1. Introduction
Studies of network development planning aim to analyze the interaction of grid configuration, generation and demand. Providing demand envisions, the scheduling goal is to deal with generation evolution (e.g., replacing out-to-date technologies with renewables and innovative ones), combined with demand load trend evolution taking into account economical, reliability, continuity and environmental factors. Further, network developing analysis represents a crucial issue due to the analytics intricacy and the big data management. These reasons have brought the definition of Transmission Evolution Planning (TEP) and Generation Evolution Planning (GEP) or combined G&TEP. With regard to TEP, the Transmission System Operators (TSOs) must consider the uncertainty of future framework of load demand required or renewable generation penetration satisfying technical constrains and ensuring reliability and security, by assessing branch doubling or new grid assets.
G&TEP is overwhelming modelled, as mono- or bi-level programming-based optimization methods to minimize costs, and it is formulated as Mixed Integer Linear Programming (MILP) [
1,
2,
3,
4], Mixed Integer Nonlinear Programming (MINP) [
5] or Robust Optimization [
6]. Generation and transmission investment costs are the essential terms present in the objective function for all the approaches [
1,
2,
3,
4,
5,
6] whereas additional aspects involve operating costs, expected energy not served (EENS), losses cost, load shedding costs.
TEP mathematical optimization methods can be grouped in two main categories: programming-based and heuristic optimizations. Linear programming (LP) [
7,
8,
9,
10,
11], MILP [
12,
13,
14,
15,
16,
17], robust optimization (RO) [
18,
19,
20,
21] or games theory [
22] are the spread methods of the first group. Instead heuristics can include K-medoids [
23], Gbest-Guided Artificial Bee Colony [
24], Particle Swarm Optimization [
25,
26], multi-criteria decision making and differential evolution [
27] social spider algorithm [
28], evolutionary algorithms [
29], semi-definitive programming and branch and bound [
30], symbiotic organism algorithm [
31], or combined search space reducer and bat-inspired algorithm [
32]
As regards objective functions, in TEP new transmission line investment costs are usually considered, but additional operating costs or penalties could be included, such as generation operating costs [
10,
19,
24,
25], matched with unserved energy costs as in [
9,
20,
21,
23,
32], or with load shedding as in [
11,
14,
15]. Other works consider loss-of-load probability and load curtailment costs [
13], renewable curtailment costs [
8], load shedding and renewable curtailment costs [
16], losses penalty factor [
30], or operational costs, generation curtailment and load shedding costs [
17]. Few works neglect generation operating costs in objective function but embed new factors as in [
26] where weighted vulnerability factors are introduced to optimize the system security, or in [
27] whose purpose is to avoid congestions by means of a penalty factor. A novel case of investment costs omission is evaluated in [
7], rather the focus is on the generation costs including fuel, O&M and CO2 costs. In multi-level TEP optimization the investment costs are considered in the last level. Detailed market aspects are modeled in [
22] where the first-level optimization pursues generation costs minimization and consumer surplus maximization, the second-level aims at maximizing zonal generation and consumer surpluses and congestion rent earned by the TSO, while the third-level goal is to minimize investment costs and maximize global generation and consumer surpluses. The authors of [
18] minimize the generation and load shedding costs in the level-I, and investment costs in level-II.
A crucial perspective in the TEP optimization constraints is the network model. The most employed formulation is the DC load flow (LF) [
1,
2,
3,
6,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
21,
22,
23,
27,
28,
31], in few papers the losses have been embedded by means of quadratic expression in the DC formulation [
5,
29,
32]. Nevertheless, other approaches exploit power transfer distribution factors (PTDFs) [
4,
7,
8], shifting factors [
11,
26], economic dispatch [
20] or power balance [
29] but all the approaches omit the behavior of losses. AC LF formulation is adopted in [
25] and [
30]: the first one compares the AC optimal power flow (OPF) model with DC-TEP formulation considering piecewise linear losses, while the second represents the losses as difference between generation and load. As reported above, the TEP optimization methods focus on economic terms, with simplified network models to reduce formulation complexity and reduce computational cost. However, as demonstrated in [
25] the AC OPF solution requires higher capacity installation and lower operating costs compared to the DC-TEP ones. Besides, [
33] investigates the influence of losses model in the TEP solution, underlining the variation in investment costs proving that in large-scale systems the losses have a relevant impact.
A separate set of approaches, guided by TSO applications, involve solution techniques of TEP problem different from optimization. The combination of DC-OPF and cost-benefit analysis (CBA) is the common framework employed to evaluate the candidate project selection. The CBA is conducted by means of: wind spillage and production costs indices [
34]; the comparison with and without weighted environmental aspects indices [
35]; present-value, welfare, investment costs indices [
36]; investment costs, congestion costs and risk costs minima [
37]; reliability indices and investment, operational and risk costs minima. In [
38] ENTSO-E CBA is evaluated through a software called SCANNER. In [
39] electrical market with Multi-Area Power-Market Simulator software is considered, with detailed model of intermittent and hydro generation, comparing results by CBA based on investment costs, transmission and generation capacity. A flow-based optimization market capacity is proposed in [
40] exploiting PTDFs and remaining available margins (RAMs) through the corridors. Moreover, phase shifter transformers (PSTs) are connected over cross-zonal branch to adjust RAMs by means of an optimization. The authors of [
41] developed a flow-based methodology, solving Day-Ahead Market (DAM) and AC LF to state the network operating conditions and evaluating the economic benefits introduced by candidate projects according to losses, generation redispatching, renewable and load curtailment reductions.
Nowadays a worldwide green energy generation and consumption transition is developing and TSOs are organizing a long-term TEP forecasting renewable generation and demand growth. In the field of scientific research this issue is tackled by multi-scenario [
2,
7,
12,
14,
20,
23,
25,
35,
38,
39], scenario clustering [
13,
16,
17] or multi-year approach [
5,
6,
9,
18,
19,
21,
36,
37]. In the first, framework generation and load are uncorrelated, in the second the gathering of scenarios is related to cost minima, while in the last the increases are correlated and predefined. The considered network uncertainties are mainly represented by intermittent renewable generation and load demand.
The power grid size is a relevant factor in terms of TEP computational cost and method extension to real power systems. Some works exploit test networks, such as 4-bus [
22,
40], Garver 6-bus [
24], 9-bus [
1], IEEE 24-bus [
4,
9,
25,
27], IEEE 30-bus [
6], IEEE 118-bus [
13,
17,
18,
41] and 120-bus [
16]. TEP methods are applied to real network models as well, such as Australian 24-bus [
7], German zonal market [
8], Romanian [
15], European Zonal Market [
23,
35], small-scale China [
26], Northern Europe [
38,
39], U.S. 240-bus [
2] and 3000-bus Northern China [
34]. Other works perform comparisons of different networks, involving test networks of different sizes [
10,
19,
20,
28], or matching simple test system with real network models such as 93-bus Colombian [
12], 46-bis and 87-bus Southern Brazilian [
28,
29,
30,
31,
32], Iranian Power grid [
5,
37].
Simulation time reduction also depends on the number of candidate projects to evaluate, and an established technical analysis is helpful to reduce the set dimension embedding the ones with higher benefits. For this purpose, in [
7] the set is determined according to RES penetration, in [
9] a relaxed version of the TEP problem is solved to quantify the investment pool with most benefits. Moreover, in [
10] following the load flow results the reinforcement for congested corridors is considered and locational marginal price advantages for new line addition, in [
20] a method based on long- and short-terms network uncertainties is developed to pinpoint the candidate investments, and in [
23] different typology of candidate projects are determined according to the potential benefits introduced. For candidate selection, in [
29] a load and an angle performance index is defined, in [
31] a search space reducer algorithm is solved, in [
34] cost-benefit incremental relationship and sensitivity factor of branch capacity and admittance are evaluated, in [
35] the probability of branch overload is considered, while in [
41] an equivalent positive and negative critical overload duration is introduced. Further methods, applications, and evaluations are reported in the review papers [
42,
43].
There are few papers that include N-k security criteria in TEP assessment. In particular, optimization problems are faced including N-1 security constraint in the formulation [
5,
6,
29,
40], or involving both N-1 and N-2 security criteria generating a set of contingency scenarios [
4]. The authors of [
10,
44,
45] evaluate an N-k security in the second stage of the procedure in order to define the set of candidate projects and/or to obtain optimal solution, whereas in [
37] a risk index is defined including the line outage probability.
From literature analysis, it stems that the formulation of a TEP problem has several facets that are hardly caught in the presence of a real-size network, where cost-benefit analysis of single development projects is usually carried out [
46,
47]. Moreover, different implications of the development projects should be assessed in the form of scenario analysis. In order to perform useful comparison among different projects by accounting non-commensurable quantities, multi-criteria analysis can be adopted, with particular reference to Analytic Hierarchy Process (AHP) due to high flexibility and applicability [
48,
49]. This technique has found application in power system problems such as Generation expansion planning, in a multi-objective model with detailed network representation in [
50] and encompassing financial, technical, environmental and social aspects in [
51], or distribution system planning [
52]. However, TEP problems represent a field of application of AHP for multi-criteria analysis. In particular, application of test network involve IEEE 24-bus test system in [
53], where a multi-objective optimization involving congestion cost, investment cost, probabilistic reliability, anti-competition and anti-flexibility indices allows to obtain the Pareto front and it is supported by AHP and TOPSIS to determine the best solution, in [
54], exploiting a two-stage TEP algorithm where cost minimization results are analyzed with AHP considering congestion cost, consumer cost, power losses and voltage deviations, and in [
27], where dynamic evolution technique is underpinned by AHP for ranking the best compromise solution. Moreover, in [
55] indices of economy, safety, flexibility and vulnerability are taken as criteria of fuzzified AHP method with different comprehensive weights for IEEE 6-bus network with different planning schemes, whereas in [
56] location marginal prices from AC load flow in IEEE 9-bus system are used for individuating candidates combined with AHP. As regards real network applications, in [
57] a combination of AHP and entropy weight is adopted to evaluate three candidate projects by means of indices of safety (including N-1 and N-2 security ones) and reliability, economy and efficiency, coordination and flexibility, social aspects and risk control. Whereas in [
58] a Brazilian network is analyzed considering AHP for probabilistic, strategic, financial, externalities and enterprise risk, and in [
59] Paraguay transmission system expansions are analyzed with AHP considering operation and inversion cost, power losses, line length and project financing. It can be noted that the analysis of AHP in TEP has seldom accounted for evolution scenarios of load and renewables, and methods are focused on network with limited extension.
In this paper, a methodology for performance analysis of a portfolio of network development projects is proposed, in order to evaluate the subset of projects towards which the TSO should focus its realization efforts according to positive implications on different technical aspects and limited economic effort. In particular, the methodology aims at assessing zonal market framework, with linear bids and inelastic demand, and AC LF analysis, in the base case network to individuate possible candidate projects. The same tools are exploited in the presence of network development candidate project, in order to calculate merit indicators on active power losses, admissible load increase and admissible renewable generation increase. These merit indicators are compared among candidate projects by means of Analytic Hierarchy Process method, in order to determine the most promising solution under different weights of criteria, representing an evaluation of various evolution scenarios. A further analysis implies the influence of investment cost as economic merit indicator, and its inclusion in AHP is carried out in order to point out the impact of economic efforts on the multi-criteria decision framework. The procedure is applied to NREL-118 test system.
The contributions of this paper can be synthesized as follows:
- -
The full network representation by means of AC LF evaluating active e reactive power flow and effective power losses.
- -
The evaluation of candidate projects set according to base case network operating condition within one year of observation.
- -
The adoption of AHP approach evaluating the candidate projects according to different weighted indices of losses reduction, admissible load increase, admissible renewable penetration, and investment effort.
- -
Differently for cost-benefit analysis dealt with in [
41], the procedure does not involve an economic quantification of technical benefits, whereas it is aimed at comparing heterogeneous implications of network development projects in a normalized way.
The remainder of the paper is organized as follows.
Section 2 is devoted to the description of the multi-stage methodology for network operation analysis, candidate project selection, performance indicator definition and comparison technique. The test system and the base case analysis are presented in
Section 3, whereas the network development projects are assessed in
Section 4 along with their comparison. Conclusions are drawn in
Section 5.
2. Methodology
The determination of the network development initiative follows a multi-step methodology, synthesized in the following points:
Study of base case operation according to techno-economic programming over a defined time horizon.
Individuation of candidate network development projects, able to produce effects on system behavior.
Carrying out of scenario analysis for each candidate project and determination of the merit indicators.
Selection of the most promising projects.
2.1. Power System Techno-Economic Operation
In order to evaluate the operating conditions of the considered power system, technical and economic considerations should be accounted. The combination of these aspects can be assessed in optimal power flow analysis [
60], however in the outline of an unbundled energy sector, the presence of energy markets should be considered. Therefore, the adopted method to determine power system operation is structured as follows.
For each operating condition to be analyzed, represented by the
-th time step in the considered time window, the procedure involves the solution of a zonal energy market with quadratic generation bids and inelastic load demand, whose formulation can be synthesized as follows:
s.t.
where:
is the number of market zones and is the zone index;
is the number of buses and is the bus index;
is the number of generators and is the generator index;
is the number of interzonal connection and is the zonal interface index;
is the total amount of time step and is time step index;
is a binary parameter and it indicates if the -th generator is connected (1) or not (0) to the -th bus;
is a binary parameter and it indicates if the -th bus is connected (1) or not (0) to the -th zone;
indicates if the -th power exchange is entering (1), exiting (−1) or not connected (0) to the -th zone;
is a binary parameter and it indicates if the -th interzonal exchange is connected to the z-th zone (assuming 1 or −1 if the positive exchange is exiting or entering the -th zone), whereas it is 0 if the -th interzonal exchange does not involve the -th zone;
is the inelastic active power demand at the -th bus at -th time;
and are the linear and quadratic bid coefficient of the -th generator;
is the availability of the -th generator at -th hour;
is the rated active power of the -th generator;
is the rated active power exchange of the -th interzonal border;
is the generated power of the -th generator at -th time;
is the power exchange at the -th zonal interface.
In particular, Equation (2) represents the power balance constraint, Equation (3) explicates the generator technical limits, and Equation (4) introduces the zonal interface limit constraints.
The absence of generators technical minimum avoids the presence of block order bids that involve entirely accepted or rejected bids conditions according to the market clearing price, for each hour. These bids entail a Mixed Integer Linear Programming with binary variables that state all-or-nothing constraints, which in turn leads to a counterintuitive market solution called Paradoxically Accepted/Rejected Blocks, described in [
61,
62].
It should be remarked that the maximum interzonal power exchange across the -th border is strictly related to the active power flow rating of all the -th branches constituting the -th border. For instance, it could represent the Available Transfer Capacity (ATC) value in N or N-1 conditions, or come from other security considerations.
The output of the energy market is represented by the power generation plan of the dispatchable generation present in the power system able to minimize the objective (e.g., reduce the generation cost) in the presence of zonal constraints. However, the impact on the behavior of network elements should be assessed as well. Therefore, a steady-state network analysis is performed, considering the distributed load flow framework with full AC formulation, developed as follows:
s.t.
where:
is the total number of branches and is the branch index;
is the total generated active power by the -th generator;
and are amplitude and phase of the nodal admittance between and buses (coming from the construction of the nodal admittance matrix ;
and ( and ) are amplitude and phase of nodal voltage at -th (-th) bus at -th time step;
is the loss participation factor of -th generator;
is the system total active power loss at time step ;
is the amount of active power losses across -th branch at -th time step;
and are binary parameters and they indicate if the -th (-th) bus is connected (1) or not (0) to the -th branch;
and are the resistance and reactance values of the -th branch;
is the imaginary unit.
The distributed load flow is considered in order to share the burden of active power losses balance—not considered in zonal energy market solution—with a limited though diffused stress on the selected generators.
In outcome of the analysis, further than the determination of nodal voltage, the amount of active, reactive and apparent power flowing across the
-th branch, named
,
,
, are determined from the following complex equation:
where the superscript * stands for complex conjugate value
This double-layer analysis is performed for each time step of the considered time horizon.
2.2. Selection of the Candidate Projects
From the power system techno-economic operation analysis, and particularly from the determination of power flowing through branches, the loading analysis of network connection can be carried out.
In particular, for each
-th branch, the loading factor in each time step
is determined as the ratio of absolute value of power flow
on active power flow rating
, as follows:
For the base case, the average value of the loading factor
throughout the considered time horizon and a duration curve of loading factor (sorting the values from the highest to the lowest, irrespective of the time step position in the horizon) can provide synthetic evaluation of the operation stress of the
-th branch, thus individuating the paths that would benefit more from a reinforcement project. The formulation of
can be generalized as follows:
According to the adopted operation planning standard, the overloaded branches can be individuated if the power flow exceeds the rating value by a suitable margin
; therefore, no overload is observed if the following condition is satisfied
From the theoretical framework of the zonal market, it could be expected that more stressed connections are placed across the zones and not within each zone. Therefore, a first selection is made considering the doubling of existing connections across each couple of zones.
However, further connection lines could be individuated as well, in order to improve the network meshing, providing different paths for power routing that could increase the efficiency, although they could represent a more costly solution. A second selection of candidate projects involves new connections between couple of nodes pertaining to different zones, not interested by existing line or existing market zone connections.
2.3. Scenario Analysis of Development Projects
The impact of the candidate project is assessed by means of a PINT approach, therefore each project is analyzed separately, as described in the following.
Differently from the determination of techno-economic benefits at target years according to defined evolution of system generation and demand, the proposed approach aims at determining the effect of the presence of development projects in the considered system through technical merit indicators.
A first indicator is represented by the variation of total active power losses induced by the presence of the
x-th candidate project. In order to perform this estimation, the energy market in Equations (1)–(4) is solved accounting for the presence of the
-th development project, affecting the inputs of the rated active power exchange
, and the load flow analysis in Equations (5)–(11) is carried out considering the influence of the
x-th candidate project on the nodal admittance matrix
. Therefore, the global power system operation is varied in each
-th time step. The indicator
quantifies the energy losses reduction benefits, for the
-th development project over
observation period, with respect to the base case network, and it can be expressed as follows:
where:
is the total number of candidate development projects and is the candidate project index;
represents the -th branch active losses in the presence of x-th candidate project at time step .
Furthermore, in order to investigate the effect of the project on possible evolution of the generation and demand, and particularly on the attainment of targets for increased energy service for users and reduced environmental impact of power system, the considered power system is subject to increase of load demand and of renewable generation scenarios.
In the load increasing scenario, the load demand is increased by 1% for each iteration for each load bus in each time step. In order to ensure proper balance and avoid power shortage, the generation capacity is incremented uniformly, of the same amount. Therefore, for each load iteration
the energy market in Equations (1)–(4) and the load flow problem in Equations (5)–(11) are solved with new input parameter of load demand
and maximum generation level
, defined as follows:
At each
-th load iteration, the branch loading
of each
-th branch in each
-th time step is determined. If no overload is detected according to the adopted planning standard, i.e., Equation (15) is satisfied, the procedure goes on to the next iteration, otherwise the procedure stops at a given iteration number
, and the admissible demand increase in the network under study is given by:
The load increase indicator of the
-th candidate project is therefore determined by difference between the result of the developed network (considering the input variation on
and
further than the scenario influence) and of the base case (subscript
), as follows:
In the renewable increasing scenario, the power generation amount by renewable-based generation technologies is increased by 1% for each
-th iteration in each time step. In this case, no intervention on power balance is operated, i.e., load demand and conventional power generation are not varied, since the aim is to investigate the effect of a growing renewable share in the power generation mix. Therefore, for each renewable iteration
the energy market in Equations (1)–(4) is solved with new input parameter of maximum generation level
applied to the
-th renewable generator (in the subset
of renewable generators, in number
) defined as follows:
According to energy market results, giving different production levels for all generators due to a new equilibrium point, the load flow analysis (5)–(11) is carried out.
At each
-th renewable iteration, the branch loading
of each
-th branch in each
-th time step is determined. If no overload is detected according to the adopted planning standard, i.e., condition (15) is satisfied, the procedure goes on to the next iteration, otherwise the procedure stops at a given iteration number
, and the admissible renewable generation increase in the network under study is given by
The renewable generation increase indicator of the
-th candidate project is therefore determined by difference between the result of the developed network (considering the input variation on
and
further than the scenario influence) and of the base case (subscript
), as follows:
A representation of the technical merit indicator determination process is depicted in the flowchart reported in
Figure 1. It can be noted that the method involves the storage of network operation analysis under different conditions and in the presence/absence of development project, therefore a specific calculation framework is necessary in order to collect the necessary information.
Finally, in order to account for economic implications, an estimation of the investment cost of the -th candidate project is carried out, according to standard building and installation cost for the single components.
2.4. Project Comparison and Selection
As highlighted in the methodology, the three merit indicators, although referring to comparable units, measured as energy amounts over a given time horizon, are determined according to different operating conditions and evolution frameworks of the system under study. Therefore, in order to carry out a proper comparison among the outcomes of the analysis of candidate projects, Analytic Hierarchy Process (AHP) is adopted [
63,
64].
The AHP is based on the determination, for each
-th criterion evaluated for
options, of the
pairwise comparison matrix
, whose element
represents the prevalence of the
-th option compared to the
-th one. If the
-th option is preferred to the
-th one, then
, in a scale of values up to 9 according to importance comparison; for equal importance it is
; moreover, the following reciprocal constraint applies:
Once the matrix
is built, its normalized version
by column is obtained, and its elements are determined as follows:
By averaging the entries of each row of
, the
score vector
for the
-th criterion is determined:
By padding the vectors
by columns, the
score matrix
is obtained:
Proceeding in the same way, the pairwise comparison matrix of the criteria is determined whose element represents the prevalence of the -th criterion compared to the -th one. Applying the same normalization and averaging process described in Equations (25) and (26), the criteria weight vector is determined.
The
vector of global scores
is therefore determined by the following matrix operation, where the element
represents the global score assigned by the AHP to the
-th option.
For the application to the proposed framework, the -th candidate project represents the generic -th option, whereas the three indicators , and represent the criteria.
In addition, the consistency check is performed on pairwise comparison matrices. Taking
as a reference, the consistency index is determined as follows:
where the first term
is a scalar determined as the average of the elements of the vector obtained by multiplying the rows of
by
and dividing by the corresponding element of
:
The consistency ratio is therefore determined as where the random index is determined as the average when elements of are random. It is considered that the consistency is acceptable if .