Multicriteria decision analysis allows to organize and structure complex decision problems characterized by multiple, often conflicting objectives. Several scholars in the field of MCDA have stressed that MCDA should not be mistaken for decision-making techniques, but rather techniques for analysis and aid (Keeney, 1982; Roy, 1996; Belton and Stewart, 2002). Keeney (1982) for example wrote that “decision analysis will not solve a decision problem, nor is it intended to.” Belton and Stewart (2002) stated that the main goal of MCDA should be “to facilitate decision-makers’ (DM) learning about and understanding of the problem faced, about their own, other parties’ and organizational priorities, values and objectives and through exploring these in the context of the problem to guide them in identifying a preferred course of action.” As discussed in the studies by Dodgson et al. (2009) and Wang et al. (2009), from the definition of the problem to the desired decision analysis outcome, four mains steps should be followed (Figure 1). This process is not necessarily sequential and may have iterations (Guitouni and Martel, 1998).

A multicriteria decision problem thus essentially consists of a set of alternatives that are evaluated on the basis of conflicting and incommensurate criteria, according to the DM’s preferences (Malczewski and Rinner, 2015). Therefore, the main elements of any multicriteria decision problem include values, alternatives, criteria and their weighting, and DMs.

Values can be defined as principles or beliefs, held by individuals or groups, which reflect their conception of what are good or desirable states or behaviors (Connelly and Richardson, 2005; Balint et al., 2011). In this sense, they rationalize actions and guide the selection or evaluation of behaviors and events. According to the study by Keeney (1992), values are typically indicated in seven different forms (ethics, traits, characteristics, guidelines, priorities, value tradeoffs, and attitude toward risk). To be of use in decision-making, they are made explicit through associated statements, criteria, objectives, and weights. Values are rather ends in given time horizon and should not be confused with means. As proposed in the study by Keeney et al. (1987), a “value tree” can be established to elicit the values of stakeholders and derive an organized hierarchy of corresponding criteria that achieve or describe the given values.

An alternative is a means toward the satisfaction of the values and criteria. In some cases, it is the aim of MCDA to short list a wide range of existing alternatives and, in others, the identification of alternatives is itself a necessary and active process (Dodgson et al., 2009). As suggested by Keeney (1992), alternatives are only important in the way that they satisfy the values (and the criteria involved), and their identification should therefore be driven by these values. Alternatives can be generated either automatically (as often the case in multiobjective optimization) or manually (Pedrycz et al., 2011).

Although there is no standard definition of the expression “criteria” (Nijkamp et al., 1990), generally speaking, a criterion represents a standard of judgment to test the acceptability of an alternative (Pedrycz et al., 2011). In multicriteria literature, it is used to describe two distinct concepts: objectives and attributes. Objectives indicate the desired direction toward which a DM wishes to move (for example, minimum cost or maximum energy efficiency) (Pedrycz et al., 2011). We can distinguish objectives from goals, which represent a threshold or target level to reach, in terms of a specific state in space and time, while the objective gives the desired direction (Hwang and Masud, 1979). Attributes are a set of characteristics chosen by DMs that measure the performance of an alternative (e.g., how it impacts employment, quality of life, environmental parameters) (Hwang and Masud, 1979; Pedrycz et al., 2011). Weights are determined to indicate the relative importance or preference of criteria (Wang et al., 2009).

The DM is but one among several types of actors in a decision process. Roy (1996) describes three main types of actors: stakeholders, third parties, and analysts. The former are those who have an important interest in the decision and directly intervene in the decision process. They consist either of individuals, a clearly defined group of individuals (an elected body, a panel of experts, etc.), or a group with less well-defined boundaries (a lobby, public opinion, etc.). Usually, the objectives and values of the different stakeholders are diverse and conflicting. Because the decision aid cannot simultaneously benefit all stakeholders comprehensively, one stakeholder is generally identified as the DM. The second type of actors are referred to as third parties, as they do not actively take part in the decision process, but are affected by its consequences, and thus whose preferences must be taken into account (typically the citizens, end-users, consumers, etc.). A third type of actor is the analyst, who plays a role in supporting the DM. In some cases, the DMs may develop the decision aid themselves, but generally this task is performed by an analyst different from the DM. This is in particular the case when the DM does not possess the technical or methodological background, or when an external party is desired to ensure a neutral and more objective approach. In summary, stakeholders are the actors who actively take part in the decision process and include the DM. These can be assisted by an analyst whose role is to provide methodological support to help answer questions posed by a stakeholder in a decision process. This support must take into account the interests of not only the stakeholders, but also third parties who are affected by the decision.

Rationales for using MCDA in energy planning have been discussed and compared to other decision support methods (cost–benefit approach, cost-effectiveness approach, energy ecological footprint, etc.) or simply to informal judgment unsupported by such methods (Dodgson et al., 2009; Wang et al., 2011). The main arguments in favor of MCDA from the studies by Pohekar and Ramachandran (2004), Dodgson et al. (2009), Browne et al. (2010), Coelho et al. (2010), Ghafghazi et al. (2010), and Wang et al. (2011) can be summarized as follows: –

Useful to resolve conflicting interests and reach compromises

Browne et al. (2010) nevertheless point out two key drawbacks of MCDA in energy planning, namely the dependency on subjective judgment in qualitative approaches and the difficulty to quantify environmental or social impacts. MCDA has also been criticized for providing inconsistent results, being highly dependent on method choice and subjective stakeholder preferences (Guitouni and Martel, 1998).

Energy planning is a complex issue for involving not only multiple criteria but also for spanning across multiple scales, from the building scale up to national or international scales (Wierzbicki et al., 2000; Thery and Zarate, 2009; Prasad et al., 2014). However, the urban scale has received particular attention in the past years regarding energy issues, as cities are acknowledged to play an important role in mitigating climate change. The International Energy Agency estimates that cities represent two-thirds of the world’s energy consumption and GHG emitters and could therefore notably curtail climate change impacts through local authority engagement (IEA, 2008). The Intergovernmental Panel on Climate Change has recently noted urban planning and other city scale interventions among the key measures for climate change mitigation (IPCC, 2014). The United Nations similarly advocates municipal-led energy conservation and carbon emission reduction plans (Habitat-Iclei, 2009). Cities themselves have recognized their strengths, as noticeable through the Aalborg charter signatories (Aalborg Charter, 1994), who declared “that the city or town is both the largest unit capable of initially addressing the many urban (…) natural resource and environmental imbalances (…) and the smallest scale at which problems can be meaningfully resolved in an integrated, holistic and sustainable fashion.” The “Covenant of Mayors” (Covenant of Mayors, 2016) is a city-based European initiative, which aims to help cities achieve the EU’s climate targets (EC, 2007, 2014), by fostering exchange of experience. The network of megacities C40 also acknowledges, through various studies, the important role cities play in addressing climate change (C40, Arup, 2014; C40, Arup, UCL, 2015).

This trend for planning energy at the urban scale is further supported by research in various fields. The urban level is believed to offer a more direct and concrete space to act regarding energy consumption and supply, allowing for effective adaptation of policy measures to local specificities (Chapuis et al., 2010; Keirstead and Schulz, 2010; Caputo and Pasetti, 2015). St. Denis and Parker (2009) identified that given recent technological advancement and better access to local knowledge, communities have become more active in planning their own energy systems. They identified several advantages at this local scale, including increased social input and participation of locals, more agile responses to opportunities and threats, more personal investment and interest of actors, and a clearer link between local consumption and generation. On the other hand, the more localized scale of the individual building scale is found less effective than wider district or urban scales, neglecting possible benefits of considering multiple buildings on the demand side (e.g., by taking shading into account when planning the building stock) and on the supply side (e.g., enabling economies of scale or energy-efficient centralized systems) (Ratti et al., 2005; Koch, 2009; Zanon and Verones, 2013; Immendoerfer et al., 2014; Petersen, 2016). As a means to effectively perform this “upscaling” (UN-Habitat, 2009; Gossop, 2011; Zanon and Verones, 2013; Cajot et al., 2015; Strasser, 2015) have argued that energy issues should integrate existing urban planning processes. A survey from UP-RES (2011) made clear the interest and need from building scale specialists to access training regarding the energy solutions on the urban scale.

The planning of energy infrastructure is not essentially new to the wide range of urban activities; however, its handling by means of integrated, cross-sector, and multiactor approaches is relatively recent. Many cities are struggling to develop new methods to successfully bring together energy issues in the framework of urban planning procedures (Zanon and Verones, 2013; Immendoerfer et al., 2014; Strasser, 2015). Because of the novelty of such approaches, the term of urban energy system planning itself can be subject to debate and is worth clarifying. Based on a compilation of fragmented definitions of its subterms from literature, we put forward a synthetic definition of UESP, in an attempt to facilitate and improve the discussion on this emerging field.

On the basis of the above, we define UESP as the inclusion of energy issues (related to the acquisition and use of energy) in the processes of urban planning (which strategically and spatially organize the development of a city) to find environmentally friendly, institutionally sound, socially acceptable, and cost-effective solutions to satisfy the demands of an urban area.

In each reviewed study, the general scope of the problem was hereby investigated, considering the geographical location in which the problem was set (classified by continent), the physical scales bounding the problem (whether the focus of the planning was set on the building, neighborhood, city, regional, state, or country scale), and the temporal scale covered by the problem (noting any temporal horizon considered by the authors in their study). The different studies were also classified according to topic and planning focus. Concerning the topics, six broad themes were found sufficient to cover the extent of the reviewed papers, as follows: (1) heating and cooling, (2) power, (3) mobility, (4) environment, (5) waste, and (6) water/wastewater. The papers were further classified according to their planning focus as follows:

System specific planning: when only a subpart of the UES is considered, e.g., heating system, electrical system, demand side analysis in residential sector

This review question counted the number of alternatives analyzed by each study and the methods used to identify them. Keeney (1992) emphasizes the importance of alternative creation, writing that it “may be more important to create alternatives than to evaluate readily available ones.” He blames decision methodologies for often neglecting this aspect or inhibiting it. If many MCDA problems in literature may lead to believe that the typical application involves a predefined set of alternatives, this is often not the case (Belton and Stewart, 2002), and the more complex problems such as those found in UESP require careful thought in the structuring of the problem and identification of alternatives. For example, Feng and Lin (1999) point out the limitations of conventional approaches for generating new urban layout plans. They argue that the urban development process should begin with a systematic generation of physical layout alternatives, but deplore that conventional processes for generating alternatives are usually considered “a ‘black box’ inside which planners are subjective and alternatives are few.” In response, they propose an optimization model to systematically elicit alternatives maximizing public comfort and convenience, with which they were able to identify four alternatives performing better than the original plan. Typical methodologies that can be used to support alternative generation are discussed in the study by Siebert and Keeney (2015), who review existing methodological approaches for creating alternatives (structured techniques, creativity techniques, value-focused thinking, etc.). Mirakyan and De Guio (2014) reviewed the main methods that can support alternative identification, including brainstorming, soft systems methodology; strengths, weaknesses, opportunities, and threats (SWOT) approach; Delphi; means-ends objective network; and network of problems. They further proposed a methodology for finding innovative alternatives in planning, where common solutions may not be satisfactory, but where satisfying solutions are not obvious. Belton and Stewart (2002) present various methods for idea generation, some of which can help think about and identify alternatives. These include checklists to stimulate idea generation, alternative-based thinking methods, where users analyze and compare alternatives to stimulate the generation of new ones, and value-focused thinking using means-ends objective network, introduced by Keeney (1992), where value elicitation is performed prior to alternative identification. Despite the advancement of research in this field, the reviewed studies usually remained implicit or fairly superficial in describing how the alternatives were generated. In this context, the review distinguished the following cases for alternative generation: by the authors themselves, based on literature; by external experts or actors; by relying on heuristic approaches such as mathematical multiobjective optimization, systematic combinations, and enumeration; or simply externally defined (e.g., the evaluation of grant request submitted). When specified, any supporting method used by experts or actors was noted [including Geographic Information System (GIS) software, SWOT analysis, communities of practice (CoP), or interactive user selection].

This question considered the number of criteria used in the analysis and any explicit references to their selection process. These approaches included author’s judgment, literature or predefined sets of indicators, experts’ recommendations, and, when specified, other supporting methods (Delphi, CoP), or any combination of these approaches.

The problem type, as posed in the study, was surveyed. This question is of particular interest in the field of MCDA and can even be considered as the first and most important step in MCDA (Chen, 2006): when addressing complex decision problems, analysts must initially consider what type of result or outcome is expected. Roy (1996) first used the term “problematic” to describe the main types of outcomes MCDA can provide. In this review, we investigated which problematics were adopted, based on the most relevant problematics proposed by Roy (1996) and Belton and Stewart (2002). According to their typologies (Figure 3), a MCDA approach can be applied to

select or choose a “best” alternative, by reducing the set of alternatives to a smallest subset (choice problematic)

It should be noted that the delimitation between problematics is not always absolute and that a certain hierarchy can exist among them. For example, a first expected outcome may be ranking, followed by the choice of a single alternative. Similarly, portfolio can be considered a type of design problematic, in which alternatives are created by combining various subcomponents. In turn, the newly created portfolios can further be ranked, sorted, described, or chosen from. In the reviewed studies where multiple problematics concurred, the predominant one was retained, relying when possible on explicit statements from the authors.

Analysts must be aware of the type of results they aim to provide (a best single solution, a ranking of all solutions, the identification of original alternatives, etc.). Often, however, this appeared to be only implicitly considered or not considered at all. As the method used may constrain the type of result, this reinforces the argument of carefully identifying the expected decision outcome early in the decision process, to avoid the generation of undesired information and associated waste of time, effort, and cost.

This question investigated the extent to which the planning problems were driven by values, rather than means or resources readily available. In the reviewed papers, values most often took the form of “characteristics of consequences that matter” stated in the introductory sections of the papers or as “priorities” and “value tradeoffs” when establishing criteria and weights.

This question is particularly relevant in the light of the recent changes in urban planning, which shifted from physical, expert-led approaches, to a wider, strategic, and collaborative form (Albers, 1986; Wachs, 2001; Scherrer, 2008). A wider scope, including in particular energy issues as well as public participation, implies that planners must cope with a wider range of values than previously, which poses a series of challenges such as their systematic incorporation in decision-making or gathering information about the values (O’Brien, 2003; Balint et al., 2011). Connelly and Richardson (2005) and O’Brien (2003) point out the general lack of explicit acknowledgment of—and distinction between—the different values that should influence and drive planning. According to them, omitting them, assuming values are known, or underestimating their diversity may lead to unacceptable outcomes. Keeney (1992) originally advocated “value-focused” thinking, as opposed to what can be called “alternative-focused” or “resource-focused” thinking. According to him, failing to elicit values before defining alternatives leads to a constrained vision of the problem, which “anchors the though process, stifling creativity and innovation.” In addition, how alternatives are identified also directly influences how well the values shall be satisfied. Often, DMs do not spend enough effort on alternative creation, missing many alternatives, including important ones, hindering the achievement of the objectives (Siebert and Keeney, 2015). Exploring how values were involved in the reviewed studies, these were classified as either value driven or alternative driven, following loosely the framework from the study by Keeney (1992). This classification is not meant to be an absolute one, but rather to foster thinking about values in MCDA and how these could improve the final outcomes. Based solely on the reported elements in the final publication, and without the full knowledge of how the decision problem was actually addressed, the problem cannot easily be declared either purely value driven, or purely alternative driven. Instead, it lies within a continuum between both ends (Figure 4), depending on the reported information regarding values, and alternatives.

Several characteristics of value-driven approaches were used to evaluate the reviewed studies (Figure 4). As such, the studies that were more explicit in their elicitation of values, whose values in turn appeared to influence the identification of criteria and alternatives, and which considered a broader range of alternatives, or motivated the narrow set evaluated, were marked as value driven. Those who only moderately exposed the guiding values, or which values did not appear to influence the decision analysis, or which did not motivate the choice of alternatives were marked as alternative driven. To help in the classification, three types of alternative sets were used to describe the studies: ○

Type 1: The study includes and compares different categories of alternatives (e.g., renewable energy technologies, refurbishment, funding options, spatial alternatives, information, and incentives)

Three specific approaches for alternative generation (previously mentioned in Section “Alternatives”) were also found to be particularly in line with value-driven thinking, as they enable a transparent and comprehensive way to identify alternatives, with little a priori in their construction. The first approach was multiobjective decision-making (MODM)-based heuristics, in which a wider range of alternatives are automatically generated by combining decision variables to best achieve predefined objectives, allowing more informed and more rational decisions (Cohon, 1978). Similarly, some GIS approaches also assumed a broad, continuous range of locations as possible solutions, as opposed to a comparison of a limited subset of preselected sites. Other systematic combinatorial approaches were found, in which theoretically all possible alternatives of a category were considered a priori. This was for example the case when enumerating all possible priority sequences for restoration of district heating pipe segments (Rochas et al., 2015). Finally, to complete the assessment of planning drivers, the stated rationales for selecting the set of alternatives, if available, were also used to classify the studies (e.g., authors who provided strong justifications for the choice of alternatives were more likely to be classified as value driven as if the alternatives were not justified or appeared to be selected arbitrarily).

It should be noted that the value- or alternative-driven nature of a problem should not necessarily be regarded as “good” or “bad.” Indeed, the problem context may be simply conditioning an alternative-driven approach, as for example in the study by Hsueh and Yan (2011), where a method for comparing urban development projects that had applied for governmental grants is proposed. Similarly, some studies might not claim to find the best possible alternative to achieve a valued goal, but rather purposely aim to answer a more specific question about a given technology. This is for example the case of De Feo et al. (2008), who compares various coagulants for urban wastewater treatment.

This section aimed at identifying the MCDA methods, as well as the supporting methods, employed to address the different problems and the rationales behind their choice. There are many ways to categorize MCDA methods. In this study, they have been divided into five main groups, loosely based on classifications proposed by Belton and Stewart (2002), Zhou et al. (2006), Løken (2007), and Mardani et al. (2015): (1) value measurement models; (2) goal, aspiration, and reference level models; and (3) outranking models; (4) MODM; and (5) other methods. Value measurement models assign numerical scores to each alternative, by aggregating criteria and weights. Among the most common approaches include the weighted sum approach or analytical hierarchy process. The second group is often gathered under the expression goal programming, and a typical example is the Technique for Order Preference by Similarity to Ideal Solutions (TOPSIS). The third group of outranking models distinguishes alternatives in a pairwise fashion for each criterion. This group is also referred to as the French school of MCDA methods, namely because of the pioneering work on the ELECTRE methods from Roy (1996). The fourth group included MODM methods, which relied either on linear programming techniques or multiobjective optimization heuristics such as evolutionary algorithms. This group also contained methods that were referred to as single-objective decision-making (SODM). Although in most cases, SODM considers just a single objective, providing a somewhat limited interpretation of a problem, their name can be in some cases misleading. SODM can indeed handle multicriteria aspects of a problem by resorting to various techniques (Savic, 2002). This can be achieved for example by aggregating various objectives in a single-objective function through a weighted sum (Yokoyama and Ito, 1995; Ma, 2012; Karmellos et al., 2015), by setting different constraints on criteria to evaluate trade offs with the objective, as in the ε-constraint method (Pérez-Fortes et al., 2012), or simply by varying the criteria being optimized and comparing outcomes (Ayoub et al., 2009). The fifth group included methods relying on fuzzy set theory (FST) and other methods developed more recently and referred to as decision-making aggregation methods in Mardani et al. (2015). FST can be either used to extend existing MCDA methods (Wang et al., 2009; Mardani et al., 2015) or used independently for criteria aggregation (Greco et al., 2016).

Several auxiliary methods are frequently found supporting the MCDA process. Referred to in this article as “supporting methods,” these include for example the abovementioned FST, the Delphi method, SWOT analysis, GIS tools, and others.

Furthermore, the stated arguments for choosing a MCDA or their supporting methods were collected during the review. These arguments can be compared to the requirements and “quality factors” discussed in the study by Mirakyan and De Guio (2015) for choosing decision methods.

Also of interest regarding the MCDA methods was the combined use of methods and, most importantly, the rationale behind their selection and combination. Strantzali and Aravossis (2016) have recently pointed out the growing trend in combination and comparison of different methods’ results. As mentioned in the section “Introduction,” Løken (2007) and Hobbs and Horn (1997) have advocated using multiple MCDA methods to tackle a similar problem, as its solution may strongly depend on the method choice. Greene et al., 1989; Crump and Logan, 2008; and Mirakyan and De Guio, 2015 have discussed the various purposes and motivations for combining methods. Greene et al. (1989) initially noted five key purposes for mixing methods, which include triangulation, complementarity, development, initiation, and expansion (Table 1). The reviewed studies indeed often involved multiple methods, and the stated or inferred purposes were systematically monitored. As already encountered by Greene et al. (1989), the purpose stated by the authors may differ from the definitions proposed, in which case, for better coherence, the inferred purpose matching the definitions was noted. In several occurrences, several purposes were identified and noted as primary, secondary, and tertiary purpose, by order of importance in the study.