These features are significant as they all result in increasing scarcity of a resource which is crucial for the operation of electricity systems, given the need to balance the electricity system second-by-second (i.e., flexibility¹). Increasing penetration of low carbon electricity generators will reduce flexibility as they are either nonschedulable as well as partly nondispatchable [e.g., in the case of solar Photovoltaic (PV) and wind generation], or are (traditionally) not designed for frequent and fast modulation (e.g., nuclear). Meanwhile, increasing penetration of large-scale (variable) RES, will increase demand for flexibility at the system level, just as the availability of traditional providers of flexibility (thermal generation) is decreasing. At the same time, increasing penetration of small-scale RES as well as of electric heating and transport is increasing demand for flexible resources at the distribution network level, to prevent voltage and congestion issues (Navarro-Espinosa & Ochoa, 2015; Navarro-Espinosa & Mancarella, 2014; Mancarella, Gan, & Strbac, 2011), which can motivate expensive network reinforcement (Martínez-Ceseña, Good, & Mancarella, 2015). While there are various types of flexible resources available, flexibility from the demand-side, from entities, such as neighborhoods, have been identified as particularly attractive sources (Strbac et al., 2012). Hence, there is clearly motivation for neighborhoods to become, as a primary objective, flexible, rather than energy positive (besides the ambiguity of the terminology itself).

Naturally, given the variable and stochastic nature of demand for energy services (e.g., heating or air conditioning), RES, and plant and networks outages, the demand for operation of flexible resources is variable and dynamic. Hence, to achieve flexibility (rather than energy positivity, in the classic sense) focus should be on strategic adoption of flexible resources and enabling ICT. This motivates methods of neighborhood/district assessment which fundamentally reflect this focus on flexibility. This requirement is, however, not straightforward. The value of flexibility varies substantially, both temporally and spatially, but also according to the preferences of the evaluator. For example, at times of low system demand, flexibility may be important especially in cases when there is little inertia in the system (as it may be in the presence of large volume of power electronics-connected RES). On the other hand, for periods of high system demand, flexibility may have substantial value as the system is more sensitive to shocks due to the increased penetration of inflexible technologies. Similarly, on congested distribution networks with little spare capacity the value of flexibility may be significant to avoid network reinforcement. Finally, whatever the physical status of the electricity system and networks, the value of flexibility may be different to different parties. For example, a consumer with a requirement for high reliability in their electricity supply (e.g., a hospital), may value reliability highly. In contrast, a relatively poor (hence price sensitive) domestic customer may be willing to accept a lower level of electricity supply reliability if this provides a discount on their bill. Similarly, users with an alternative energy-service supply (e.g., back-up electricity generation or an alternative supply of fuel/alternative energy service providing device) may be willing to accept lower electricity supply reliability. In this context, reliability [and to some extent even resilience (Panteli & Mancarella, 2015)] could be provided by flexible neighborhood-level demand-side resources that might respond to economic incentives to decrease their electricity consumption (or increase their production in the case of embedded generation).

Accounting for these varying, dynamic conditions and preferences is not an easy task for one central assessor, making assessment of flexibility difficult. However, in a liberalized system, where (generally variable and dynamic) price signals indicate the value of agent behavior, economic value can be considered the most suitable proxy metric to measure flexibility (Chapter 6). Although it is not a perfect measure, as it relies on well-functioning (i.e., cost-reflective) markets, which may not always be the case, it is the best available. Further its efficacy as a measure of flexibility should only increase as technology increases the penetration of variable, dynamic pricing (Faruqui, Harris, & Hledik, 2010), and reforms (such as including the price of disconnections and voltage reduction in imbalance penalties, and making the penalties reflect the marginal cost of balancing actions) increase the cost reflectiveness of prices (Flamm & Scott, 2014).

Given the identification of economic value as the appropriate proxy measure of the flexibility of a neighborhood, Section 2 examines how the economic value metric of neighborhoods varies given objectives of “classic” energy positivity and flexibility. Quantitative examples demonstrate the material difference in economic value which results from each objective. Section 3 then considers, through further quantitative examples, how optimization of a neighborhood with respect to various objectives which are commonly conflated (i.e., economic value, CO₂ emission, and self-sufficiency) are, in fact, not equivalent. Recognizing that decision-makers may want to evaluate neighborhoods according to many criteria, rather than solely according to flexibility, Section 4 introduces the concept of multicriteria analysis, outlining a framework for assessment of interventions according to multiple criteria. Section 5 offers some concluding remarks.

However, it is important to note that while the aforementioned objectives may motivate economically valuable behavior, given the general coincidence of economic and low-carbon drivers (i.e., high CO₂ generation is often expensive), and of economic and local energy drivers (as on-site generation avoids UoS costs and can avoid some elements of taxation), pursuing a noneconomic objective may result in a suboptimal economic value for the EPN.

To illustrate this, consider as an example that the operation of neighborhood 2 is optimized for a typical winter day with respect to three different criteria, which could be the criteria of an employed energy optimization engine (such as that explored in Chapter 4):

The neighborhoods are optimized for a typical winter day, with energy price and grid CO₂ emission rates (based on the UK context) as in Fig. 2.10.

For each case, the relevant CO₂ emission (including both local and grid emissions), the amount of energy traded with the grid (as a measure of neighborhood self-sufficiency) and the total energy cost are quantified. The results for the three studies are shown in Table 2.1.

Results in Table 2.1 clearly highlight that while some optimization criteria may produce results which are desirable with respect to other metrics, a result will be generally suboptimal with respect to a given criteria unless it is optimized specifically with respect to that criteria.

More detail is shown in Figs.  2.11–2.13, which show CHP electricity generation, local and grid CO₂ emission, and energy cost in the cases of optimization with respect to economic value, CO₂ emission, and self-sufficiency, for neighborhood 2, respectively. As shown through comparison of Figs. 2.11 and 2.12 the profiles for CO₂ minimization and self-sufficiency maximization are quite similar, as both the CO₂ minimization and self-sufficiency maximization objectives are served by minimizing imports from the grid. The slight difference arises as CO₂ emissions are minimized by directing grid consumption to a later period, where grid emission rates are lower. In contrast, the grid is used more often when maximizing economic value, as although there are penalties associated with grid usage (as UoS fees and taxes are charged on import, but not accrued as profits during export), these do not produce as high an incentive as in the previous cases to avoid electricity imports (and hence export, during other periods). The relationship between electricity prices and grid emission rates means that operating the neighborhood based on optimal economic gain also produces reasonably low emissions (as the CHP is motivated to operate at high price times, reducing the neighborhood grid consumption and hence grid emissions). It is worth noting that a neighborhood primarily heated by EHP would have a different result, as correlation of high grid emission rates and low prices would motivate consumption at times of high grid emissions.

A multicriteria CBA is meant to compare different intervention options based on relevant criteria. In the context of this book, the interventions may represent a vision of the EPN (e.g., investment in particular infrastructure and provision of a specific portfolio of services) from the perspective of an actor (e.g., consumers, retailers, and so forth) under a given market framework. For this purpose, it is convenient to present the performance of the interventions according to different criteria in the form of a matrix (performance matrix) as shown in Table 2.2. In the examples later, performance criteria include Net Present Value (NPV), payback time, Internal Rate of Return (IRR), and CO₂ emissions.⁸ The performance of each intervention option has, for illustration purposes, been set arbitrarily and, for the sake of simplicity, each criterion is expressed as either attractive or unattractive. In practice, each particular criterion would typically be assigned a numerical value.

The basic technique (and thus first step) to perform the multicriteria analysis is direct analysis. Direct analysis consists of inspecting if any of the options performs as well or better than all the other options according to each criterion. In this example, this would be option A, which is attractive according to all criteria under consideration. That is, this option offers high benefits (high NPV) on the short term (low payback times), high premium for the capital invested (high IRR) and low environmental impact (low CO₂ emissions), all of which are attractive for the corresponding criteria. In practice, there might not be an option that meets the direct analysis requirements, thus other multicriteria techniques have to be used.

If it can be reasonably assumed that all criteria are independent of each other (e.g., the perception of the NPV will not change regardless of the payback time) and can be assigned a weight, the linear additive model may be applicable for the analysis. This technique consists of assigning a weight to each criterion and multiplying the score of each project by the weight. This would produce a single criterion for each intervention option based on which the best alternative can be determined. Such an approach can also be used to form a multicriteria objective, in the operational domain. Such an objective could be used to weight the economic value, CO₂ and self-sufficiency objectives discussed in Section 3. In order to illustrate this, assume that each option is credited 1 point per criterion if its performance is attractive. Accordingly, option B is worth 3 points (for its performance according to the payback time, IRR, and CO₂ criteria) and is deemed a better alternative than option C, which would only be credited 2 points (for its performance according to the NPV and payback time criteria), while option A would be deemed the best option given its 4 points (for its performance according to all criteria). Clearly the weight assigned to each criterion is a key parameter, as it has significant impact on the outcome of this technique.

Now, if it is not reasonable to assume that all criteria are independent of each other (e.g., a high NPV is desired but short payback times are preferred) an analytical hierarchy process could be used. In this case a pairwise comparison approach (based on a particular analysis or judgment) is used to determine the proper weights for the different criteria and intervention options. In order to illustrate this process, consider the pairwise weights shown in Table 2.3. Table 2.3, which illustrates how a given criterion (row) is valued with respect to other criteria (column) (e.g., the NPV criterion is deemed twice as valuable as the IRR criterion and thrice more valuable than the CO₂ emissions criterion).

The weights for each criterion associated with the example of pairwise comparison are presented in Table 2.4. As it can be seen, the payback time is deemed the most valuable criterion, followed by the NPV, IRR, and CO₂ emissions. An explanation of the mathematics required to process this matrix to obtain the weight for each criterion is beyond the scope of this chapter. It will only be mentioned that the weights can be determined as the eigenvector associated with the maximum eigenvalue of the matrix (for more information, see Weisstein, 2016).

Following the weighting procedure undertaken previously, a similar procedure is performed for the assessment of the intervention options. For the sake of simplicity, option A is not considered in this weighting procedure (as it would be deemed the best in this analysis due to its attractive performance according to all criteria) and an investment option is deemed twice as valuable as another if its performance under a given criterion is better according to Table 2.2. The results of such a pairwise comparison and the associated weights are shown in Tables 2.5 and 2.6, respectively.

Finally, the weights for both criteria and intervention options are combined (i.e., the sum of the weight of the option multiplied by the weight of the corresponding criterion). The results show that option B is marginally a better option than option C under the selected criteria and weights (Table 2.7).

In cases where district assessors are interested in multiple assessment criteria, such as energy positivity and flexibility, but extensible to any other criteria (such as CO₂ emissions and economic value, as discussed in Section 3), the methods presented here are clearly useful, in the EPN planning phase. Further, the linear additive model (in which criteria are assigned weights equaling unity) may be employed for operational optimization. This may enable simultaneous consideration of multiple criteria (such as economic value, CO₂ emissions, and self-sufficiency, as in Section 3). In all cases, assessors must determine the weights that should be assigned to various criteria. Assuming districts are primarily concerned with minimizing costs, most weight is likely to be given to the economic objective. Thus, if price signals can be considered an effective indicator of the value of flexibility, economic value is demonstrated to be a suitable proxy for flexibility (Section 1).

Through development and explanation of the concept of flexibility, comparison to the inflexible, static, classical definition of energy positivity, and through quantitative studies, this chapter demonstrates the suitability of flexibility, and, indeed, the inadequacy of the concept of energy positivity, especially when ill-defined and focused on the electrical side only, as measures of desirable behavior in current and future electricity systems. However, in a power system context, the key message is that the value of energy services and electricity supply reliability varies by individual, and that the value of demand and generation, and hence system and network redundancy, vary in time and space. Hence, neighborhoods should be optimized and assessed using objectives and metrics which reflect this. Naturally, this requires suitable models and optimization algorithms (as detailed in Chapter 4) and suitably advanced ICT for monitoring, communication and to undertake complex optimization (Chapter 5). Another key message is that, in the absence of any other practical objective/metric, economic value should, in a liberalized electricity system, be used as a proxy for flexibility. Further, this objective/metric can be made a better proxy through improving the cost reflectivity of prices. In this context, economic value should be assessed through quantification of the district business case, which is discussed in detail in Chapter 6.