in

A robust multiple-objective decision-making paradigm based on the water–energy–food security nexus under changing climate uncertainties

As stated, the primary goal of this study is to promote an objective decision support framework for water resource planning and management purposes within the context of the WEF security nexus, which takes into account the uncertainties imposed by the climate change phenomenon. Such a framework is “robust” since it takes the multi-dimensionality of water-related problems into account while addressing the uncertainties imposed by climate change projections. The basic components of this decision-making paradigm are depicted in Fig. 1. In principle, while this framework is sensitive to the uncertainties associated with the climate change projections, it can provide a dynamic water resources planning and management scheme promoted within the WEF security network. Thus, in addition to the status quo, a series of climate change projections (i.e., RCP 2.6, RCP 4.5, and RCP 8.5) are also integrated into the proposed decision support framework. In essence, the main components of the proposed framework are simulation and operation of the water resources system based on the standard operation policy (SOP), evaluating the system’s efficiency through a series of quantitative performance criteria, and finally, applying the MADM-based framework to opt for a robust system renovation setting.

Figure 1

Basic components of the robust decision-making paradigm for water resources planning and management.

Full size image

Simulating the water resources system

SOP is a primitive, and perhaps the most-well-known real-time operation policy in water resources planning and management14. The core principle here is to minimize the prioritized water shortage at the current time step with no conservation policy (e.g., hedging rules) in place. SOP, as a standard rule curve (RC), determines how the operator should behave at any given state of a reservoir15,16. This rule curve is established as an attempt to balance various water demands including but not limited to flood control, hydropower, water supply, and recreation17. A SOP operating system attempts to release water to meet a water demand at the current time, with no regard to the future.

In general, SOP can be mathematically expressed as18:

$$R_{t} = left{ {begin{array}{*{20}c} {D_{t} } {AW_{t} } 0 end{array} – S_{min } } right.begin{array}{*{20}c} {} & {if} & {AW_{t} > S_{min } } {} & {if} & {AW_{t} > S_{min } } {} & {if} & {AW_{t} le S_{min } } end{array} begin{array}{*{20}c} {} & {and} & {AW_{t} – S_{min } ge D_{t} } {} & {and} & {AW_{t} – S_{min } < D_{t} } {} & {} & {} end{array} quad t = { 1},{ 2},{ 3}, , … , ,T$$

(1a)

where

$$AW_{t} = S_{t} + Q_{t} – Loss_{t}$$

(1b)

in which Rt = amount of water supplied during the tth time step; Dt = consumers’ water demand during the tth time step; AWt = amount of available water during the tth time step; St = amount of stored water during the tth time step; Smin = dead storage of the reservoir; Qt = inflow during the tth time step; Losst = net water loss (i.e., precipitation minus evaporation) of the reservoir during the tth time step; and T = total number of time steps in the operational horizon.

In practice, however, a different type of water demand leads to a different interpretation of water shortage. There are cases in which the stakeholders’ needs are represented by a set of volumetric demand targets, and the decision-makers’ objective would be to minimize the water deficit based on a set of priorities for these demands. This is a typical case for agricultural, domestic, industrial, and environmental demands. For hydropower generation, however, a conventional interpretation of SOP would be to generate maximum electricity permitted by the power plant capacity (PPC) at each given time step19. For a hydropower system, the amount of water needed to reach a power plant capacity is given by19:

$$R_{t} = frac{{86400 times PF times Countday_{t} times PPC}}{{gamma_{w} times g times eta times Delta H_{t} }}$$

(2)

in which, γw = water specific weight; g = gravitational acceleration; η = efficiency of the hydropower system; ΔHt = height difference between the reservoir water level and the tailwater level at time step t; Countdayt = number of days within time step t; and PF = plant factor of the hydropower system.

As stated earlier, applying an SOP-based plan requires a set of pre-defined priorities to advise decision-makers concerning the order, in which each of these demands is to be met. The major water demands include drinking, industry, environment, agriculture, and hydropower. Thus, according to the SOP’s principle, the decision-makers, first, allocate the available water to meet the demand of the stakeholder with the highest priority (i.e., the domestic and industrial demand). After this first water demand is fully satisfied, the available water can be used for the next demand. Such an allocation process continues until no water is available. It should be noted, however, that if the released water in each stage passes through the penstock equipped with the turbines, electricity can be generated. The amount of energy generated in previous stages must be accounted for before computing the amount of water released for hydropower purposes.

Performance criteria

Performance criteria are, in essence, quantitative measures that can provide a practical insight for the decision-makers regarding the status of a system. This definition covers a broad spectrum of mathematical representations, which can range from simple mathematical formulas such as the average of a specific output to more complex and probability-based entities20,21. The most fundamental and universal probability-based performance criteria are reliability, resiliency, and vulnerability22,23,24. In essence, reliability is the probability of successful function of a system; resiliency measures the probability of successful functioning following a system failure; and vulnerability quantifies the severity of failure during an operation horizon25. It should be noted that these three criteria assess different aspects of a water resources system, and as such, they complement one another26. For more information regarding these probabilistic performance criteria, the readers can refer to Sandoval-Solis et al.27 and Zolghadr-Asli et al.20.

In this study, the concept of levelized cost of energy (LCOE) is utilized for economic evaluation. The LCOE of a given hydropower system is the ratio of lifetime costs to lifetime electricity generation, both of which are discounted back to a common year using a discount rate that reflects the average cost of capital28. The LCOE of renewable energy systems depends on the technology, geographic criteria, capital and operating costs, and the efficiency of the system. The LCOE can be mathematically expressed as follows29:

$$LCOE = frac{{sumnolimits_{t = 1}^{n} {frac{{I_{t} + M_{t} + F_{t} }}{{left( {1 + r} right)^{t} }}} }}{{sumnolimits_{t = 1}^{n} {frac{{E_{t} }}{{left( {1 + r} right)^{t} }}} }}$$

(3)

in which It = investment expenditures in year t; Mt = operation and maintenance expenditures in year t; Ft = fuel expenditures in year t; Et = electricity generation in year t; r = discount rate; and n = economic life expectancy of the system.

MADM

MADM is an umbrella term to describe a series of frameworks, which aim to help individuals or a group of individuals to prioritize a series of discretely defined alternatives with regard to a set of evaluation attributes30,31. MADM can provide the necessary means to conduct planning and management under changing circumstances such as those under climate change conditions10,32. According to one of the basic principles of MADM, the decision-maker can use the similarity of the feasible alternatives and the preferential result and/or incongruity of the undesirable alternatives. The notion mentioned above is, chiefly, the core principle of the reference-dependent theory33. Accordingly, the reference-based branch of the MADM methods can, itself, be classified into two major groups: screening methods and ranking methods. Screening methods eliminate alternatives that cannot satisfy the pre-determined conditions for the desirable solution, while ranking methods order all the alternatives from the best to the worst34.

Pioneered by Hwang and Yoon35, the technique for order references by similarity to an ideal solution (TOPSIS) is a compensatory, objective MADM solving method rooted from the basic principles of the reference-dependent theory. The core idea is that the chosen alternative should have the shortest distance from the ideal solution and the farthest distance from the negative-ideal solution36. The basic computation algorithm of TOPSIS can be summarized as follows37,38:

Step I: Construct the original decision matrix (X), where m feasible alternatives are to be evaluated based on n evaluation criteria:

$$X = left[ {begin{array}{*{20}c} {x_{11} } & {x_{12} } & cdots & {x_{1n} } {x_{21} } & {x_{22} } & cdots & {x_{2n} } vdots & vdots & ddots & vdots {x_{m1} } & {x_{m2} } & cdots & {x_{mn} } end{array} } right]$$

(4)

in which xij = the element of the ith alternative concerning the jth criterion.

Step II: Defining the reference alternatives [i.e., the ideal solution (s+) and the negative-ideal solution (s)]. To do so, first, the elements of the decision matrix that are associated with negative criteria must be redefined by using the following equation:

$$x_{ij}^{ * } = frac{1}{{x_{ij} }}$$

(5a)

The elements of the decision matrix that are associated with positive criteria would remain the same:

$$x_{ij}^{ * } = x_{ij}$$

(5b)

The ideal alternative is an arbitrarily defined vector, which describes the aspired solution to the given problem, while the inferior alternative is an arbitrarily defined solution that represents the most undesirable option for the given MADM problem. Here, the ideal and negative-ideal solutions would be represented with two separate vectors where each pair of the corresponding elements in these vectors is, respectively, the maximum and minimum values of (x_{ij}^{ * }) with regard to each of the evaluation criteria.

Step III: Each element of the decision matrix should be normalized by using the following equation:

$$p_{ij} = frac{{x_{ij}^{ * } }}{{sqrt {sumnolimits_{i = 1}^{m} {x_{ij}^{ * 2} } } }}$$

(6)

in which pij = the normalized performance value for the ith alternative with respect to the jth criterion.

Step IV: The weighted normalized preference value (zij) can be computed as follows:

$$z_{ij} = p_{ij} times w_{j} quad forall i,j$$

(7)

in which wj = the weight (i.e., the importance value) of the jth criterion. The weights assigned to the evolution criteria reflect their relative importance to the decision-makers. The higher the weights are, the more crucial their roles would be in the selection process. Chiefly, these weighting mechanisms are either subjective in nature or follow an objective procedure. In the subjective approaches, the weights of the attributes are assigned based on the performance information given by the decision-maker, whereas in the objective approaches, the weights of the evaluation attributes would be obtained by using the objective information extracted from the decision matrix39. Shannon’s Entropy method, used in this study as the weight assignment mechanism, is a well-known objective weighting technique40. This method tends to assign the highest weight to an evaluation attribute with the highest dispersity in its values. For more information on the computational framework of this method, the readers can refer to Lotfi and Fallahnejad41.

Step V: In this step, every given alternative is compared to the reference points, namely, the ideal and inferior alternatives. The described procedure, which is known as the separation measurement in TOPSIS, can be mathematically expressed as follows35:

$$D_{i}^{ + } = sqrt {sumlimits_{j = 1}^{n} {left( {z_{ij} – z_{j}^{ + } } right)^{2} } }$$

(8)

And

$$D_{i}^{ – } = sqrt {sumlimits_{j = 1}^{n} {left( {z_{ij} – z_{j}^{ – } } right)^{2} } }$$

(9)

in which (D_{j}^{ + }) and (D_{j}^{ – }) = separation measurements of the jth criterion with respect to the ideal and inferior alternatives, respectively.

Step VI: The relative closeness to the ideal solution (χi), which can be used to rank the desirability of the feasible alternative, can be computed as follows35:

$$chi_{i} = frac{{D_{i}^{ – } }}{{D_{i}^{ + } + D_{i}^{ – } }}quad forall i$$

(10)

The further this distance (i.e., larger values of χi), the more desirable the alternative would be.

Robust multi-attribute framework

As stated, each climate change scenario depicts a unique future with regard to the changing climate, which in turn introduces an element of uncertainty to the projected performance of water resources systems during their operation horizon. Furthermore, downscaling methods, which link these projected changes in the global climatic pattern to a local or regional scale, can be another source of uncertainty. Naturally, for long-lasting water infrastructure such as a hydropower system, addressing these uncertainties in a proper and timely manner can be one of the key components of a robust project. Thus, this study aims to not only evaluate the system’s performance under the status quo but also assess the credibility of the system under the projected climate change conditions.

The other characteristic one might expect from a robust project is its ability to take into account the multi-dimensionality nature of water-related infrastructure. Most notably, addressing the WEF security nexus must be a priority in water resources planning and management. Resultantly, any robust decision-making paradigm for water resources planning and management purposes should also account for the other pillars of the WEF nexus (i.e., energy and food sectors), as they would be consequentially affected by such decisions. It is also important to note that these sectors could be affected by the climate change phenomenon. The other crucial feature of a robust decision-making paradigm is that it should be able to account for the socio-economic, environmental, and technical factors that determine the overall quality of the project. Such a decision-making paradigm is depicted in Fig. 2. This notion in practice, however, can typically lead to a mega decision matrix composed of numerous criteria and alternatives that can be overwhelming if the subjective MADM methods are to be employed. This study, thus, employs an objective MADM framework (i.e., TOPSIS/Entropy) to help overcome the above-described problem. The basic idea is to promote a universal and practical decision support framework that enables the water resources planners and managers to account for the intricacies of the WEF security nexus while simultaneously taking the uncertainties of climate change projections into account. Figure 3 illustrates the flowchart of the proposed decision support framework.

Figure 2

Schematic diagram of the MADM problem.

Full size image
Figure 3

Flowchart of the proposed framework.

Full size image


Source: Ecology - nature.com

Coupling power and hydrogen sector pathways to benefit decarbonization

Saving seaweed with machine learning