Renewable and Sustainable Energy Reviews 81 (2018) 1879–1886

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

Forecasting the impact of renewable energies in competition with MARK

non-renewable sources
⁎
Claudia Furlan , Cinzia Mortarino
Department of Statistical Sciences, University of Padua, Italy

A R T I C L E I N F O A BS T RAC T

Keywords: The diffusion dynamics of traditional and clean energy systems, in the energy mix, are studied in a competition
Competition modeling, where their interacting life cycles are jointly described. Competition has only been considered
Energy policy recently since diffusion competition models are not easy to be implemented. Data are represented by yearly
Renewable energy consumptions (in Mtoe) of traditional sources (coal, oil, gas, and nuclear systems) and a selection of renewable
Diffusion of innovations
sources (hydroelectric, wind, and solar power) for the US, Europe, China, and India from 1965 to 2014. For
each case, we will investigate whether the diffusion level of traditional sources sustains or prevents the spread of
renewables and vice versa. For the US and Europe, we will take into account the link between the 2008–2009
financial crisis and the slowdown in consumptions. As a matter of fact, the energy sector is characterized by
uncertainty due to the depletion of finite energy sources, the technological and economical problems of
renewables, and the hypothesis of shale gas as a possible bridge to renewables that could delay the spread of
latter ones. In this scenario, six-year forecasts are provided until 2020.

1. Introduction
By now, it is a matter of fact that the energy mix is slowly changing, but
it is still uncertain which direction will be feasible in both the short and long
term. Clean technologies have great potential for producing electricity, but
their development is slow and depends on several factors, among them
incentive schemes and technological problems that need to be overcome.
Hydroelectricity often represents an exception among the clean technolo-
gies since, in many countries, its exploitation is stable. Non-renewable
energies are moving to the last phase of their life cycle, even if it may
happen that a new technological innovation can increase the accessibility to
a source, at least for a while. For example, fracking has given a new era of
gas through shale gas, especially in the United States. The new growing
trend of natural gas was unpredictable until a decade ago through national
historical data. In fact, new technological discoveries can suddenly change
the life cycle of a source. However, even if the life cycle of a particular
source is hypothetically extended by a technological discovery, a sensible
choice could be waiting before exploiting that direction in order to
understand if the new technology can be considered safe. For example,
Klein and Whalley [1] encourage caution in considering shale gas as the
bridge to renewables since it has negative environmental impacts due to
hydraulic fracturing [2–4]. Its effects are worse than the natural gas from
conventional sources, its environmental impact (groundwater and surface
water contamination and wastewater generation, fugitive methane emis-
sions, and increased seismic activity) has not yet been adequately discussed
[3], and its drilling operations have not yet been adequately regulated [5].
Other countries, such as China [6] and India [7,8], are considering shale
gas, but these negative effects could change the investments in the energy
sector.
Given this uncertainty, the forecasts released in 2013 by the EIA say
that world energy consumption will increase by 56% between 2010 and
2040 [9], mostly due to non-OECD countries, and the energy market will
expand. In this market, the different energy systems behave as competitors.
The entry of new technological innovations may compensate for the exit of
technologies for depleted energy sources. At the same time, new technol-
ogies may foster the expansion of the market potential size, especially for
emerging countries. Usually, it takes time before a new technological
innovation arrives to represent a non-negligible market share. This has
been the case of green technologies that, for years, have faced difficulty in
being adopted due to technological discrepancies with the pre-existing
electric system and because they strongly rely on incentive schemes to be
economically competitive. Moreover, the competitive effects would also
affect the time entry of new technologies and life cycle of pre-existing
technologies.
The contribution of statistics here, in the diffusion of an innovation
field, is to provide estimates of the market size of each energy system in
the future energy mix by extrapolating information from historical
data. The best way to deal with this is to use a single complex system in

⁎
Corresponding author.
E-mail addresses: furlan@stat.unipd.it (C. Furlan), mortarino@stat.unipd.it (C. Mortarino).

http://dx.doi.org/10.1016/j.rser.2017.05.284
Received 1 July 2016; Accepted 29 May 2017
Available online 09 June 2017
1364-0321/ © 2017 Elsevier Ltd. All rights reserved.
C. Furlan, C. Mortarino
order to correctly identify the competition effects [10]. In the past,
several attempts were made by analyzing one source at a time, both
renewable and non-renewable, in a univariate diffusion framework—for
example: oil [11], natural gas [12,13], nuclear energy [14–16], wind
power [17,18], biomass energy [19], and photovoltaic system [20,21].
Only recently has competition modeling been considered [22–
25,10] since competition is not an easy tool, especially with diffusion
models that imply nonlinear estimation. At the moment, competition
between two competitors at a time has been considered in the modeling
phase.
In the energy sector, the aim of this type of study is to investigate
how the diffusion of a technology influences or competes with the
diffusion of other technologies. Vestrucci et al. [26] presented the
Italian case with competition between firewood and coal, studied in a
diffusion framework through a logistic model. Huh and Lee [27] used a
diffusion model, taking competition into account through competitors'
prices, for the description of renewables in South Korea. Duan et al.
[28] used a Lotka–Volterra model to study the evolution of wind and
PV solar technologies in leading countries. Guidolin and Guseo [29]
analyzed the transition in Germany from nuclear power to renewables
(wind and photovoltaic) through an extended Lotka–Volterra model.
In this paper, we attempt to simultaneously analyze all energy sources
by focusing on the competition between renewable and non-renewable
energy systems. In particular, we consider the total consumptions (in Mtoe)
from non-renewable and renewable energies. For non-renewable energies,
we refer to coal, oil, gas and nuclear energy (COGN) sources, while for the
Renewables, we refer to hydroelectric, wind, and solar power.
As a case study, we will consider the consumptions of the US,
Europe, China, and India, which represent four big actors in the world
energy trade. From the modeling point of view, since we are dealing
with synchronic competition, we will start from the Savin and
Terwiesch's model [23]. However, as discussed in Guseo and
Mortarino [25], we will further generalize the model in order to
estimate totally distinct cross-product competition effects; that is, both
the competition effects of the COGN sources on the Renewables and the
Renewables on COGN sources have their own parameters. In this way,
it is possible for each group of energy systems (that is, the COGN and
the Renewables) to isolate the internal contribution of its growth from
the external contribution due to the development of the competitor. For
cases where the life cycle is not uniform and the diffusion process
highlights subsequent waves, we will take this aspect into account to
improve the description and increase the forecasts' reliability. As in the
model proposed by Guseo and Mortarino [25], parameters would be
allowed to change from one wave to the subsequent wave.
We also propose six-year forecasts until 2020 with 3σ predictive
bands built under the hypothesis of heteroscedasticity [10]. We focused
on short-term forecasts since a wider window may lose meaning in the
energy sector now. These years are characterized by high levels of
uncertainty in the energy sector due to the depletion of Uranium 235
and other finite sources, the feasibility of shale technology, the
technological and economical problems of renewable energies, the
increasing demand of electricity in developing countries, and the
necessity of reducing CO2 emissions.
This paper is organized as follows. In Section 2, we briefly present
the competition models used. In Section 3, we show and discuss the
results of the model fitting to the consumptions of the US, Europe,
China, and India. In Section 4, we present concluding remarks.
2. Competition models
Competition exists if there is a market and at least two substitute
goods to be sold to potential customers. In the 1960s, the diffusion
modeling of a single product started. For this reason, the terminology
used comes from the quantitative marketing field, in which the
diffusion modeling has the highest number of applications. When a
product is launched in the market, the first customers are called
1880
Renewable and Sustainable Energy Reviews 81 (2018) 1879–1886
innovators since they were informed about the existence of the product
by an external source of information such as advertising. However, the
product actually takes off when the imitators start to buy it. The
imitators are those customers that were informed by internal source
such as word-of-mouth (WOM), describing interaction among adopters
and non-adopters. Simultaneous to the description of products' diffu-
sion, Rogers [30] introduced the issue of technology diffusion, giving
the definition of what can be considered an innovation. In particular, in
this paper, we deal with competition among technological innovations:
the customers, here, are the final consumers of energy, the sales are
represented by consumptions, the product is the energy, the COGN
(Coal, Oil, Gas, and Nuclear) sources and Renewables represent the two
competing brands in the market, the market potential is the maximum
level of consumption achievable in the energy market, and the residual
market is the energy consumption that the energy sources can still
ensure. Here we assume, for simplicity, a finite potential for both
competitors. For COGN, this is reasonable, while for the Renewables,
this choice is limited to current technologies that may be expanded in
following waves.
Savin and Terwiesch [23] propose a model that splits the inter-
personal communication effects due to WOM into (so-called) within-
brand and cross-brand effects (unbalanced models). The Unbalanced
Competition Regime Change Diachronic (UCRCD) model [25] extends
the previous model to diachronic competition (two products that enter
the market at different launch dates), allowing the parameters of the
first entrant to change at the competition's start. Since, in this paper,
the competition is synchronic, we will refer to only the competition
phase of the UCRCD model, leaving out the stand-alone part of the
model where the first competitor stays alone in the market.
Let z1 (t ) be the cumulative sales of the first product (i.e., cumulative
consumptions of COGN), z2 (t ) be the cumulative sales of the second
product (i.e., cumulative consumptions of the Renewables), and
z (t ) = z1 (t ) + z2 (t ) be the cumulative sales of the category. Let m be
the common market potential and [m − z (t )] be the common residual
market. Let z′1 (t ) be the instantaneous sales of the first product and
z′2 (t ) be the instantaneous sales of the second product. Then, the
expression of the proposed model UUC (Unrestricted Unbalanced
Competition) model becomes:
⎡ z (t ) z (t ) ⎤ ⎡ z (t ) ⎤
z′1 (t ) = m ⎢ p1 + (q1 + δ ) 1 + q1 2 ⎥ ⎢1 − ⎥
⎣ m m ⎦⎣ m ⎦
⎡ z (t ) z (t ) ⎤ ⎡ z (t ) ⎤
z′2 (t ) = m ⎢ p2 + (q2 − γ ) 1 + q2 2 ⎥ ⎢1 − ⎥
⎣ m m ⎦⎣ m ⎦
z (t ) = z1 (t ) + z2 (t ), (1)
The term “unrestricted” comes from Guseo and Mortarino [25], where
the constraint δ = γ —included both in the model by Savin and Terwiesch
[23] and in the UCRCD model—is discussed. Here, the more flexible UUC
model, where δ and γ are free, will be used. The development of the first
product is characterized by the innovation parameter, p1, the within-
brand imitation coefficient, (q1 + δ ), and the cross-brand imitation
coefficient, q1. Analogously, the development of the second product is
characterized by the innovation parameter p2 , the cross-brand imitation
coefficient, (q2 − γ ), and the within-brand imitation coefficient, q2. In this
unrestricted version, the within-brand and the cross-brand coefficients
become product-specific; that is, δ refers exclusively to the first product
and γ to the second product. In particular, δ represents the difference
between the within-brand and cross-brand coefficients for the first
product, while γ is the difference for the second product: δ and γ are
positive if the within-brand effect is stronger than the cross-brand effect,
for the corresponding product, and negative in the opposite case, as
discussed in Guseo and Mortarino [25].
In the following, we give a few more details to explain how the
model interprets the competition dynamic that exists in a market with
two products. In the competition regime, the instantaneous sales of a
product z′i , i = 1, 2 , are the sum of the sales due to innovators and the
C. Furlan, C. Mortarino Renewable and Sustainable Energy Reviews 81 (2018) 1879–1886

sales due to imitators. In the UUC model, as in the UCRCD model, z1/ m Observe that, according to the features that inspired the UCRCD
and z2 / m represent the relative diffusion levels of the first and the model, all parameters are allowed to assume different values in the two
second products in the market, respectively, at time t. That is, the new periods, a and b. If we need to add shocks that perturb the diffusions of
sales at time t of a product, due to the imitation effect, can be split into the two competitors differently, we may use the following structure
the new sales arising from WOM by adopters of the same product and (where the shocks affect the second period, b, only):
the new sales arising from WOM of the competitor's adopters. (q1 + δ )
⎧ ⎡ z (t ) z (t ) ⎤
is the within-brand WOM coefficient of the first competitor, and it z′1 (t ) = ⎨ma ⎢ pa1 + (qa1 + δa ) 1 + qa1 2 ⎥ It < t ⋆
describes the effect on new sales of the first product due to the level of ⎩ ⎣ m m ⎦
spread of the first product, which is measured by z1/ m (the market ⎡ z (t ) z (t ) ⎤ ⎫⎡ z (t ) ⎤
fraction already occupied by the first competitor at time t). q1 is the + mb ⎢ pb1 + (qb1 + δb ) 1 + qb1 2 ⎥ x1 (t ) It ≥ t ⋆⎬ ⎢1 − ⎥
⎣ m m ⎦ ⎭⎣ m ⎦
cross-brand WOM coefficient of the first competitor: it describes the
⎧ ⎡ z (t ) z (t ) ⎤
effect on new sales of the first product due to the level of spread of the z′2 (t ) = ⎨ma ⎢ pa2 + (qa2 − γa ) 1 + qa2 2 ⎥ It < t ⋆
second product, which is measured by z2 / m (the market fraction already ⎩ ⎣ m m ⎦
occupied by the second competitor at time t). Analogously, q2 is the ⎡ z (t ) z (t ) ⎤ ⎫⎡ z (t ) ⎤
+ mb ⎢ pb2 + (qb2 − γb ) 1 + qb2 2 ⎥ x2 (t ) It ≥ t ⋆⎬ ⎢1 − ⎥
within-brand WOM coefficient of the second competitor and q2 − γ is ⎣ m m ⎦ ⎭⎣ m ⎦
the cross-brand WOM coefficient of the second competitor.
m = ma It < t ⋆ + mb It ≥ t ⋆
It is possible that the adoption of a technology is accelerated or
delayed by externalities such as incentive schemes and financial crises. z (t ) = z1 (t ) + z2 (t ) It ≥ t ⋆. (6)
This effect has been modeled, for univariate growth models, through
In the case of exponential shock for both competitors, we would have
the introduction of an intervention function, x(t) [31]. For values of x(t)
x1 (t ) and x2 (t ) from Eq. (2), with parameters (a1, b1, c1) for the first
greater than 1, the adoption process is accelerated over time; other-
shock and (a2 , b 2 , c2 ) for the second one.
wise, it is delayed. Following Guseo and Dalla Valle [32], x(t) may be
To estimate the parameters, we use a nonlinear regressive frame-
expressed in the form of an exponential shock as follows:
work with dependent variables given by the observed instantaneous
x (t ) = 1 − c1 exp b1 (t − a1) I[t ≥ a1], (2) sales of the two products, si(t):
where a1 is the time of the beginning of the shock, b1 expresses how ⎡ s1 (t )⎤ ⎡ z′1 (t )⎤ ⎡ ε1 (t )⎤
rapidly the shock decays toward 0 and is usually negative, and c1 ⎢ ⎥=⎢ ⎥+⎢ ⎥,
⎣ s2 (t )⎦ ⎣ z′2 (t )⎦ ⎣ ε2 (t )⎦
indicates the intensity of the shock. Another possibility suggested by
Guseo and Dalla Valle [32] is a rectangular shock: where z′i (t ) may be defined by a UUC model, Eq. (1); a UUC model with
a shock, Eq. (4); a TW-UUC model with an exponential shock for each
x (t ) = 1 + c1 I[a1≤ t ≤ b1], (3)
competitor, Eq. (6); εi (t ) represents the residual term, i = 1, 2 . The
where c1 indicates the intensity of the shock, and a1 and b1 express the parameter set for the UUC model is (m, p1 , q2, p2 , q2, δ, γ ), for the UUC
time window when it affects diffusion. Guseo and Mortarino [24] model with a single shock is (m, p1 , q2, p2 , q2, δ, γ , a1, b1, c1), and for the
proved that it is also possible to include an intervention function x(t) in TW-UUC model with a different shock for each competitor is
a competition framework, applying it simultaneously to both compe- (ma, pa1 , qa1, δa, pa2 , qa2, γa, mb, pb1 , qb1, δb, pb2 , qb2, γb, a1, b1, c1, a2, b2, c2 ).
titors. In this paper, we include the effect of an externality in the UUC The estimates can be obtained through a nonlinear least squares
model by using a shock to affect the first competitor's diffusion only: algorithm, which does not require assumptions regarding the distribu-
tion of εi (t ).
⎡ z (t ) z (t ) ⎤ ⎡ z (t ) ⎤
z′1 (t ) = m ⎢ p1 + (q1 + δ ) 1 + q1 2 ⎥ ⎢1 − ⎥ x (t )
⎣ m m ⎦⎣ m ⎦
⎡ z (t ) z (t ) ⎤ ⎡ z (t ) ⎤ 2.1. Comparison among models
z′2 (t ) = m ⎢ p2 + (q2 − γ ) 1 + q2 2 ⎥ ⎢1 − ⎥
⎣ m m ⎦⎣ m ⎦
z (t ) = z1 (t ) + z2 (t ), (4) When alternative models are evaluated for the same application, it
is necessary to use a procedure to identify the model that performs
where x(t) is the intervention function represented by Eq. (2) or Eq. (3) better. In this work, the alternative models are represented by the UUC
and parameters (a1, b1, c1) are three further parameters to be estimated. model of Eq. (1) and the UUC model with a shock of Eq. (4), or by the
For situations when data cover a large time-span, it is unlikely that TW-UUC model of Eq. (5) and the TW-UUC model with a shock for
a single structure is adequate to describe the diffusion of the products each competitor of Eq. (6). Since, in each case, the alternative models
under study. For this reason, a multi-wave approach [33–35,16] may are nested, an F test can be used to study the significance of the
be useful to achieve a better representation of the local dynamics of an extended model [36,37,10]. Let M2 be the nested model that is
adoption process. For example, a two-wave model extending the UUC compared with the larger M1 model, n be the number of observations,
model (TW-UUC) may be written as follows. If we denote by t ⋆ the k be the number of parameters of M1, and s be the incremental number
change-point between period a and period b, we obtain: of parameters of M1 not included in M2. The F test is:
⎧ ⎡ z (t ) z (t ) ⎤ ∼2 ∼2
z′1 (t ) = ⎨ma ⎢ pa1 + (qa1 + δa ) 1 + qa1 2 ⎥ It < t ⋆ F = [R (n − k )]/[(1 − R ) s], (7)
⎩ ⎣ m m ⎦
∼2
⎡ z (t ) z (t ) ⎤ ⎫⎡ z (t ) ⎤ where R is the normalized squared multiple partial correlation
+ mb ⎢ pb1 + (qb1 + δb ) 1 + qb1 2 ⎥ It ≥ t ⋆⎬ ⎢1 − ⎥ coefficient, obtained as follows:
⎣ m m ⎦ ⎭⎣ m ⎦
⎧ ⎡ z (t ) z (t ) ⎤ ∼2
R = (RM2 1 − RM2 2 )/(1 − RM2 2 ),
z′2 (t ) = ⎨ma ⎢ pa2 + (qa2 − γa ) 1 + qa2 2 ⎥ It < t ⋆
⎩ ⎣ m m ⎦
2 2
and RM1 and RM2 denote the determination index of models M1 and M2 ,
⎡ z (t ) z (t ) ⎤ ⎫⎡ z (t ) ⎤
+ mb ⎢ pb2 + (qb2 − γb ) 1 + qb2 2 ⎥ It ≥ t ⋆⎬ ⎢1 − ⎥ respectively.
⎣ m m ⎦ ⎭⎣ m ⎦ If the stochastic component of the regression model is i.i.d.
m = ma It < t ⋆ + mb It ≥ t ⋆ according to a Gaussian distribution, the F test of Eq. (7), under the
null hypothesis, is distributed as a Snedecor's F with (s, n − k ) degrees
z (t ) = z1 (t ) + z2 (t ) It ≥ t ⋆. (5)
of freedom.

1881
C. Furlan, C. Mortarino Renewable and Sustainable Energy Reviews 81 (2018) 1879–1886

3. Case studies

In this paper, we consider the consumptions (Mtoe) of COGN

sources and Renewables, provided by [38], for the US, Europe,1 China,
and India. Our data cover the period 1965–2014, with a few excep-
tions: nuclear energy started in India in 1969 and in China in 1993;
solar energy started in the US in 1983, in Europe in 1984, in China in
1990, and in India in 2010; wind power started in Europe in 1978, in
the US in 1983, in China in 1990, and in India in 1995.
Once we aggregate consumptions (Mtoe) of the COGN sources and
Renewables, the competition figures out to be synchronic for all four
cases studied. Renewable and non-renewable technologies have very
different behaviors and are at completely different stages of their life
cycle. Solid circles in Figs. 1–4 represent the total consumption of
COGN sources, while filled squares represent the total consumption of
Renewables in the US, Europe, China, and India. Solid lines represent
the fitted values with six-year forecasts until 2020, and broken lines
represent 3σ predictive bands [10] (Appendix 3). In the sequel, each
case is analyzed in detail.

3.1. The US Fig. 1. The US. Consumptions (MToe) from COGN sources (solid circles) and
Renewables (filled squares). The solid lines correspond to a TW-UUC model, Eq. (6),
In 1965, US COGN consumption reached nearly 12500 Mtoe, while with different exponential shocks, Eq. (2), for the two competitors, while broken lines
the use of solar and wind power started in 1983, and until that time, represent 3σ predictive bands.

the Renewables were only represented by hydroelectric energy. The

pattern shown in Fig. 1 clearly suggests fitting a model with two waves. Table 1
In the time series of consumptions of COGN sources, a rapid decrease The US. Estimates, standard errors, and marginal linearized 95% confidence intervals of
the TW-UUC model, Eq. (6), with a different exponential shock, Eq. (2), for each
is visible in years 2008–2009. This period represents the years of the competitor. The rising times of the shocks, given by 1965 + a1 for COGN and by 1965 + a2
financial crisis in which oil consumption decreased and gas consump- for the Renewables, have not been estimated, but they minimize the residual deviance of
tion remained stable, interrupting—only in these years—the positive the model. R2 is provided, and the F test is given with respect to the nested TW-UUC
trend due to shale gas extraction (see source-specific data in British model without exogenous shocks, Eq. (5).
Petroleum [38]). To improve the fitting and capture the financial crisis
Par. Estimate St. error 95% CI
effect, which is exogenous to the main diffusion process, an exponential
shock was used here. ma 54374 1458 (51516, 57232)
Ultimately, for these data, the TW-UUC model of Eq. (6) was pa1 0.02273 0.00047 (0.02181, 0.02366)
qa1 −0.97725 1.66930 (−4.24901, 2.29451)
applied with different exponential shocks, Eq. (2), for the two
δa 1.09459 1.73252 (−2.30108, 4.49026)
competitors. The change-point has been fixed to t ⋆ = 1983, the year pa2 0.00080 0.00023 (0.00036, 0.00125)
when solar and wind power started to be connected to the grid. The qa2 −0.67720 1.51605 (−3.64860, 2.29420)
fitted model is represented by the solid line in Fig. 1. 3σ predictive γa −0.70541 1.57234 (−3.78715, 2.37632)
bands are represented by the broken lines. Table 1 shows the results mb 199080 19965 (159949, 238212)
pb1 0.00324 0.00030 (0.00265, 0.00382)
and the R2. The fitting is overall good (R2 = 0.99945). A TW-UUC qb1 0.74796 0.52836 (−0.28760, 1.78352)
model, without shocks, was also estimated to test whether the shocks δb −0.73324 0.54953 (−1.81030, 0.34382)
are significant. Since the value of the F test is 10.95, which is larger than pb2 0.00043 0.00029 (−0.00015, 0.00102)
the threshold of 4, the two shocks are identified as significant (Table 1). qb2 −0.16909 0.42111 (−0.99446, 0.65628)
γb −0.17533 0.43550 (−1.02889, 0.67823)
The rising time of both shocks has not been estimated simulta-
a1 44
neously with the other parameters. Since a1 and a2 represent discrete b1 0.01823 0.08268 (−0.14382, 0.18028)
time points, their values have been fixed in an adequate set of integer c1 −0.06345 0.01003 (−0.08311, −0.04378)
values and, for each configuration of a1 and a2 , the remaining a2 35
parameters have been estimated conditionally upon that choice. In b2 0.35219 0.52001 (−0.66702, 1.37140)
c2 0.00699 0.05083 (−0.09265, 0.10662)
Table 1, the configuration that minimizes the residual deviance of the
model is reported. Thus, the shock rose at year 1965 + a1 = 2009 for the R2 = 0.99945 F test =10.95
COGN time series and at year 1965 + a2 = 2000 for the Renewables.
The former shock was estimated to be negative (c1 = −0.06345) and not
yet absorbed in time (positive b1). The latter shock was estimated to be the beginning of their life cycle—in particular, their positive trend of
positive (c2=0.00699) and not yet absorbed in time (positive b2). It the last few years is due to wind rather than solar power (see source-
basically corresponds to the beginning of new specific incentives passed specific data in British Petroleum [38]).
for wind power: in the Energy Policy Act of 1992, the wind power The market potential in the second wave (mb = 199080) is almost
technology was penalized with respect to solar system, and only in four times greater than the one estimated for the first wave
1999, the incentive schemes were modified in favor of wind power [17]. (ma = 54374), which confirms that a new era for energy production
Considering that consumptions in the US have been relatively stable started in the 1980s. After the energy crisis in the 1970s, the 1980s Oil
since 2010 [39,40], the model predicts, in the next years, a trend with a Glut,2 and the oil crisis in 1984, the independence from oil was strongly
mild increase for both types of energy sources. The Renewables are at

1 2
European Union, excluding Estonia, Latvia and Lithuania prior to 1985 and Slovenia A serious surplus of crude oil caused by falling demand following the decline of
prior to 1991. consumption of the 1970s Energy Crisis.

1882
C. Furlan, C. Mortarino Renewable and Sustainable Energy Reviews 81 (2018) 1879–1886

promoted by investing in nuclear energy and renewable energies. Table 2

Moreover, part of the research was devoted to enlarge the available Within-brand and cross-brand WOM effects sustaining COGN sources and Renewables
for each model. For the US, only the second wave is considered; for this case, subscripts b
sources of oil and gas, as in the last decade with shale gas and oil.
are avoided.
The innovative effect of COGN sources decreases after t ⋆ = 1983
( pa1 = 0.02273 and pb1 = 0.00324 ); in fact, the innovative contribution of COGN The Renewables
a technology, in general, decreases over time. This effect is especially
small for the Renewables ( pa2 = 0.00080 and pb2 = 0.00043). Within Cross Within Cross
(q1 + δ ) (q1) (q2 ) (q2 − γ )
Table 2 shows the within-brand and cross-brand effects for both
types of energy sources. For the US, we reported those figures only for The US 0.01472 0.74796 −0.16909 0.00624
the second wave, which is the most interesting both for comparison Europe −0.04815 2.22408 0.94433 −0.04085
with other countries and predictions. To allow a better comparison China 0.08371 −0.34859 0.22280 −0.00625
India 0.05880 −0.05146 −0.00146 0.00323
with the model parametrization of the other cases studied, the sub-
script b of parameters is dropped in Table 2: so, only in Table 2, pb1 is
reduced to p1, pb2 to p2, δb to δ, and γb to γ.
COGN sources have both positive within-brand (qb1 + δb = 0.01472)
and cross-brand (qb1 = 0.74796) coefficients. This means that the
diffusion of COGN sources is sustained by their own spread as well
as the spread of Renewables. The knowledge about the technological
problems of using clean technologies, as the connection to grid and
electricity storage, awareness about the need for an incentive policy,
and the oil industry's lobbying pressure, may have stimulated, espe-
cially in the past, the exploitation of non-renewables for as long as
possible.
The Renewables have a negative within-brand coefficient
(qb2 = −0.16909) and a positive cross-brand coefficient
(qb2 − γb = 0.00624). The sign of the first coefficient means that the
Renewables did not have the strength to spread without the incentive
schemes that the government signed to promote their adoption, which
is captured by the positive exponential shock. The sign of the second
coefficient means that the increase of Renewables' consumption
benefits from the spread of COGN sources. The point that the
traditional sources drive the green energies to develop may be the
consequence of two different aspects: in periods when high levels of
energy are needed, attention is also focused on Renewables to sustain
growth; at the same time, the recent awareness that this country's Fig. 2. Europe. Consumptions (MToe) from COGN sources (solid circles) and
contribution is essential to achieve the required reduction of CO2 Renewables (filled squares). The solid lines correspond to a UUC model, Eq. (4), with
emissions drives energy policies to sustain Renewables' deployment. a rectangular shock, Eq. (3), for COGN consumptions, while the broken lines correspond
Nowadays, especially in the US, there is a debate about considering to 3σ predictive bands.
shale gas as an alternative to Renewables as long as it is available. In
fact, in the last 10 years, the consumption of gas experienced an As for the US, the role of innovators is negligible, especially for
incredibly positive trend due to the recent supply of shale gas. Renewables. Table 2 shows a negative within-brand (q1 + δ = −0.04815)
However, it is too early to forecast whether shale gas will delay the and a positive cross-brand effect (q1 = 2.22408) sustaining COGN
spread of renewable energy [1]. sources; that is, the diffusion level of COGN sources is reduced by
the spread of COGN sources, and it is increased by Renewables' spread.
3.2. Europe With regard to the negative sign of the within-brand coefficient, it may
be explained as the effect of the declining pattern over the last few
The consumptions in Europe have a different pattern with respect years. With regard to the cross-brand coefficient, the knowledge about
to the other case studies. Fig. 2 shows that consumptions of COGN problems related to the adoption of clean technologies might have
sources have consistently fallen down since 2008. So far, from data, supported, especially in the past, the exploitation of non-renewable
there are not signals of a possible new rise if the economic and/or sources.
technological conditions remain as they are now. The effects of the The Renewables have a positive within-brand effect (q2 = 0.94433)
crisis, which started in 2008, are still visible; in fact, even if we sum the and a negative cross-brand effect (q2 − δ = −0.04085). This means that
consumptions of both competitors, we still obtain a declining pattern in Renewables' consumption is supported by their own diffusion and that
the last few years, confirming that COGN consumption is decreasing it is negatively influenced by the spread of COGN sources. Renewables
both due to the Renewables' competition and a general decrease in have a bigger tradition in Europe with respect to the US and China and
energy demand. have recently grown due to investments in wind power (see source-
We fitted a UUC model, Eq. (4), with a rectangular shock, Eq. (3), specific data in British Petroleum [38]). In the last 10–15 years, Europe
affecting COGN only. Results are presented in Fig. 2 and Table 3. The has signed good incentive schemes for producing clean energies,
model gives sensible predictions for both competitors. As in the previous especially in some countries. The reasons are several: among them,
case, the time window of the shock has been estimated through a the intention to stop nuclear energy in some countries [16,29] and the
discrete optimization algorithm, and the remaining parameters have awareness that many countries in Europe completely depend on
been estimated conditionally on that choice. Thus, the shock occurs unstable foreign countries for gas and oil. With regard to the cross-
between 1965 + a1 = 1980 and 1965 + b1 = 2002 , and its impact is brand coefficient, two conditions may have contributed to obtaining a
estimated to be negative (c1 = −0.10923). The shock is significant since negative coefficient: the last economical crisis and the fact that
the F test, which compares this model with a UUC model without shocks, traditional energy sources have delayed the spread of clean energies.
assumes a huge value (F=159) if compared with the threshold of 4.

1883
C. Furlan, C. Mortarino Renewable and Sustainable Energy Reviews 81 (2018) 1879–1886

Table 3
Europe. Estimates, standard errors, and marginal linearized 95% confidence intervals of
the UUC model, Eq. (4), with a rectangular shock, Eq. (3), for COGN consumption. The
time window of the shock, given by 1965 + a1 and 1965 + b1, has not been estimated, but
it minimizes the residual deviance of the model. R2 is provided, and the F test is given
with respect to the nested UUC model, Eq. (1).

Par. Estimate St. error 95% CI

m 102835 1936 (98988, 106682)

p1 0.00906 0.00009 (0.00888, 0.00925)
q1 2.22408 0.56296 (1.10567, 3.34249)
p2 0.00031 0.00009 (0.00013, 0.00049)
q2 0.94433 0.16982 (0.60695, 1.28171)
δ −2.27223 0.58737 (−3.43913, −1.10532)
γ 0.98518 0.17771 (0.63212, 1.33825)
a1 15
b1 37
c1 −0.10923 0.00770 (−0.12454, −0.09393)

R2 = 0.99849 F test =159

3.3. China

The generated electricity production in China is growing at a rapid Fig. 3. China. Consumptions (MToe) from COGN sources (solid circles) and Renewables
(filled squares). The solid lines correspond to a UUC model, Eq. (4), with a rectangular
speed [39], but the demand cannot only be met by fossil fuels. For the
shock, Eq. (3), for COGN consumptions, while broken lines correspond to the 3σ
need of reducing CO2 emissions, given the abundance of renewable predictive bands.
sources, the Chinese government has passed several policies since 2005
to strongly support the growth of renewables [39]. These policies were Table 4
also intended to sustain local industries (e.g., photovoltaic cell manu- China. Estimates, standard errors, and marginal linearized 95% confidence intervals of
facturers). the UUC model, Eq. (4), with a rectangular shock, Eq. (3), for COGN consumptions. The
Data in Fig. 3 show a steeper growth of the consumptions after time window of the shock, given by 1965 + a1 and 1965 + b1, has not been estimated, but
it minimizes the residual deviance of the model. R2 is provided, and the F test is given
2000 and a short slow-down just before it. We thought to model this
with respect to the nested UUC model, Eq. (1).
pattern using a UUC model with a rectangular shock only in the time
series of COGN consumptions. As explained in Section 2, this model Par. Estimate St. error 95% CI
corresponds to Eq. (4) with a rectangular shock, Eq. (3), in the time
m 5.67 × 1014 1.33 × 10−11 (5.67 × 1014, 5.67 × 1014)
series of COGN consumptions. Results of the fitted model are
p1 2.09 × 10−13 2.16 × 10−14 (1.66 × 10−13, 2.52 × 10−13)
presented in Table 4. As before, the time window of the shock has
q1 −0.34859 0.09648 (−0.54026, −0.15692)
been obtained through a discrete optimization. Thus, the shock is p2 1.42 × 10−14 2.13 × 10−14 (−2.81 × 10−14, 5.65 × 10−14)
positioned between 1965 + a1 = 1995 and 1965 + b1 = 2003, and its q2 0.22280 0.08342 (0.05707, 0.388538)
impact is estimated to be negative (c1 = −0.20308). The shock is δ 0.43230 0.10159 (0.23048, 0.63412)
significant since the F test, comparing this model with a UUC model γ 0.22905 0.08779 (0.05463, 0.40347)
without shock, assumes a huge value (F = 160.23). a1 30
b1 38
The estimate of the market potential is very high since data show a
c1 −0.20308 0.01499 (−0.23287, −0.17329)
rapid growth. However, it is known that, for Bass-family models, this
parameter is not fully reliable when the process is far from peaking, as R2 = 0.99583 F test =160.23
in this case. For the same reason, the marginal linearized 95% interval
confidences of p2 show a kind of instability very close to zero, which
means that, for China, the innovative effect of Renewables is negligible. Renewables is also found in the analysis of Lin et al. [42] and
The within-brand and cross-brand WOM coefficients are shown in confirmed by other papers included in the review they propose. Lin
Table 2. The within-brand coefficients (q1 + δ and q2) are positive for et al. [42], in particular, say that “the increase in the share of fossil fuel
both competitors, which means that the diffusion of each competitor in energy consumption leads to a … reduction in renewable energy
has a positive effect on its spread. For Renewables, the government has consumption in China.” In fact, besides the efforts for pushing
played a significant role by passing an endless stream of incentives to Renewables, it is still easier to invest in conventional sources in terms
achieve a remarkably rapid growth of renewables [41]. In fact, the of efficiency, abundance of energy generation, and economic competi-
speed of growth has been so high that it has even led to problems such tiveness.
as overcapacity and lack of funds [39].
The cross-brand coefficients (q1 and q2 − γ ) of both competitors are 3.4. India
negative. With respect to COGN sources, this means that the spread of
Renewables has a negative effect on the consumption of COGN sources. In India, 62.1% of the power generation still comes from coal
To meet the growing energy demand and the required reduction of CO2 thermal power, while 27.4% comes from renewables [43]. As of March
emissions, the government is strongly supporting the development of 31, 2014, 21318 villages (3.6%) in remote areas did not have electricity
green energies to support the economy. However, producing energy yet, which makes renewable energy a suitable and convenient choice
from Renewables is not free of complications: the different develop- [44]. Moreover, India is the fourth largest CO2 emitter in the world, and
ment plans, signed by the government to promote Renewables, should its rapid economic growth requires both an increase of the energy power
be reinforced to solve technological problems such as the connection to generation and a reduction of the environmental impact [45]. In order to
wind and solar power grids [41]. meet India's goal of reducing the CO2 emissions (from 2005 levels) by
With respect to Renewables, the negative association between the 20–25% by 2020, increasing the share of renewables in the energy mix
share of fossil fuels in total energy consumption and the adoption of has been promoted by the National Action Plan of 2008 [45].

1884
C. Furlan, C. Mortarino
To India data, we applied the UUC model, Eq. (1). Results are
shown in Fig. 4 and Table 5. The life cycle of both competitors is far
from reaching maturity in this country. As a consequence, the estimate
for m is quite unstable, and uncertainty also affects the evolutionary
parameters.
The innovative components of both competitors, p1 and p2 , are
found to be negligible for this country too. The within-brand effect
(q1 + δ ) of COGN is positive, as for the US and China, but the within-
brand effect q2 for Renewables comes out to be slightly negative, which
means that the spread of Renewables is not positively associated to
their consumption. In this country, the demand continues to exceed the
supply, despite the efforts aimed at increasing the energy generation
capacity [44]. This is counter-productive, for example, for the produc-
tion of energy from the installed pumped storage stations of the large
hydro. In fact, the lack of an energy surplus affects the exploitation of
the large hydro, which represents 14.41% of Indian power generation.
A direct consequence is that the pumped storage stations of the large
hydro, which have been more intensively committed since 2001, are
operating in the pumping mode for fewer hours than contemplated;
thus, most of the pumped storage stations achieved 40–50% of the
projected design energy [46]. Moreover, India has a more limited
capacity of hydropower than China [47] and does not invest in
Renewables as much as China does. If India wants to compete with
China, the Indian government should increase the funds dedicated to
promote renewables [47].
The cross-brand effect (q2 − γ ) of COGN sources on Renewables is
positive, as for the US case, which indicates that the spread of COGN
has a positive effect on the consumption from Renewables. The cross-
brand effect q1 of Renewables on COGN consumption is negative,
which indicates that the spread of Renewables has a negative effect on
the consumption of COGN. Ultimately, in this country, Renewables
have a negative impact both on themselves and on COGN sources,
while COGN sources have a positive impact both on themselves and on
Renewables. Renewables still do not have enough advantages to
develop by themselves but are driven by the necessity of reducing the
share of COGN sources. On the other hand, COGN sources drive
themselves to spread, but this growth is partially limited by the
necessity of investing in renewables.
Fig. 4. India. Consumptions (MToe) from COGN sources (solid circles) and Renewables
(filled squares). The solid lines correspond to a UUC model, Eq. (1), while the broken
lines correspond to the 3σ predictive bands.
1885
Renewable and Sustainable Energy Reviews 81 (2018) 1879–1886
Table 5
India. Estimates, standard errors, and marginal linearized 95% confidence intervals of the
UUC model, Eq. (1). R2 is provided.
Par. Estimate St. error 95% CI
m 2.12 × 1013 1.03 × 1019 (−2.01 × 1019, 2.01 × 1019)
p1 2.06 × 10−12 9.99 × 10−7 (−1.96 × 106, 1.96 × 106)
q1 −0.05146 0.03398 (−0.11807, 0.01515)
p2 3.19 × 10−13 1.54 × 10−7 (−3.03 × 10−7, 3.03 × 10−7)
q2 −0.00146 0.03398 (−0.06806, 0.06513)
δ 0.11026 0.03666 (0.03840, 0.18212)
γ −0.00469 0.03665 (−0.07653, 0.06715)
R2 = 0.99829
4. Conclusions
In this study, the competition between traditional energies (COGN)
and Renewables is simultaneously modeled, in terms of consumptions
(Mtoe), in a diffusion framework. Four big actors in the world energy
market are considered: the US, Europe, China, and India. The two
types of energy systems (i.e., the competitors) are at different stages of
their life cycle, and the competition dynamics are also different among
the cases studied. In fact, the four areas analyzed show very different
WOM patterns. None of them, however, exhibit a completely coopera-
tive behavior between COGN sources and Renewables, which turn out
to actually compete. This result suggests that separate forecasts for
consumptions of different sources could be partially unreliable because
this interaction is neglected. The different competition structures
strongly depend upon the maturity levels of diffusion, which is very
different between emerging countries (China and India) and more
stable economies (the US and Europe). In addition, the specific policies
adopted within each country or region had a different impact on
consumptions' evolution.
The impact of the 2008–2009 economical crisis is apparent in
Europe, where the consumptions have reduced over the last few years,
especially of COGN sources. This effect is also visible in US data, where
it has been captured as a negative externality. COGN consumptions
also reduced in the US during the crisis but are now relatively stable—
unlike in Europe. The reduction of COGN consumptions in Europe is
also due to a partial substitution of traditional energies with clean
energies; in this area, renewables have a longer tradition than in other
analyzed countries, and good incentive schemes were passed in the past
to promote this transition.
The predictions until 2020 show, in the US, a mildly positive trend
for both COGN and Renewables consumptions and, in Europe, a
declining pattern for COGN and a moderate increase for Renewables
consumption. In China, consumptions of both energy types are
predicted to steeply increase. In India, only COGN consumptions are
predicted to considerably increase, and Renewables consumption will
moderately increase. The different situation between China and India
about Renewables is apparent: India allocated a lower amount of funds
to promote clean energies, and at the same time, it has a more limited
capacity of renewable energy systems than China (e.g., for hydroelectric
power). To this end, India would benefit from developing Renewables
to create surplus energy to make the pumped storage stations work and
to provide electricity to remote areas.
Acknowledgements
This work was supported by the Energy Research Centre Giorgio
Levi Cases, University of Padua, Italy; Grant 2014 “Innovation
Diffusion Processes: Competition and Substitution in Energy
Technologies.”
C. Furlan, C. Mortarino
References
[1] Klein Sharon JW, Whalley Stephanie. Comparing the sustainability of U.S.
electricity options through multi-criteria decision analysis. Energy Policy
2015;79:127–49.
[2] Melikoglu M. Shale gas:analysis of its role in the global energy market. Renew
Sustain Energy Rev 2014;37(9):460–8.
[3] Wang Qiang, Chen Xi, Jha Awadhesh N, Rogers Howard. Natural gas from shale
formation–the evolution, evidences and challenges of shale gas revolution in United
States. Renew Sustain Energy Rev 2014;30(2):1–28.
[4] Vengosh Avner, Jackson Robert B, Warner Nathaniel, Darrah Thomas H, Kondash
Andrew. A critical review of the risks to water resources from unconventional shale
gas development and hydraulic fracturing in the Unites States. Environ Sci Technol
2014;48:8334–48.
[5] Centner Terence J, O'Connell Laura Kathryn. Unfinished business in the regulation
of shale gas production in the United States. Sci Total Environ 2014;476–477(4/
1):359–67.
[6] Xin-gang Zhao, Ya-hui Yang. The current situation of shale gas in Sichuan, China.
Renew Sustain Energy Rev 2015;50:653–64.
[7] Garg P. Energy scenario and vision 2020 in India. J Sustain Energy Environ
2012;3:7–17.
[8] Shale Gas International. India announces reforms to oil and gas industry. Available
〈http://www.shalegas.international/2015/11/22/india-announces-reforms-to-oil-
and-gas-industry/〉[accessed 27 November 2015], 2015.
[9] Today in Energy. EIA projects world energy consumption will increase 56% by
2040. Vol. July. Available 〈https://www.eia.gov/todayinenergy/detail.cfm?
Id=12251〉[accessed 12 April 2016], 2013.
[10] Guseo R, Mortarino C. Modeling competition between two pharmaceutical drugs
using innovation diffusion models. Ann Appl Stat 2015;9(4):2073–89.
[11] Guseo R, Dalla Valle A, Guidolin M. World oil depletion models: price effects
compared with strategic or technological interventions. Technol Forecast Social
Change 2007;74(4):452–69.
[12] Guseo R, Mortarino M, Abud Darda Md. Homogeneous and heterogeneous
diffusion models: algerian natural gas production. Technol Forecast Social Change
2015;90(B):366–78.
[13] Abud Darda Md, Guseo R, Mortarino C. Nonlinear production path and an
alternative reserves estimate for South Asian natural gas. Renew Sustain Energy
Rev 2015;47:654–64.
[14] Guidolin Mariangela, Guseo Renato. A nuclear power renaissance?. Technol
Forecast Social Change 2012;79(9):1746–60.
[15] Dalla Valle A, Furlan C. Diffusion of nuclear energy in some developing countries.
Technol Forecast Social Change 2014;81:143–53.
[16] Furlan Claudia, Guidolin Mariangela, Guseo Renato. Has the Fukushima accident
influenced short-term consumption in the evolution of nuclear energy? An analysis
of the world and seven leading countries. Technol Forecast Social Change
2016;107:37–49.
[17] Dalla Valle A, Furlan C. Forecasting accuracy of wind power technology diffusion
models across countries. Int J Forecast 2011;27(2):592–601.
[18] Riddhi Panse, Kathuria Vinish. Modelling diffusion of wind power across countries.
Int J Innov Manag 2015;19(4).
[19] Toka Agorasti, Iakovou Eleftherios, Vlachos Dimitrios, Tsolakis Naoum,
Grigoriadou Anastasia-Loukia. Managing the diffusion of biomass in the residential
energy sector: an illustrative real-world case study. Appl Energy 2014;129:56–69.
[20] Guidolin Mariangela, Mortarino Cinzia. Cross-country diffusion of photovoltaic
systems: modelling choices and forecasts for national adoption patterns. Technol
Forecast Social Change 2010;77(2):279–96.
[21] Palmer J, Sorda G, Madlener R. Modeling the diffusion of residential photovoltaic
systems in Italy:an agent-based simulation. Technol Forecast Social Change
2015;99:106–31.
1886
Renewable and Sustainable Energy Reviews 81 (2018) 1879–1886
[22] Krishnan Trichy V, Bass Frank M, Kumar V. Impact of a late entrant on the
diffusion of a new product/service. J Mark Res 2000;37(2):269–78.
[23] Savin S, Terwiesch C. Optimal product launch times in a duopoly: balancing life-
cycle revenues with product cost. Oper Res 2005;53(1):26–47.
[24] Renato Guseo, Mortarino Cinzia. Sequential market entries and competition
modelling in multi-innovation diffusions. Eur J Oper Res 2012;216(3):658–67.
[25] Guseo Renato, Mortarino Cinzia. Within-brand and cross-brand word-of-mouth for
sequential multi-innovation diffusions. IMA J Manag Math 2014;25(3):287–311.
[26] Vestrucci Paolo, Schiavi Serena, Orlandelli Carlo Maria. Long term dynamics of
energy systems: the Italian case. Technol Forecast Social Change 2015;96:266–76.
[27] Huh Sung-Yoon, Lee Chul-Yong. Diffusion of renewable energy technologies in
South Korea on incorporating their competitive interrelationships. Energy Policy
2014;69:248–57.
[28] Duan Hong-Bo, Zhu Lei, Fan Ying. A cross-country study on the relationship
between diffusion of wind and photovoltaic solar technology. Technol Forecast
Social Change 2014;83:156–69.
[29] Guidolin Mariangela, Guseo Renato. The German energy transition: modeling
competition and substitution between nuclear power and renewable energy
technologies. Renew Sustain Energy Rev 2016;60:1498–504.
[30] Rogers Everett M. Diffusion of innovations. New York: The Free Press; 1962.
[31] Bass Frank M, Krishnan Trichy V, Jain Dipak C. Why the Bass model fits without
decision variables. Mark Sci 1994;13(3):203–23.
[32] Guseo R, Dalla Valle A. Oil and gas depletion: diffusion models and forecasting
under strategic intervention. Stat Methods Appl 2005;14(3):375–87.
[33] Mohr SH, Evans GM. Mathematical model forecasts year conventional oil will peak.
Oil Gas J 2007;105(17):45–6.
[34] Guseo R. Worldwide cheap and heavy oil productions: a long-term energy model.
Energy Policy 2011;39(9):5572–7.
[35] Guseo R, Guidolin M. Heterogeneity in diffusion of innovations modelling: a few
fundamental types. Technol Forecast Social Change 2015;90:514–24.
[36] Seber GAF. The linear hypothesis: a general theory, 2nd ed.. London: Griffin; 1980.
[37] Seber GAF, Wild CJ. Nonlinear regression. New York: Wiley; 1989.
[38] Petroleum British. BP statistical review of world energy 2015. Available 〈http://
www.bp.com〉[accessed 29 December 2015], 2015.
[39] Jianfei Shen, Luo Chen. Overall review of renewable energy subsidy policies in
China–contradictions of intentions and effects. Renew Sustain Energy Rev
2015;41:1478–88.
[40] Today in Energy. Future world energy demand driven by trends in developing
countries. Vol. December. Available 〈https://www.eia.gov/todayinenergy/detail.
cfm?Id=14011〉[accessed 23 April 2016], 2013.
[41] Zhao Zhen-Yu, Chen Yu-Long, Chang Rui-Dong. How to stimulate renewable
energy power generation effectively?–China's incentive approaches and lessons.
Renew Energy 2016;92:147–56.
[42] Lin Boqiang, Omoju Oluwasola E, Okonkwo Jennifer U. Factors influencing
renewable electricity consumption in China. Renew Sustain Energy Rev
2016;55:687–96.
[43] Indian Electricity Scenario. Ministry of Power, Govt. of India. Available 〈http://
powermin.nic.in/power-sector-glance-all-india〉[accessed 18 April 2016], 2016.
[44] Tripathi Lata, Mishra AK, Dubey Anil Kumar, Tripathi CB, Baredar Prashant.
Renewable energy: an overview on its contribution in current energy scenario of
India. Renew Sustain Energy Rev 2016;60:226–33.
[45] Chandel SS, Shrivastva Rajnish, Sharma Vikrant, Ramasamy P. Overview of the
initiatives in renewable energy sector under the national action plan on climate
change in India. Renew Sustain Energy Rev 2016;54:866–73.
[46] Swain M. Pumped Storage. Hydro Power Plant. Electrical India, November:127–
136, 2013.
[47] Ahmed Saeed, Mahmood Anzar, Hasan Ahmad, Ahmad Guftaar, Sidhu Sardar,
Uddin Butt Muhammad Fasih. A comparative review of China, India and Pakistan
renewable energy sectors and sharing opportunities. Renew Sustain Energy Rev
2016;57:216–25.