Forecasting short-term electricity consumption using a semantics-based genetic programming framework: The South Italy case
Introduction
Load forecasting is the task of predicting electricity demand on different time scales, in minutes (very short-term), hours/days (short-term), and months and years (long-term). This information can be used to plan and schedule operations on power systems (dispatch, unit commitment, network analysis) in a way to control the flow of electricity in an optimal way, with respect to various aspects (like for instance quality of service, reliability, costs). An accurate load forecasting has great benefits for electric utilities and both negative and positive errors lead to increased operating costs. Overestimation of demand leads to an unnecessary energy production or purchase and, on the contrary, underestimation causes unmet demand with a higher probability of failures and costly operations. Several factors influence electricity demand: day of the week and holidays (the so-called “calendar effects”), special or unusual events, economic situation and weather conditions. In warm countries, the last factor is particularly critical during summer, when the use of refrigeration, irrigation and air conditioning is more common than in the rest of the year.
With the recent trend of deregulation of electricity markets, energy demand forecasting has gained even more importance. In the market environment, precise forecasting is the basis of electrical energy trade and spot price establishment for the system to gain the minimum electricity purchasing cost.
All these facts show the importance of having reliable predictive models that can be used for an accurate energy demand forecasting. In this paper, the goal is to propose a new and sophisticated computational method that can be used to automatically generate models for making accurate predictions on the energy demand. This method is based on Artificial Intelligence (AI). The application of an AI technique is aimed at overcoming the limitations of traditional statistics based linear regression methods. Although these techniques and models are reliable, they are generally unable to adapt to unusual events, like for instance sudden changes in the weather conditions and varied holiday activities, which form a highly non-linear relationship with the daily load. Hence, their load predictions in the presence of such events are often not as satisfactory as desired, and consequently, more sophisticated methods are needed in order to accurately map the correlation between all the variables. AI methods for forecasting have shown an ability to give better performance in dealing with non-linearity and other difficulties in modeling of time series. Their advantage lies mainly in the fact that they do not require any complex mathematical formulations or knowledge of predetermined relationships between inputs and outputs.
In this study, the focus is on Genetic Programming (GP) (Koza, 1992, Vanneschi and Poli, 2012), that is one of the youngest AI techniques. In particular, a recently defined and very promising variant of standard GP is proposed, integrating the concept of semantic awareness in the algorithm used to generate predictive models.
To analyze the appropriateness of the proposed computational method for energy Short-Term Load Forecasting (STLF), the energy consumption in a particular area of Italy has been investigated. As reported by the main Italian energy provider, electricity consumption in Italy is rising. Early data on the trend of the demand in 2010 showed a 1.8% increase compared to 2009, the highest positive variation from 2007 to date. The total energy required in Italy equaled 326.2 billion kWh. In 2010, the electricity demand was met at 86.5% with the national production and for the remaining part (13.5%) with the balance of electricity exchanged with foreign countries. This amount of imported energy makes Italy the world's biggest importer of electricity. Due to its reliance on expensive fossil fuels and imports, Italians pay approximately 45% more than the EU average for electricity.
The area under exam includes the regions of southern Italy. South Italy regions have been taken into account because they represent a very interesting test case, given the important contribution that they have in building the global Italian energy demand and considering the high variability of energy demand in this area in recent years. Moreover, this area is challenging for the STLF problem: while AI-based models and traditional statistics-based models are able to produce accurate predictions for a large part of the Italian area, the regional area considered in this study presents particular features that often cause the forecasting models to not produce satisfactory predictions. The main motivations for the high variability in terms of energy demand in this area are discussed in Section 3.1.
The paper is organized as follows: Section 2 presents the variant of GP proposed in this study for addressing the STLF problem. Section 3 describes the data that have been considered in this study and reports experimental results comparing the proposed approach to the standard GP algorithm and other state of the art techniques. Section 4 concludes the paper, highlighting the main contributions of this work. In the final part of the manuscript, appendices contain general introductions of basic concepts for non-experts; more in particular, Appendix A describes the energy load forecasting problem, while Appendix B introduces the standard GP method.
Section snippets
Methodology
Models lie in the core of any technology in any industry, be it finance, manufacturing, services, mining, or information technology. The task of data-driven modeling lies in using a limited number of observations of system variables for inferring relationships among these variables. The design of reliable learning machines for data-driven modeling tasks is of strategic importance, as there are many systems that cannot be accurately modeled by classical mathematical or statistical techniques.
Experimental study
This section begins with a brief description of the main characteristics of the regions of the south of Italy, motivating the fact that the study of these regions is particularly interesting for energy STLF. Subsequently, we describe the data, the used experimental settings and the obtained results for the energy STLF in that geographic area.
Conclusions
Electricity short term load forecasting is important for the power industry, especially in the context of the ongoing deregulation of the electricity market. Proper demand forecasts help the market participants to maximize their profits and/or reduce their possible losses by preparing an appropriate bidding strategy. In this study, the Short-Term Load Forecasting (STLF) problem has been considered, focusing on the energy demand of the South Italy regions. This area has been chosen due to the
Acknowledgments
The authors acknowledge project MassGP (PTDC/EEI-CTP/2975/2012), FCT, Portugal.
References (20)
- et al.
Semantically driven mutation in genetic programming
- et al.
Climate variations in Italy in the last 130 years
- et al.
A c++ framework for geometric semantic genetic programming
Genet. Program Evolvable Mach.
(2014) Neural networks: a comprehensive foundation
(1999)Multivariate isotonic regression and its algorithms, Wichita State University, College of Liberal Arts and Sciences
Department of Mathematics and Statistics
(2009)Promoting phenotypic diversity in genetic programming
Genetic programming: on the programming of computers by means of natural selection
(1992)Human-competitive results produced by genetic programming
Genet. Program Evolvable Mach.
(2010)- et al.
Approximating geometric crossover in semantic space
- et al.
Mixed-Data Classificatory Programs I — agglomerative systems
Aust. Comput. J.
(1967)
Cited by (36)
Forecasting electricity consumption based on machine learning to improve performance: A case study for the organization of petroleum exporting countries (OPEC)
2020, Computers and Electrical EngineeringCitation Excerpt :Moreover, the PSONN converges to the best accuracy, faster than the GANN and standard ANN. In [20] proposes the integration of semantic awareness (SA) with genetic programming (GP) called SA-GP for a short term load forecasting problem in electric power operations. The SA is incorporated to improve the performance of the GP algorithm.
Machine learning in energy economics and finance: A review
2019, Energy EconomicsCitation Excerpt :Electricity demand forecasting is one of the traditional domains for ML in energy economics. Many papers (e.g. Pao, 2006; Azadeh et al., 2008; Lai et al., 2008; Ardakani and Ardehali, 2014; Castelli et al., 2015) have built hybrid models to predict low-frequency energy demand trends in different countries. The overall structure and components of the majority of these papers look similar: a traditional time-series model, SVM, ANN model (in particular a WNN), and more recently feedforward and feedback multi-layer models.
Alignment-based genetic programming for real life applications
2019, Swarm and Evolutionary ComputationCitation Excerpt :On the other hand, the large majority of the existing semantic methods are indirect, in the sense that they use traditional syntax-based genetic operators and only accept newly created solutions based on some semantic criteria [5]. This feature makes GSGP suitable for addressing problems characterized by a vast amount of data, still maintaining an acceptable running time [8,9]. To continue this promising research stream, in this paper, we introduce a novel GP system, aimed at exploiting semantic awareness in a totally different way, compared to GSGP and GSGP_LS.
Forecasting mid-long term electric energy consumption through bagging ARIMA and exponential smoothing methods
2018, EnergyCitation Excerpt :Nevertheless, mid/long-term load forecasting assumes a particular importance for electric power utility planning. Even though precise short-term forecasting forms the basis of the electrical energy trade and spot price calculation [5], several decisions are made on the basis of mid/long-term energy demand forecasting, such as the construction of new generation facilities, the purchase of existing generating units, the development of transmission and distribution systems, among others [1]. Kaboli et al. [6] add that mid/long-term forecasting is especially interesting for companies operating in a deregulated environment, as it provides them precious insights on the market need of energy, the maintenance schedule of power plants, the fuel supplies and occasional needs of electrical energy imports/exports.