Ethiopian energy consumption forecast

This Energy consumption forecast is vital and has a great economic impact. Mathematical models developed for energy forecast can also serve as inputs for further studies. This study is intended to develop an energy consumption forecast using the grey prediction model GM (1,1), based on the actual energy consumption data from the year 2008 to 2017. The models are developed for the total, solid biomass, oil products, and electrical energy consumption; and the accuracy for each model is ratified. These developed forecasting models were used to anticipate six-year Ethiopian consumption of major energy types. The outcomes of models for all four energy consumption types show an upward trend; simulating and forecasting are found suited with the grey system model with development coefficient values less than 0.3 for all selected energy forms


Introduction
Ethiopia is among the least-developed nations in the world. About 34% of its population live below the poverty threshold. Access to modern energy resources and energy creation is at the lowest level, and is mainly reliant on biomass. The energy segments of the country are estimated to be 92.4% of waste and biomass, 5.7% oil and 1.6% hydro-power sources [1]. However, the World Bank report shows that the economy is among the fastest rising, with an annual growth rate of 9.8% since 2008. The country has also set a plan and is striving to realize its development till 2025 with climate resilient actions. The country's commitment to the climate issue was proven by taking the lead in delivering the Nationally Determined Commitment of all developed countries to the United Nations Framework Convention on Climate Change.
It is known that the social and economic growth of a country is mainly supported by energy, one of the strategic resources. As a nation occupied with an incredibly yearning and quick energy improvement program, Ethiopia has a feature of harvesting mainly of hydro based renewable energy and targeting to increase the electricity access from 26% in 2014 to 60% by 2040 [2]. The country's clean energy demand versus the supply gap creates an unsuitable and unhealthy living status; besides to this shortage of access to clean energy brings resemblance of the citizens to use biomass-based energy sources as a first choice. The dependence on biomass-based energy sources in-turn becomes a cause for deforestation and leads to soil erosion and fertility loss for which the economy of a country is hindered by. To minimize the environmental loads and improve the living status of the community, the energy sector should properly be managed with a forecast of short and long term strategy.
Forecasting is a technique which uses historical data as inputs to make informed estimates that are predictive in determining the direction of future trends; now a days it is becoming a powerful tool for policy makers in different sectors and industries. With this regard, forecasting the energy consumption trend of Ethiopia is taken as the focus of this work for better decisionmaking process.
It is believed that energy consumption is an intricate scheme with vulnerable and limited helpful data, and hence its prediction is a troublesome issue. Issues related to economy, urbanization, industrialization and the environment make energy remain obscure to a great extent [3]. In the recent decades studies related to energy problems are getting due attention with a special focus on developing countries in which the energy system is evolving at a high level [4]. The study done on the energy demand-forecasting methods summarized and showed that different models such as regression, time series, grey model and some others can come up with better prediction accuracy [4][5]. However, these models have limitations of demanding 30 to 100 or more initial data/observations, it is therefore meaningful to find an efficient and feasible model for forecasting the pattern of an evaluated structure with scarce observations. In poor countries like Ethiopia, where proper data of a long-range of time and full information are not found, it is reasonable to use grey system theory in building forecasting models [6]. The grey forecasting theory which initially proposed by Deng [7] mainly addresses ambiguous problems with limited data and minimal knowledge. Now a day, grey system theory is getting a wider application in various manufacturing and service industries, education, and environment related fields [8][9][10][11][12]. Though lots of Grey model (GM) types are developed so far, its computational strength and ease to handle makes the first-order grey model be a favoured tool for different research works [13]. The GM(1,1) prediction model seems to obtain the lowest postforecasting errors and it is more suitable to make a shortterm prediction [14].
Procedures for the GM(1,1 ) model are briefly discussed by Xiao et al. [15] as the starting procedure called the whitening equation is a differential equation, preceded by discretising this differential equation so that the basic form is attained. Thereafter, evaluating the system parameters using the least-squares method, after that using the whitening equation and system parameters the data sequence is simulated and predicted. Just four or more samples are used in such a model to analyse the system's characteristics and evolution.
Different studies on Ethiopia's energy scenario have been done so far, some of which are done in the last few years focusing on access to sustainable energy [16], power sector development [17], residential electricity demand [18], Regional urban energy use trend [19], allocation of biomass energy sources [20] review on bigas technology [21], energy scenarios for sustainable strategies [22], and very few on energy demand projection using different approaches and tools. However, to the author's knowledge, there is no literature providing a comprehensive analysis of Ethiopian energy demand projections for different energy source types using the grey theory.
Such models would enable policy makers and energy planners to reconcile the energy supply with the projected growth in demand. With all above listed benefits and for its significant importance to counter-act against energy crisis, and structure the future energy demand recommended by Brinckerhoff [23], Ethiopian major energy types of consumption forecast is developed. It is a new work especially in East Africa for Ethiopia. The author believes that this paper will be a valuable input and have a better acceptance as it refers to the case of Ethiopia where the trend of having focus to energy is yet not developed. Besides to this such energy forecasting research could help other developing countries adopt it and have a related work on, for better energy planning and management.

The Grey Model (1,1)
Grey systems theory has been grown up over almost thirty years of growth as a recently emerging scientific field with its own theoretical framework comprising of system analysis, measurement, simulation, estimation, decision-making, control, and optimization techniques. One of the fundamental pieces of this theory is the Grey model. This approach accepts grey module as its establishment sorting out apparently non continuous data sequences through data processing. Afterwards model values will be computed and obtained from the new sequentially built differential equations. Residual errors are possibly be found by taking the real measured value as a reference, and hence the model can be rectified and thus be conferred with a better degree of forecasting accuracy. There are numerous sorts of grey forecasting models, yet Grey Model (1,1) is the commonly utilized type, besides this different grey based models are risen up from it. Meanwhile GM(1,1) is characterized by differential conditions to portray the inner and evolving behaviour of the system predictability for extrapolation.

The Grey
Model (1,1) Development 1. By having some equivalent time span sections of a case, ordered values are attained and are called as original numerical sequences (0) (Eq. 1).
The arbitrator for randomness of the initial data is treated by generating the First-Order Accumulated Generating Operation (IAGO) from the original time sequence (0) , as Eqs. 2 and 3. Where: Whenever (0) values are non-negative, the (1) sequence values will have an increasing trend, with an exception value of (0) (1)= (1) (1).
2. First, we have to use quasi-smoothness sequence and quasi exponential law to verify that GM(1,1) can be built, if both are fulfilled at a time then the criteria is satisfied to build the model.

The Grey Model (1,1) Precision Verification
The developed model through the above procedures can be ratified using different techniques. Posterior error, Residual error and Degree of grey association are among the most commonly applied methods to check and confirm the accuracy of Grey Model(1,1). With the recommendation of Huang et al. [24] and Lian et al. [25], this research prefers to use 'Posterior error' for model accuracy checking. The following are steps implemented in checking models.
Step 2: Finding for an average ( − ) and root-meansquare error ( 2 2 ) of deviation errors (residual errors) First we have to find the absolute residual, ( ) = The residual variance S 2 2 is then find from Eq. 16.
Step 3: Finding for mean square deviation ratio and small error probability The common formula used to verify the forecasting accuracy is Eq. 18.
As can be seen from Table 1, the accuracy classes indicate that as worth higher, probability of leading to a minor error gets larger; thus, when the value is small, the value becomes large; hence prediction accuracy becomes nice. If and maintain within a permissible range, we can use Grey Model (1,1) for prediction.

Energy Consumption Prediction with the Grey Model (1,1)
Using the procedures and steps described in section 2.2 GM(1,1) model for electric, oil products, solid biomass and total energy consumption have developed for the country based on the data of year 2008-2017 adopted from the Knoema library catalog [1]. With these the original sequence (0) shown in Table 2 has developed.
In this case, we use the consumption of electric energy as an example to illustrate how these models are built and measured. Table 2 Electric, oil products, solid biomass and total energy consumption data of Ethiopia from 2008 to 2017, (Unit: kilotons of oil equivalent (ktoe))

S. No. Year
Original Using Table 1, precision of the established model is in a 2 nd precision level, where it can qualify to be used for making prediction. Calculated values are shown in Table 3. Based on Ghalehkhondabi et al. [26], mid-longterm forecasting can be done using grey system as long as the development coefficient − is < 0.3. This implies that the model can be utilized to forecast the country's electric energy utilization for the year 2020-2025. Findings from numerical manipulations using Eq. 6-9 in section 3, the development coefficient for the model becomes − = 0.1287 < 0.3, and is given as under.

Sample years
Real  Fig. 2. Ethiopian oil products energy consumption forecast Table 5 Predicted energy in (ktoe) & deviations of Ethiopian solid biomass energy utilization (1ktoe is equivalent to 11.63 kWh)

S.
No.

Sample Years
Real   Based on the utilization data of each energy types from the year 2008 to 2017, prediction models of the country's electric, oil, solid biomass and total energy respectively are developed for the coming six years as shown below.

Conclusion
In this study, electricity, oil products, solid biomass, and total energy consumption prediction of Ethiopia from 2020 to 2025 were made based on the grey model. Accordingly, the following conclusions are drawn.
The stimulative accuracy level of the forecasting models examined through the \posterior error" method found for electric, oil products, solid biomass and total energy are: 94.83%, 97.65%, 99.78% and 99.78 respectively. These stimulative accuracy level values except for electric energy consumption model all lie in the first precision level range. However, all the developed models met the standards for mid to long term prediction.
In comparison to the 2017 (last actual data used), findings indicate that electric, oil products, solid biomass, and the total energy consumption rise averagely at a rate of 24.83%, 14.75%, 2.84% and 3.86% respectively every single year. This rising pattern could possibly be due to the rapid population growth and the country's strive to attain the second growth and transformation plan. The findings from the forecast indicate that the country's energy structure seems to mainly depend on solid biomass energy sources followed as a secondly consumed energy of oil products.
The overall level of accuracy is higher as more initial data is acquired. In general, these prediction results can be used as references in providing information to the policy makers, and other concerned governmental and non-governmental sectors in charge, especially in establishing and adjusting the energy consumption structure in relation to the environmental and economic values.