Electricity load forecasting: a systematic review
Isaac Kofi Nti, Moses Teimeh, […]Adebayo Felix Adekoya
Journal of Electrical Systems and Information Technology
Abstract
The economic growth of every nation is highly related to its electricity infrastructure, network, and availability since electricity has become the central part of everyday life in this modern world. Hence, the global demand for electricity for residential and commercial purposes has seen an incredible increase. On the other side, electricity prices keep fluctuating over the past years and not mentioning the inadequacy in electricity generation to meet global demand. As a solution to this, numerous studies aimed at estimating future electrical energy demand for residential and commercial purposes to enable electricity generators, distributors, and suppliers to plan effectively ahead and promote energy conservation among the users. Notwithstanding, load forecasting is one of the major problems facing the power industry since the inception of electric power. The current study tried to undertake a systematic and critical review of about seventy-seven (77) relevant previous works reported in academic journals over nine years (2010–2020) in electricity demand forecasting. Specifically, attention was given to the following themes: (i) The forecasting algorithms used and their fitting ability in this field, (ii) the theories and factors affecting electricity consumption and the origin of research work, (iii) the relevant accuracy and error metrics applied in electricity load forecasting, and (iv) the forecasting period. The results revealed that 90% out of the top nine models used in electricity forecasting was artificial intelligence based, with artificial neural network (ANN) representing 28%. In this scope, ANN models were primarily used for short-term electricity forecasting where electrical energy consumption patterns are complicated. Concerning the accuracy metrics used, it was observed that root-mean-square error (RMSE) (38%) was the most used error metric among electricity forecasters, followed by mean absolute percentage error MAPE (35%). The study further revealed that 50% of electricity demand forecasting was based on weather and economic parameters, 8.33% on household lifestyle, 38.33% on historical energy consumption, and 3.33% on stock indices. Finally, we recap the challenges and opportunities for further research in electricity load forecasting locally and globally.
Background
Electricity is the pivot in upholding highly technologically advanced industrialisation in every economy [1,2,3]. Almost every activity done in this modern era hinges on electricity. The demand and usage of electric energy increase globally as the years past [4]; however, the process of generating, transmitting, and distributing electrical energy remains complicated and costly. Hence, effective grid management is an essential role in reducing the cost of energy production and increased in generating the capacity to meet the growing demand in electric energy [5].
Accordingly, effective grid management involves proper load demand planning, adequate maintenance schedule for generating, transmission and distribution lines, and efficient load distribution through the supply lines. Therefore, an accurate load forecasting will go a long way to maximise the efficiency of the planning process in the power generation industries [5, 6]. As a means to improve the accuracy of Electrical Energy Demand (EED) forecasting, several computational and statistical techniques have been applied to enhance forecast models [7].
EED forecasting techniques can be clustered into three (3), namely correlation, extrapolation, and a combination of both. The Extrapolation techniques (Trend analysis) involve fitting trend curves to primary historical data of electrical energy demand in a way to mirror the growth trend itself [7, 8]. Here, the future value of electricity demand is obtained from estimating the trend curve function at the preferred future point. Despite its simplicity, its results are very realistic in some instances [8].
On the other hand, correlation techniques (End-use and Economic models) involve relating the system load to several economic and demographic factors [7, 8]. Thus, the techniques ensure that the analysts capture the association existing between load increase patterns and other measurable factors. However, the disadvantage lies in the forecasting of economic and demographic factors, which is more complicated than the load forecast itself [7, 8]. Usually, economic and demographic factors such as population, building permits, heating, employment, ventilation, air conditioning system information, weather data, building structure, and business are used in correlation techniques [7,8,9]. Nevertheless, some researchers group EED forecasting models into two, viz. data-driven (artificial intelligence) methods (same as the extrapolation techniques) and engineering methods (same as correlation the techniques) [9]. All the same, no single method is accepted scientifically superior in all situations.
Also, proper planning and useful applications of electric load forecasting require particular “forecasting intervals,” also referred to as “lead time”. Based on the lead time, load forecasting can be grouped into four (4), namely: very short-term load forecasting (VSTLF), short-term load forecasting (STLF), medium-term load forecasting (MTLF) and long-term load forecasting (LTLF) [6, 7, 10]. The VSTLF is applicable in real-time control, and its predicting period is within minutes to 1 h ahead. The STLF is for making forecasting within 1 h to 7 days or month ahead [11]. It is usually used for the day-to-day operations of the utility industry, such as scheduling the generation and transmission of electric energy. The MTLF is used for forecasting of fuel purchase, maintenance, utility assessments. Its forecasting period ranges from 1 week to 1 year. While the LTLF is for making forecasting beyond a year to 20 years ahead, it is suitable for forecasting the construction of new generations, strategic planning, and changes in the electric energy supply and delivery system [10].
Notwithstanding the above-mentioned techniques and approaches available, EED forecasting is seen to be complicated and cannot easily be solved with simple mathematical formulas [2]. Also, Hong and Fan [12] pointed out that electric load forecasting has been a primary problem for the electric power industries, since the inception of the electric power. Regardless of the difficulty in electric load forecasting, the optimal and proficient economic set-up of electric power systems has continually occupied a vital position in the electric power industries [13]. This exercise permits the utility industries to examine the dynamic growth in load demand patterns to facilitate continuity planning for a better and accurate power system expansion. Consequently, inaccurate prediction leads to power shortage, which can lead to “dumsor” and unneeded development in the power system leading to unwanted expenditure [7, 14]. Besides, a robust EED forecasting is essential in developing countries having a low rate of electrification to facilitate a way for supporting the active development of the power systems [15].
Based on the sensitive nature of electricity demand forecasting in the power industries, there is a need for researchers and professionals to identify the challenges and opportunities in this area. Besides, as argued by Moher et al. [16], systematic reviews are the established reference for generating evidence in any research field for further studies. Our partial search of literature resulted in the following [10, 12, 17,18,19,20,21] papers that focused on comprehensive systematic review concerning the methods, models, and several methodologies used in electric load forecasting. Hammad et al. [10] compared forty-five (45) academic papers on electric load forecasting based on inputs, outputs, time frame, the scale of the project, and value. They revealed that despite the simplicity of regression models, they are mostly useful for long-term load forecasting compared with AI-based models such as ANN, Fuzzy logic, and SVM, which are appropriate for short-term forecasting.
Similarly, Hong and Fan [12] carried out a tutorial review of probabilistic EED forecasting. The paper focused on EED forecasting methodologies, special techniques, common misunderstandings and evaluation methods. Wang et al. [19] presented a comprehensive review of factors that affects EED forecasting, such as forecast model, evaluation metric, and input parameters. The paper reported that the commonly used evaluation metrics were the mean absolute error, MAPE, and RMSE. Likewise, Kuster et al. [22] presented a systematic review of 113 studies in electricity forecasting. The paper examined the timeframe, inputs, outputs, data sample size, scale, error type as criteria for comparing models aimed at identifying which model best suited for a case or scenario.
Also, Zhou et al. [17] presented a review of electric load classification in the smart-grid environment. The paper focused on the commonly used clustering techniques and well-known evaluation methods for EED forecasting. Another study in [21] presented a review of short-term EED forecasting based on artificial intelligence. Mele [20] presented an overview of the primary machine learning techniques used for furcating short-term EED. Gonzalez-Briones et al. [18] examined the critical machine learning models for EED forecasting using a 1-year dataset of a shoe store. Panda et al. [23] presented a comprehensive review of LTLF studies examining the various techniques and approaches adopted in LTLF.
The above-discussed works of literature show that two studies [20, 21] address a comprehensive review on STLF, [23] addresses forecasting models based on LTLF. The study in [24] was entirely dedicated to STLF. Only a fraction (10%) of above systematic review studies included STLF, MTLF and LTLF papers in their review; however, as argued in [10], the lead time (forecasting interval) is a factor that positively influences the performance of a chosen model for EED forecasting studies. Again, a high percentage of these studies [10, 12, 17,18,19,20,21,22, 24] concentrated on the methods (models), input parameter, and timeframe. Nevertheless, Wang et al. [19] revealed that the primary factors that influence EED forecasting models are property (characteristic) parameters of the building and weather parameters include. Besides, these parameters are territorial dependant and cultural bond. Thus, the weather pattern is not the same world-wide neither do we use the same building architecture and materials globally.
Notwithstanding, a higher percentage of previous systematic review studies overlooked the origin of studies and dataset of EED forecasting paper. Also, only a few studies [12, 17, 19] that examined the evaluation metrics used in EED forecasting models. However, as pointed out in [17], there is no single validity index that can correctly deal with any dataset and offer better performance always.
Despite all these review studies [10, 12, 17,18,19,20,21,22, 24] on electricity load forecasting, a comprehensive systematic review of electricity load forecasting that takes into account all possible factors, such as the forecasting load (commercial, residential and combined), the forecast model (conventional and AI), model evaluation metrics and forecasting type (STLF, MTLF, and LTLF) that influences EED forecast models is still an open gate for research. Hence, to fill in the gap, this study presents an extensive systematic review of state-of-the-art literature based on electrical energy demand forecasting. The current review is classified according to the forecasting load (commercial, residential, and combined), the forecast model (conventional, AI and hybrids), model evaluation metrics, and forecasting type (STLF, MTLF, and LTLF). The Preferred-Reporting Items for Systematic-Review and Meta-Analysis (PRISMA) flow diagram was adopted for this study based on its ability to advance the value and quality of the systematic review as compared with other guidelines
……
most used algorithms for electricity forecasting
Conclusions
The current study sought to reviewed state-of-the-art literature on electricity load forecasting to identify the challenges and opportunities for future studies. The outcome of the study revealed that electricity load forecasting is seen to be complicated for both engineers and academician and is still an ongoing area of research. The key findings are summarised as follows.
1.
Several studies (90%) have applied AI in electrical energy demand forecasting as compared with traditional engineering and statistical method (10%) to address energy prediction problems; however, there are not enough studies benchmarking the performance of these methods.
2.
There are few studies on EED forecasting in Africa countries (12 out of 67). Though the continent has progressive achievement in the creation of Regional Power Pools (PPP) over the last two decades, the continent still suffers from a lousy power network in most of its countries, leaving millions of people in Africa without electricity.
3.
Temperature and rainfall as an input parameter to the EED forecasting model are seen to have a divergent view. At the same time, some sections of research recorded an improvement in accuracy and reported no improvement in accuracy when introduced and input. However, the current study attributes this to the difference in automorphic temperature globally and the different economic status among countries. An additional investigation will bring more clarity to the literature.
4.
This study revealed that EED forecasting in the residential sector had seen little attention. On the other hand, Guo et al. [78] argue that the basic unit of electricity consumption is home.
5.
It was observed that there had been a global increase in residential electricity demand, this according to the report in [49] can be attributed to the growing rate of buying electrical equipment and appliances of low quality due to higher living standards. However, a further probe into Soares et al. [49] assertion will bring clarity to literature because of the discrepancy in opinions in literature.
6.
Lastly, the study revealed that there is a limited number of studies on load forecasting studies in Ghana. We, therefore, recommend rigorous researchers in this field in the country to enhance the economic growth of the country.
Our future study will focus on identifying the relationship between household lifestyle factors and electricity consumption in Ghana and predict load consumption based on identified factors since it is an area that has seen little or no attention in Ghana.
