Vishwa Ranjan

Will India achieve its target to half its oil import burden by 2030 - A multi-factor study A Dissertation Proposal for Executive Post Graduate Program in Management (NMP) by Vishwa Ranjan under the guidance of Shri GSP Singh Chief General Manager (City Gas Distribution) Pipelines Head Office, Noida Indian Oil Corporation Limited (IOCL) Dr. Sajal Ghosh Associate Professor, Economics Area Chair– Economics, MDI, Gurgaon Management Development Institute Gurgaon-th November 2019 1| P a g e Abstract “Will India achieve its target to half its oil import burden by 2030 - A multi-factor study” by Vishwa Ranjan Introduction: There has always been a very strong relationship between energy consumption and economic growth of a country. After the historic introduction of economic liberalization in 1991, India has constantly featured among the top growing major economies in the world. Along with China, India has been the flagbearer of the high economic growth saga in the South East Asia region. With GDP at $2.972 trillion, about to touch $3 trillion and average growth rate of 6-7% in the past decade, India has emerged as the hotspot of economic activities in the world. In order to sustain its young population (1/3rd of India’s population are under 18 years of age), needs the fastest economic growth in order to sustain its growing aspirations. Apart from coal which is abundant in India, crude oil has the largest share in the energy mix of India. But sadly, domestic production of crude oil never picked up its desired volume forcing India to import about 83.9% of its crude oil requirements from OPEC and non-OPEC Nations. Indian Prime Minister in 2015, had set the target of reduction of 10% in its crude oil import from 77% in 2013-14 to 67% in 2022 and to 50% by 2030. Import dependency is defined as the ratio of crude oil import to total crude oil requirement. Despite several policy interventions by the Government of India such as replacing NELP1(Profit sharing model) to HELP2(Revenue Sharing Model) in 2017, crude oil production in India has been stagnant and even experience a decline of 2.2% as it reduced to 39.5 Million Tonnes in 2018 from 40.4 Million Tonnes in 2017. It is evident from data of various components of energy mix of India that path to reduce its crude oil import dependence lies on replacing its substantial portion of energy mix by energy sources which are abundant in India such as Coal or have lesser import dependence such as Natural Gas. Apart from Natural Gas and Coal the answer lies with maximum adoption on renewables such as solar, wind, hydel, biogas, biofuel, biomethane etc. Nuclear energy and hydrogen-based fuel cells also have picked up recently as clean sources of energies. Recently, NITI Aayog, the official thinktank of Government of India has come up with India’s energy and emission outlook to 2047 in order to chart out its detailed plan to transform its energy mix with dual aim to reduce its import dependence and to reduce its carbon footprint to honor its commitment made at COP-21 at Paris in 2015. India Nationally Determined Contributions (NDCs) 1 New Exploration Licensing Policy http://petroleum.nic.in/sites/default/files/2exp.policy.NELP2015.pdf 2 Hydrocarbon Exploration and Licensing Policy https://www.pmindia.gov.in/en/news_updates/hydrocarbon-exploration-and-licensing-policy-help/ 6| P a g e set three major goals - increase the share of non-fossil fuels to 40% of the total electricity generation capacity, to reduce the emission intensity of the economy by 33 to 35% by 2030 from 2005 level, and to create additional carbon sink of 2.5 -3 billion tonnes of CO2 equivalent through additional forest and tree cover. Problem Formulation: Energy forecasts have been a pet subject among researchers. But there have been very few studies on the energy mix of India in the light of the Govt’s vision to reduce its import dependence and its commitment before the world to bring down its emission intensity to the level of 2005. This study aims to suggest an ideal yet achievable energy mix of India in 2030 and derive its crude oil import dependency from the energy mix of India determined by forecasting its total energy requirement by 2030 and then filling it by the energy sources which ranks higher in terms of cleanliness as well as self-sufficiency. In order to do this, we have forecasted the energy generated by Coal, Natural Gas, Renewables, Hydroelectric and Nuclear sources and subtracted that from the total energy demand by 2030 to arrive at the share of energy required to be generated from crude oil. This study makes the substantiated assumption about the complete replacement of crude oil by these sources in light of the recent policy initiatives and enhanced government spending on these sources. It also leaves us with future scope of determining the exact extent of replacement of crude oil by these alternative sources of energy. With the availability of data from 1965(in some cases from 1981) up to 2018, the forecasts have been done for next 12 years till 2030. Research Methodology: Study of energy forecasts on the basis of its yearly demand, availability, consumption and production primarily falls under the realm of non-exploratory research and the univariate time series nature of data demands the application of forecasting techniques such as ARIMA. Since the types of energy sources considered for this study and their different aspects such as consumption and production make the consideration set pretty huge we have restricted ourselves to univariate ARIMA3 which is considered to be one of the best statistical method for forecasting timeseries data. We have used the hottest language among the new generation data scientists i.e. python. In order to remove manual error, we have used auto-arima model based on the AIC method of model selection. For this study, non-exploratory research method with secondary data collection for potential endogenous/exogenous variables have been followed. Various endogenous variables were identified based on the discussion with guide, industry experts, previous researches in the field, various reports published by different arms of government such as NITI Aayog, non-govt international agencies such as IEF, WEF and a few world energy leaders such British Petroleum. Data Collection: This study required historic data of different variables which forms the energy mix of a country. In order to ensure uniformity and fairness, most of the data have been collected from the 68th 3 How to create an ARIMA model forecast for Time series forecasting https://machinelearningmastery.com/arimafor-time-series-forecasting-with-python/ 7| P a g e edition of BP Statistical Review of World Energy4 published by energy giant British Petroleum in 2019. Most of academic research conducted around the world considers BP Statistical Review of World Energy as the most reliable source of energy data. However, data related to India have been verified against the data published by Ministry of Petroleum and Natural Gas, Govt of India and were found consistent. Some portion of the data were also matched to the extent possible with Indiastat5, an online source of data available through e-library of MDI. Based on the requirement of the study and recommendations by various industry experts following data were collected for this research: • • • • • • • • • Energy Consumption of India, China, US and World in Mtoe (from 1965 to 2018) Per Capita Energy Consumption of India, China, US and World in Gigajoule (from 1965 to 2018) Production and Consumption of Crude Oil in India, China, US and World in Million Tonnes (from 1965 to 2018) Production and Consumption of Natural Gas in India, China, US and World in Mtoe (from 1965 to 2018) Production and Consumption of Coal in India, China, US and World in Mtoe (from 1965 to 2018) Production and Consumption of Coal in India, China, US and World in Mtoe (from 1965 to 2018) Production and Consumption of Nuclear Fuel in India, China, US and World in Twh/Mtoe (from 1965 to 2018) Production and Consumption of Hydroelectricity in India, China, US and World in Twh/Mtoe (from 1965 to 2018) Production and Consumption of renewables (from all sources) in India, China, US and World in Twh/Mtoe (from 1965 to 2018) Methodological Approach/ Data Analysis and Modeling: In course of this study, keeping in mind the wide spectrum of data to be analyzed, we restricted ourselves to using only univariate ARIMA model for forecasting future production/consumption of various components of energy mix of India. In place of manual selection of model based on ADF and correlogram of various series, we chose to go an AIC based model selection. Visual inspection of correlogram can sometimes be very confusing and the researcher has to resort to trial and error in order to select the best model. AIC is considered to be the most suitable estimator of out of sample prediction error, in other terms gives the quality of statistical models for a given set 4 BP Statistical Review of World Energy https://www.bp.com/content/dam/bp/businesssites/en/global/corporate/pdfs/energy-economics/statistical-review/bp-stats-review-2019-fullreport.pdf 5 Indiastat website https://www.indiastat.com/ 8| P a g e of data. AIC in other words tell us how close a model is with the reality if at all it exists. AIC efficiently handles the tradeoff between goodness of fit as well as the simplicity of model and hence keeps the model free from the risk of both overfitting as well as underfitting. An attempt was also made to use Long Short-Term Memory6 (LSTM) model for in-sample and out of sample forecasting methods as an alternate method. But since the sample size was quite small < 70, training LSTM proved to be a difficult task and in most of the cases ARIMA came out as a better method with lower MAPE. Findings & Recommendations: The basic premise of this study was to arrive at the crude oil demand in 2030 by calculating the total energy requirement and subtracting the sum of all other sources of energy forecasted in 2030 with an assumption that the time series data of other sources of energy will reflect the impacts of recent policy interventions of the Government of India. The study shows that the share of Crude Oil In the energy mix of India is not expected to undergo any drastic change. Scenario -1 shows a marginal decline of 2% whereas scenario-2 which was expected to exhibit a greater decline show even lesser decline of just by 0.48%. Coal will retain its prominence in the energy mix of India. In scenario one its share will marginally increase from 55.9% in 2018 to 56.03% in 2030 but whereas in scenarios 2 its share will see a negligible fall. Share of Natural Gas, Nuclear and Hydropower will remain stagnant. Only silver lining is the doubling of share of renewables in both the scenarios. Based on our study India’s import oil dependency by 2030 will touch 90%(89.42% in scenario-1 and 90.19% in scenario-2). Target to achieve 50% reduction in oil import dependency thus sounds too ambitious. Short term target should be set accounting for all domestic and international ground realties. Focus should be on increasing the contribution of renewables such as solar, wind, tidal, biofuels and biogas etc. In Northern India, burning of agricultural residuals poses a grave environmental threat and pollution concerns. Futuristic technologies of converting agricultural wastes and residuals to fuel or 2-G bio-ethanol should be promoted on an unprecedented scale. In order to make these technologically viable, a complete ecosystem from collection of agricultural wastes and residuals to its final processing should be developed. Nuclear fuel has huge potential in India as it has large Plutonium reserves and has access to Nuclear Supplier Group for nuclear fuel. But the fuel which has the potential to make maximum impact is Natural Gas. Unlike Crude Oil, Natural Gas is a buyer’s market. Natural Gas can replace solid and liquid fuels in all sectors such 6 Understanding RNN and LSTM https://towardsdatascience.com/understanding-rnn-and-lstm-f7cdf6dfc14e 9| P a g e as transport, industrial, power generation as well as domestic cooking. After the 9th and 10th round of CGD bidding half of India’s geography and 70% of its population has been covered under City Gas Distribution. Directed programmes and mass awareness campaigns need to be carried out to make people migrate from liquid fuels and LPG to CNG and PNG. Last but not the least long-term target should be set and corresponding infrastructures must be created to increase the share of Natural Gas, Renewables, Hydropower and Nuclear fuels. The complete ecosystem for electric vehicles must be created. 10| P a g e Figure 14 ARIMA Model for Consumption of Energy in India The selected model is ARIMA (2,2,1) with AIC = 249.430. By plotting the model diagnostics, we get Figure 15 Model Diagnostics - Consumption of Energy in India It is clear that Standardized residual has reached the white noise and estimated density of errors follows a normal distribution. The we proceed to in-sample forecasting by forecasting 20% of the sample size and calculated MAPE of the forecast. 41| P a g e Figure 16 In Sample Forecast - Consumption of Energy in India MAPE of the model comes out to be 1.55% for Test Data of 20% of the sample, which can be considered as very good. Finally, we proceed to forecast the energy consumption of India for the period-. Figure 17 Final Forecast - Consumption of Energy in India 4.1.2 Forecasting Crude Oil Consumption/Demand of India Figure 18 Consumption of Energy in India 42| P a g e Plot of India’s Crude Oil Consumption Data -) showed no missing or abnormal data. Figure 19 ACF and PACF for Crude Oil Consumption Data ACF is exhibiting a fast decay and suggest the presence of trend in the series. Figure 20 Decomposition of Crude Oil Consumption Data Decomposition of series shows that there is only trend component and seasonality component is absent. Figure 21 ARIMA Model for Crude Oil Consumption Data The selected model is ARIMA (0,2,2) with AIC = 178.109 Figure 22 Model Diagnostics for Crude Oil Consumption Data 43| P a g e Model diagnostics confirms that Standardized residual has reached the white noise and estimated density of errors follows a normal distribution. The we proceed to in-sample forecasting by forecasting 20% of the sample size and calculated MAPE of the forecast. Figure 23 In Sample Forecast of India’s Crude Oil Consumption MAPE of the in-sample forecast comes out to be 3.40% which again can be termed as very good. Finally, we proceed to forecast the Crude Oil consumption of India for the period-. Figure 24 Final forecast of India’s Crude Oil Consumption 44| P a g e 4.1.3 Forecasting Gas Consumption of India By following the same procedure, forecast of Gas Consumption of India was done. Final forecast is summarized as below: Model : ARIMA(1,2,1) AIC : 57.256 MAPE : 12.37% Figure 25 ARIMA Model for Gas Consumption Figure 26 Final forecast for India’s Gas Consumption 45| P a g e Chapter 5: Results and Discussion on findings 5.1 Inter Model Comparison Findings of both models i.e. ARIMA and LSTM were compared based on the MAPE Errors of the model. Table 2 Model Comparison between ARIMA and LSTM Sl.no Subject Forecast of ARIMA Model (p,d,q) ARIMA MAPE (%) LSTM MAPE (%) Selected Model (2,2,1) (0,2,2) 1.55% 3.40% 5.49% 56.73% ARIMA ARIMA Quality of Forecast based on MAPE <10%- High 10%-20% - Good 20%-50%- Reasonable High High 1 2 3 4 Energy Crude Oil Demand Natural Gas Coal (1,2,1) (2,2,1) 12.37% 3.49% 21.74% 7.37% ARIMA ARIMA Good High 5 6 Nuclear Hydropower (0,2,2) (1,2,1) 5.39% 18.03% 52.7% 21.1% ARIMA ARIMA High Good 7 Renewables (3,2,1) 8.26% 79.9% ARIMA High 8 Crude Oil (1,2,1) Production 7.37% 6.73% LSTM High In 7 out of 8 cases, ARIMA performed better than LSTM. This is contrary to popular beliefs. The reason for such behavior could be small sample size of 55 data points. LSTM required bigger sample to learn effectively and predict accurately. For further calculations and study only ARIMA results were considered due to their overall better performance than LSTM. 5.2 Summary of forecast findings Results of forecasting of demand of Energy, Crude Oil, Natural Gas, Coal, Nuclear, Hydropower and renewables can be summarized as below: 58| P a g e Table 3 Model Summary Sl.no Subject of ARIMA AIC Forecast Model (p,d,q) 1 2 MAPE (%) Value Forecasted Quality of Forecast in 2018 value in based on MAPE (Mtoe) 2030 <10%- High (Mtoe) 10%-20% - Good 20%-50%- Reasonable- High- High 3 4 Energy Crude Oil Demand Natural Gas Coal (2,2,1) (0,2,2) -%-% (1,2,1) (2,2,1) -%-% 452.2 - Good High 5 6 Nuclear (0,2,2) Hydropower (1,2,1) -%-% 31.6 - High Good 7 Renewables -29.525 8 Crude Oil (1,2,1) Production (3,2,1) 8.26% 27.5 83.54 High -% 39.5 38.31 High 5.3 Scenario Analysis of Energy Mix of India in 2030 After the modal comparison, only ARIMA Model output was considered based on the lower values of MAPE Errors. Based on the above forecast results Energy Mix of India in 2030 has been calculated using two scenarios: Scenario 1: In scenario 1, demands for each component of energy mix of India was individually forecasted. Energy mix was derived after the summation of each component’s projected values in 2030. Scenario 2: In this scenario, first we forecasted the total energy requirement of India in 2030. And then we forecasted all components of Energy Mix of India other than that of crude oil. After that we subtracted the sum of all components from the total energy requirement of India to arrive at the demand of crude oil in 2030. The assumption was made that difference of crude oil requirement forecasted in scenario 1 and scenario 2 will give the amount of crude oil replaced or imported in excess owing to the changed composition of energy mix. These are summarized in the following table: 59| P a g e Table 4 Energy Mix Of India 2018 vs 2030 Energy Mix of India 2018 Components of Energy In Mtoe Mix Crude Oil 239.1 Natural Gas Coal - Nuclear Hydropower Renewables Total Energy Demand - 2030 (Scenario 1: Based on Individual Forecast) 2030 (Scenario 2: Based on Replacement Assumption) % In Mtoe % In Mtoe % 29.5 % 6.2% 55.9 % 1.1% 3.9% 3.4% 362.07 27.49% 390.47 29.02% - 6.00% 56.03% - 5.87% 54.85% - 1.06% 3.08% 6.34% - 1.04% 3.02% 6.21% After this we calculated the import oil dependence in 2018 and two scenarios explained above in 2030 using the following formuls : Import Oil Dependence = 1- (Crude Oil Production/Total Crude Oil Demand) Table 5 Oil Import Dependence 2018 vs 2030 Oil Import Dependence 2018 2030(Scenario-1) 2030(Scenario-2) 60| P a g e Crude Oil Demand Crude Oil Production Import Dependency Crude Oil Demand Crude Oil Production Import Dependency Crude Oil Demand Crude Oil Production Import Dependency -%-%-%