Transmission System Planning Considering

Solar Distributed Generation Penetration

Phillipe Vilaa Gomes1, 2, Joo Tom Saraiva1, 2
1
Faculdade de Engenharia da Universidade do Porto, Porto, Portugal.
2
INESC TEC, Porto, Portugal.
email: phillipe.gomes@fe.up.pt, jsaraiva@fe.up.pt

Abstract In recent years, power systems have been watching commissioning date to attend a pre fixed objective (investment
important advancements related with Plug-in-Electrical cost, operational cost, GHG emissions, reliability, etc.).
Vehicles (PEVs), Demand Side Management (DSM), Distributed
Generation (DG), Microgrid and Smart Grid installations that The TEP problem has non-linear and non-convex nature
directly affect distribution networks while impacting indirectly which leads to a huge computational effort. In order to
on Transmission studies. These changes will lead to an extra overcome this burden, relaxed models are often used as static
flexibility on the transmission-distribution boundary and to a approaches or formulations based on the DC power flow.
significant modification of the load patterns, that are an Although their solutions are considered reasonable in the
essential input to planning studies. In this scope, this paper literature, these models dont incorporate a holistic view over
describes a multiyear Transmission Expansion Planning (TEP) the entire planning problem, the reactive power, the branch
solved by Evolutionary Particle Swarm Optimization (EPSO) losses and the voltage limits on the bars are often disregarded,
and incorporating the impact of solar DG penetration. The and so its computational effort is lighter. It is then clear that
primary substation load profiles and the solar generation these models dont represent the real world problem so that
profiles are taken into account on the planning problem. The less realistic solutions can be obtained. In alternative, the
numerical simulations were conducted using the IEEE 24 bus original problem should be considered eventually using
reliability test system in which the planning horizon is 3 years bioinspired metaheuristics to solve it (instead of traditional
and the load growth is 2.5 % per year. If demand and solar DG mathematical optimization techniques). This kind of tools are
peaks are coincident, then the liquid demand seen by the
able to give optimal or suboptimal solutions taking advantage
transmission network gets reduced enabling a reduction of
investment costs. In the tested cases, these peaks were not
of patterns recognized in the nature as behaviors of fireflies,
coincident so that the optimal expansion plan remains bats, ants and swarms for instance.
unchanged even though the injected power from DG is large. In order to solve the TEP problem, and apart from other
This stresses the fact that solar DG may not on an isolated way considerations regarding the evolution over time of other input
contribute to alleviate the demand seen by transmission variables (generation, equipments, etc.), we usually consider a
networks but should be associated with storage devices or scenario for the system demand for each year of the planning
demand side management programs.
horizon. For each of these years and for the transmission
Index Terms Multiyear Transmission Expansion Planning,
assets available in that year an optimal power flow analysis
AC Optimal Power Flow, Solar DG, Evolutionary PSO. should be done in order to check if the system operates
properly, i.e., if there is no congestion on the lines, if Power
I. INTRODUCTION Not Supplied (PNS) is zero and if the voltage profile is
adequate. These demand scenarios can then be addressed
The purpose of a TEP problem is to determine how a using probabilistic or deterministic approaches or considering
transmission system should evolve over time in the most uncertainties representing vague information provided by the
economical way according to the planner's main drivers planner for instance under the form of fuzzy concepts.
which, in the case, can correspond to the load growth, the Regarding deterministic approaches, it is typically considered
connection of new generation sources or new demand centers, the forecasted annual peak demand as the demand scenario
the equipment aging, the improvement of competition between that the system must supply in a secure and reliable way.
generation companies, changes in export/import patterns
between neighbor systems and variations in the supply The advent of changes namely in distribution networks
reliability requirements of customers [1]. Therefore, using a related with the increasing presence of DG, the development
pre-defined list of candidate equipments (transmission lines, of micro and smart grids and the increasing number of PEVs
cables, transformers, etc.) that can be inserted on the grid, the must also be incorporated in long-term expansion planning
TEP problem aims at identifying the ones to be built and their methodologies in view of their impact on the network
depending on regional incentive policies. In this paper, we

978-1-5090-5499-2/17/$31.00 2017 IEEE

considered a deterministic load scenario taking into account an These incentives are justified by the potential benefits that the
increased penetration of solar DG on the IEEE 24 bus RTS in deployment of these distributed sources can bring to entire
order to investigate how the deployment of this type of DG system and to the society as a whole. According to [3] these
impacts the definition of the transmission expansion plans. advantages are related with the possible postponement of
The investigation of this impact can therefore be considered as expansion investments in distribution and transmission
the main contribution of this paper. systems, the reduction of the environmental impact, the
reduction of the network loading levels, the reduction of losses
Regarding the structure of the paper, following this and the diversification of energy sources.
Introduction, Section II reviews some concepts related with
distributed generation, Sections III presents the mathematical Although the solar DG penetration is still low in most
formulation of the dynamic TEP problem and Section IV countries, the global incident solar radiation can reach large
details the main blocks of the EPSO algorithm. Finally, values ranging from 900 to 1250 kWh/m throughout the year
Section V addresses the methodology used in this paper. in Germany, 900-1650kWh/m in France, 1200-1850 kWh/m
Section VI presents the simulation results and Section VII in Spain and 4200-6700 kWh/m in Brazil. Therefore, several
draws the main conclusions of this research. studies investigate the barriers to DG penetration [4] as well as
its impact on the grids [5].
II. DISTRIBUTED GENERATION
Distributed generation is usually as associated to small In TEP problems DG is usually modeled as a negative real
generation units connected to the power grid either on the load, that is, the DG reduces locally the real load on a primary
customer side or to the distribution network [2]. The size of a substation. Fig. 1 shows this behavior for a bus over one day
typical DG unit ranges from 1 kW to 10 or 20 MW but in (24 hours) on the test system used in this paper. The new
some cases, as for large wind parks or solar PV stations, it can demand that will be used to carry out the expansion planning
reach 100 or 200 MW. Besides, regarding to its primary studies takes into account the solar DG penetration. In this
source, DG can be classified as renewable (solar, wind, hydro) example, solar generation starts at 6 am and ends at 6 pm and
or nonrenewable (internal combustion engine) energy stations. has a generation peak at 12:00. It should be stressed that the
expansion study should continue to be done in a way to ensure
The exploitation of energy generation from renewable the safe operation during the peak demand (19:00), and this
sources has been increasing in several countries, in most cases drive is not changed while considering the injection of
induced by regulatory incentives for small distributed electricity from PV sources on the grid nodes.
generation such as Feed-in Tariffs and Net Metering schemes.

Figure 1. Illustration of the impact of PV generation of the demand profile along one day.

III. TEP MATHEMATICAL FORMULATION of the horizon. Furthermore, the formulation includes
The mathematical formulation adopted in this paper for the physical constraints related with the branches and generator
TEP problem takes into account the investment cost ( Cinv , p ) capacity limits, financial limitations related with the capital
available in each year or along the entire planning horizon
and a penalization term ( ) for Power Not Supplied (PNS). and quality of supply constraints associated to reliability
This means that one possible expansion plan is characterized indices. According to these ideas, the associated TEP
by the investment cost associated to the insertion of new problem can be formulated by (1) to (4).
equipments on the grid on specific years and by the expected
np
PNS in each year of the horizon when the annual peak load is
applied on the system. A solution is considered feasible when
Minimize
p =1
p .C inv , p + .PNS (1)
it is able to ensure a safe operation of the system over the
Subject to:
planning horizon, that is, it displays zero PNS in every year
 Physical Constraints (2) IV. EPSO ALGORITHM
Financial Constraints (3) EPSO is a powerful tool that combines concepts of
Quality of service Constraints (4) evolutionary computation and multi agent population taking
advantage of the standard blocks that are typical in Genetic
In this formulation p is the present-worth value coefficient
Algorithm and in Particle Swarm Optimization. This tool is
given by (5), d is the discount rate and p is the index able to combine the best features of these two groups of
associated to each period in the planning horizon. techniques and therefore has excellent performance in solving
complex problems such as the one addressed in this paper.
1
p = (5)
(1 + d ) p This algorithm is based on the evolution of a set of
particles, each of them representing solutions for the problem.
Although the AC power flow (AC-OPF) requires a larger Along the evolution process the particles evolve according to
computational burden, this is the most adequate model to deal a fitness function and continue to improve in each iteration
with TEP problems because it considers the reactive power, until the process reaches a pre-established stopping criterium,
the losses and the bus voltage limits. Therefore, we used an and the best solution of the last population is provided to the
AC-OPF based model in order to compute the PNS when the user. Fig. 2 details the main blocks of the EPSO algorithm that
system supplies the annual peak demand for each planning will also be described in the next paragraphs.
year and incorporating the new equipments included in the
expansion plan. Given the fact that the AC-OPF problem has
to be solved a large number of times, associated to the years in
the planning horizon and with the trial expansion plans to test,
we used the MATPOWER tool [6] to assist the solution of the
mentioned AC-OPF problem.
The AC-OPF used in this paper is formulated by (6) to (14):
For a given trial expansion plan being tested, this problem is
solved for each year in the planning horizon considering in
each year the new equipments that are included in that
solution until the year under analysis.
Min PNS (6)
subject to P(V , , n) PG + PD = 0 (7)
Q(V , , n) QG + QD = 0 (8)
PG min PG PG max (9) Figure 2. Main blocks of the EPSO Algorithm.

QG min QG QG max (10) The initial population is created randomly and the
Vmin V Vmax (11) developed approach uses a Tabu List to prevent repeating the
o o
same particle in the population, that is, to increase the
( N + N ) S from ( N + N ) S max (12) diversity of the particles in the initial population. In the
o o replication block each population is cloned r times in order to
( N + N ) S to ( N + N ) S max (13)
create new populations that will be mutated in the next block.
0 n nmax (14) In the mutation block the weights and the best particle found
In this formulation, the objective function (6) corresponds until now (gbest) for all the populations are mutated using (15)
to the minimization of PNS, PG and QG are the active and and (16) in which the symbol * denotes the mutation operator.
reactive power generation, PD and QD are the active and This process increases the diversity of the individuals under
analysis.
reactive power demand, V is the voltage magnitude, Sij from and
*
o 1
to are the apparent flows in branch ij, N and N are wijit +1 = 0,5 + rand () (15)
Sij *
1 + exp( wijit )
diagonal matrixes containing the inserted equipments and the
base topology equipments. * *
(16)
gbest = gbest + round (2. wiit4+1 1)
The solution having the lowest value for Eq. (1) is Therefore, new populations (offsprings) are created in the
considered the optimal solution and will be analyzed
recombination block based on the PSO movement rule.
considering issues as the operation costs, reliability, losses and
environmental impact. According to (17) the position of a particle i in iteration it+1 is
the result of its position in iteration it plus the velocity vector
given by (18). This procedure is repeated for all particles in profile and these values can be assessed in [9] taking into
the cloned populations. account the Monday of the 2nd week for the Winter, the
Saturday of the 13th week for the Spring, the Friday of the
xiit +1 = xiit + viit +1 (17) 24th week for the Summer and the Sunday of the 41st week
* * * for Autumn. The methodology discussed in this paper is
viit +1 = w
it +1
i1
it +1
.viit + w
i2 .( pbesti xiit ) + w it +1
i3 .( gbest xiit ).P (18) illustrated in Fig. 3.

The first term in (18) represents the inertia of the particle,

the second term represents its individual knowledge and the
last term represents the collective knowledge of the swarm. P
is the communication factor described in [7]. It typically takes
values 0 or 1 so that if 0 is used for a position of the particle
vector then the collective knowledge is not passed to this
particle in the next iteration. The evaluation block checks if a
candidate plan ensures a safe operation (without PNS) and so
its load shedding cost is zero. Otherwise the PNS is
multiplied by a penalty factor (in $/MW). Once this step is
finished, all the particles are characterized by their investment
and expected load shedding costs. In the selection block a
tournament selection is performed to build the new
population having the same size of the initial one. The
iterative process continues until the best solution remains
unchanged along a pre-defined number of iterations. In the
selection block it was also used a Tabu List in order to ensure Figure 3. Methodology for DG impact on TEP
the diversity of the new population. This list checks if one
As an example, Fig. 4 represents the new load profile for
solution is already in the population. If so it modifies this
solution selecting randomly one investment in this solution each season of the primary substation connected to bus 1
and postponing it by moving it one year ahead. assuming a 10% DG penetration (as suggested by Fig. 1). In
this paper, the maximum solar generation (solar peak at 12:00)
TEP problems have integer nature which leads to the for each bar is assumed to be proportional to its annual peak
phenomenon of combinatorial explosion of investment load. As an example, Fig. 5 presents the DG generation in
alternative plans. This characteristic typically leads to a high MW in each bus for the three penetration levels that were
computational effort to identify good quality plans. In order to considered, 10%, 15% and 20%. As a whole, the 10% level of
tackle this problem, the described EPSO algorithm was penetration is associated to an installed capacity of 570 MW,
implemented using parallel computing. 15% is associated to 855 MW and 20% to 1140 MW for the
V. METHODOLOGY annual peak demand of 5700 MW considered in the first year.
In this paper the penetration of solar DG was organized in 4 VI. RESULTS
scenarios: 0%, 10%, 15% and 20% of the annual peak demand This section presents the results obtained using the proposed
connected to each bus. After the optimization process carried TEP approach when applied to a modified version of the IEEE
out by the EPSO algorithm is done (considering the new peak RTS 24 bus system [9]. The system that was used has some
demand for each scenario), the DG impact is analyzed differences regarding the original one, as described below:
considering four items: reliability, economic aspects, losses
and environmental aspects. The probabilistic index Expected i. the loads are modeled as negative real power
Energy Not Supplied (EENS) is used to measure the system injections with associated negative costs as described
reliability. EENS reflects not only the base topology and in [10];
physical limits of the system under analysis but it also ii. the values of all the loads were duplicated and the
incorporates uncertainties associated with the non-ideal installed capacity of all generators were tripled (real
behavior of system components. In order to estimate the and reactive) in order to turn the transmission network
EENS it is not enough to sample components out of service. In more stressed;
fact it is necessary to sample operation and repair times of the iii. the modified system has 9 hydro generators that have
components in order to access the duration of each state which initial maximum capacity of 2880 MW (28,2% of the
justifies using a chronological Monte Carlo as detailed in [8]. total capacity of the system) as follows:
- 6 hydro generators of 150 MW in bus 22;
Regarding the economic analysis, four load blocks - 1 hydro generator of 1050 MW in bus 23;
representing one day of each season of the year are used in - 1 hydro generator of 465 MW in bus 15;
order to estimate the operation costs. Each primary substation - 1 hydro generator of 465 MW in bus 16;
(represented by the buses of the system) have their own load
 Figure 4. New load profile for the bus 1 (with 10% DG penetration).

Figure 5. DG penetration by bus and scenario at 12 am (Peak of solar generation).

The initial peak demand is 5700 MW and the total penetration of 0%, 10%, 15% and 20%) the annual peak
generation capacity is 10215 MW. Regarding the CO2 demand is the same which means that the optimal expansion
emissions, we use data from [11]. The load growth was set at plan identified by the EPSO algorithm remains unchanged.
2,5% per year and the discount rate was set at 5% per year, the This plan was obtained running 112 iterations in 42 minutes.
number of particles in each population of the EPSO algorithm The evolution of the solution is shown in Fig. 6 below and for
was set at 50, the planning horizon is 3 years and the PNS each analyzed case Table I provides the values of the indices
penalization cost was set at 109 $/MW. The EPSO algorithm used to analyze the obtained expansion plan considering the
stops after running 50 iterations with the same best solution. DG penetrations that were tested.
The simulations were run in MATLAB with an Intel i7,
The best solution found by EPSO algorithm corresponds to
3.4GHz, 8 GB RAM.
the installation of:
As the annual peak load for this system occurs after 6 pm,
- Year 1 - a transformer between bus 3 and 24, one
the solar distributed generation has no impact in decreasing it
138 kV cable connecting bus 6 to 10, one 138 kV
since no associated storage devices are being considered.
line connecting buses 7 to 8;
Additionally, the optimization process takes into account only
the minimization of the investment cost for new equipments - Year 2 - one 138 kV line connecting buses 1 to 5 in
on the grid while ensuring that the annual peak demand is the second year.
supplied. Therefore, for the considered cases (solar DG
These equipments allow the system to gain flexibility
 enough to accommodate the demand in Period 3 so that no new equipment is required in this period.
Table 1: Comparison of the results for the 4 DG penetration scenarios.

DG Inv. Cost Op. Cost EENS Trans. Losses CO2 Emissions

Cases
penetration (106 US$) (106 US$) (MWh/year) (MW) (t CO2 )
1 0% 98.05 47.60 38295,98 14808.07 62131.47
2 10 % 98.05 39.08 29563,18 14072.61 51495.52
3 15 % 98.05 35.43 27126,98 13690.19 47620.69
4 20 % 98.05 32.47 25739,72 13372.85 43649.91

seen by the transmission network remains unchanged. As a

result, even if the injected solar DG is increased to 20% of the
peak demand in each bus, the optimal expansion plan remains
unchanged. This suggests that solar DG is not able in an
isolated way to reduce the liquid demand seen by transmission
networks and thus contribute to postpone transmission
investments and reduce the corresponding cost. Therefore, as
main conclusion of this paper, solar DG should be associated
to storage devices or demand side management programs
should be implemented in order to reduce the investment effort
in transmission networks.
ACKNOWLEDGMENT
The first author thanks CAPES Foundation, Ministry of
Figure 6. Evolution of the best particle of the EPSO Algorithm. Education of Brazil, for financing this research.

VII. CONCLUSIONS REFERENCES

[1] W. Li, Probabilistic transmission system planning. Wiley-IEEE Press,
This paper describes a multiyear transmission expansion 2011.
planning considering solar distributed generation penetration. [2] J. H. Zhao, J. Foster, Z. Y. Dong, and K. P. Wong, Flexible
The planning problem is conducted considering the investment Transmission Network Planning Considering Distributed Generation
cost and a penalization for Power Not Supplied over the Impacts, IEEE Trans. Power Syst., vol. 26, no. 3, pp. 14341443, Aug.
2011.
planning horizon. On the other hand, Evolutionary Particle [3] R. S. Maciel, M. Rosa, V. Miranda, and A. Padilha-Feltrin, Multi-
Swarm Optimization is used to solve the optimization problem objective evolutionary particle swarm optimization in the assessment of
via a parallel computing approach as a way to reduce the the impact of distributed generation, Electr. Power Syst. Res., vol. 89,
pp. 100108, Aug. 2012.
computation time. The develop approach models solar [4] P. Dondi, D. Bayoumi, C. Haederli, D. Julian, and M. Suter, Network
distributed generation as negative loads in each bar considering integration of distributed power generation, J. Power Sources, vol. 106,
the solar penetration proportional to the demand. no. 1, pp. 19, 2002.
[5] S. Haffner, L. F. A. Pereira, L. A. Pereira, and L. S. Barreto, Multistage
The main motivation to develop this TEP approach comes Model for Distribution Expansion Planning With Distributed
GenerationPart I: Problem Formulation, IEEE Trans. Power Deliv.,
from the need to consider in the transmission planning task the vol. 23, no. 2, pp. 915923, Apr. 2008.
new load behavior present in the transmission-distribution [6] R. D. Zimmerman, C. E. Murillo-Sanchez, and R. J. Thomas,
boundary. The new load profiles are becoming increasingly MATPOWER: Steady-State Operations, Planning, and Analysis Tools
for Power Systems Research and Education, IEEE Trans. Power Syst.,
affected by Plug-in-Electrical Vehicles, Demand Side vol. 26, no. 1, pp. 1219, Feb. 2011.
Management Programs, Distributed Generation (DG) [7] V. Miranda, H. Keko, and . J. Duque, Stochastic Star Communication
penetration and Microgrid and Smart Grid installations. Topology in Evolutionary Particle Swarms (EPSO), Int. J. Comput.
Intell. Res., vol. 4, no. 2, p. 105, 2008.
Although affecting in a direct way distribution networks, these
[8] P. Vilaa Gomes, J. Silva, and J. T. Saraiva, Multiyear and Multi-
aspects impact indirectly on the transmission grids. Criteria AC Transmission Expansion Planning Model Considering
Reliability and Investment Costs, in IEEE EEM Porto 2016, 2016.
The results presented in Table I indicate that the penetration [9] P. Subcommittee, IEEE Reliability Test System, IEEE Trans. Power
of solar distributed power generation can provide a better and Appar. Syst., vol. PAS-98, no. 6, pp. 20472054, Nov. 1979.
safer operation for the system while reducing the operation [10] P. V. Gomes and J. T. Saraiva, Hybrid Discrete Evolutionary PSO for
AC Dynamic Transmission Expansion Planning, in IEEE Energycon
costs, the transmission losses and the CO2 emission level while conference, 2016.
contributing to diversify the generation mix. In this case the [11] C. A. Correa, R. Bolanos, A. Garces, and A. Molina, Multiobjective
peak of injected solar DG is not coincident with the annual environmental transmission network expansion planning, in 2013 IEEE
Grenoble Conference, 2013, pp. 15.
demand peak period which means that the annual peak demand