Clustering of the Electricity Consumption Time Series in the Big Data Era

Dear Colleagues,

We are living in the era of emerging technological advancements. The days when almost everything was done manually are gone, and now we live in the time where a lot of activities have been taken over by machines, software, and automatic processes. In this context, and machine learning (ML) has a special place in the advancements being made today. ML applies science to computers and machines to allow them to develop human-like intelligence. Through this technology, machines are able to perform some of the simple to complex tasks that humans need to do on a regular basis.

The concept of Big Data is defined by Gartner as high volume, high velocity, high variety and high veracity data that require new processing paradigms to enable insight discovery, improve decision making, and optimize processes. The potential of Big Data is highlighted in its definition; however, the realization of this potential depends on improving traditional approaches or developing new approaches capable of handling this data. Due to its potential, Big Data has been referred to as a revolution that will transform how we live, work, and think. The main purpose of this revolution is to make use of large amounts of data to enable knowledge discovery and better decision-making.

The ability to extract value from Big Data depends on data analytics, which can be carried out using ML systems. While ML provides significant support in various areas such as time series forecasting, the road to excellence is long. This is because ML has not been able to overcome a number of challenges—especially in the Big Data era—that still stand in the way of progress. Examples of those challenges in terms of Big Data characteristics are:
— for volume: processing performance, feature engineering; for variety: data heterogeneity, dirty and noisy data;

— for velocity: real time processing/streaming, concept drift; and

— for veracity: data provenance.

Data stream clustering has lately become an increasingly popular research topic due to the fact that nowadays more data is produced than there is available storage space. This makes traditional mining approaches that require looking at data more than once unusable. Despite many techniques having been proposed, few have tackled the problem of clustering data streams composed of multiple time series, which we refer to as the problem of clustering streaming time series. Researchers recognize that time series are pervasive in many scientific endeavors, such as electricity load forecasting.

Currently, most techniques for streaming time series clustering rely on a similarity measure based on the Pearson correlation, which is easily computable in an incremental manner, and quite useful in a streaming scenario. However, when relying only on this particular measure, the similarities detected are mainly based on time series shapes and trends, which do not necessarily reflect their true underlying model.

In learning from data streams, which most often occur in non-stationary environments, we can encounter concept drift—a phenomenon by which the data distribution changes and makes the current prediction model inaccurate or obsolete. This can cause the degradation of predictive accuracy if the learning process overlooks the drift and does not adapt to it. Detecting and adapting to concept drift is not a trivial task, and several methods exist that can be roughly split into performance monitoring algorithms and distribution comparing algorithms.

In the light of these challenges, the main scope of this Special Issue is to develop new methods applicable to various machine learning algorithms to improve their quality and robustness in the context of electricity load clustering.

Dr. Krzysztof Gajowniczek
Prof. Arkadiusz Orlowski
Guest Editor

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Energies is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2600 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• Machine learning
• Big Data
• Smart Metering
• Time Series Clustering
• Stream Mining