aim of classification is to use the derived function to predict the group to which a data point belongs using an unknown class label. By using the classification technique, one can learn the rules that form categories of data [26]. Data classification has been applied in several fields such as; medicine, credit ranking, customer behaviour, and strategic management [27]. Some of the artificial intelligent-based techniques which has been applied in the classification of various data are; Bayesian classifier, Ensemble classifier, Artificial Neural Network (ANN), k-Nearest Neighbour ($k$-NN), Support Vector Machine, SVM [14,28]. The $k$–NN is a common instance-based learning algorithm, which is used to classify an unknown object by ranking the objects neighbour amidst the training data. The result of this then intelligently predicts the class of the new objects. The k in $k$–NN is a pointer to the number of proximate neighbours which are been evaluated in order to determine the class to which a dataset belongs [26,29]. Apart from its simplicity, $k$–NN has proven to be an effective and efficient algorithm used in solving several real-life classification problems with good generalization and accuracy [30]. $k$–NN has been appraised as a transparent machine learning method with high predictive ability, even with little or no prior information about the data distribution [15,31]. $k$–NN stands out as a data mining method of choice for several classification problems due to its notable advantages, which competitively edge-out other methods. Some of its advantages include [31];

• Ability to handle training data that are too large to fit in-memory,

• It can measure the similarities between training tuples and test tuples without prior knowledge about data distribution,

• Reduction of error due to the inaccurate assumption.

• Lower computation time and high prediction accuracy.

The literature review has highlighted the success of $k$–NN in the classification of data for several applications [15,17,31,32].

This study proposes a proof of concept for the classification of biomass based on the traditional classification of properties as proposed by Khan et al. [11] using $k$–NN classification algorithm. The research seeks to innovate a consistent, flexible, direct and easy to implement classification method for biomass properties. Identifying biomass classes is very vital to understanding the mechanism which lead to the varying biomass feedstock behaviour and for an improved prediction of biomass properties [33,34]. This will engender an informed decision making about biomass resources management and feedstock production for industrial utilization especially as related to energy generation. Also, power plant developers will be enabled to ascertain the quality of the feedstock coming from various supply chain. Another objective of this study is to extend the boundary of knowledge in terms of biomass selection decision towards the hybridization of biomass sources for energy generation.