Flexibility and Energy System Models

Long-term energy system models have several inherent modeling limitation to address short- term RE variability. Thereby, these models are weak in evaluating flexibility requirement for different system configurations.

Time Resolution in Energy System Models

Planning horizon of energy system models is often multi-year or decade. The entire time horizon is split into some periods, each consisting of an equal or a different number of years.

A selected set of inter-annual time slices is used to represent a year to limit computational complexity and model size [89]. Thus, within a model period, a year is depicted by only a selected number of seasons, days of week, and time of day [20]. Though it is theoretically possible to represent days in hourly or sub-hourly resolution, chronology of demand or RE time series are not considered by these models. It is also noteworthy that, during model simulation, dynamics of assumptions associated with time slice are not taken into account and are represented only by their annual averages throughout the period.

Figure 2.4 further illustrates the inability of time slice definition to represent RE vari- ability. It shows the effect of temporal aggregation for a system model which considers one representative day in hourly resolution for each month,i.e. 24 time slices per month. The sub-figure on the left indicates hourly capacity factor of a 4 kW PV power plant in New Delhi,

2.3 Power System Flexibility and Planning Models 23

0 20 40 60

0 200 400 600

Hour

Capacity Factor (%)

Hourly capacity factors in June

0 20 40

0 5 10 15 20 25

Timeslice

Capacity Factor (%)

Capacity factors for 24 time slices representating June

Figure 2.4Effect of time slice wise aggregation of hourly solar generation potential for June in Delhi, India (28.63⁰N, 76.94⁰E)

India (28.63⁰N, 76.94⁰E)) throughout the month of June [96]⁴. From the hourly generation values, time slice wise capacity factors are calculated and displayed in the sub-figure on the right side. It is clearly observed that 24 time slices cannot reflect inter-day as well as inter-hour variability of PV generation potential of the whole month. Therefore, short-term RE intermittency from RE generators are not traceable by time slices alone.

Spatial Resolution in Energy System Models

In energy system models, the geographic region under study (country, world,etc.) is often considered as a single copper plate⁵. Therefore, the effect of spatial variation of RE resource (capacity factor,etc.), and inter or intra-regional commodity flow (fuel, electricity, water,etc.) are often not tracked. With multi-region models, region specific RE power plant capacity factors, resource potential and cost of commodity flow between regions can be incorporated into the modeling paradigm. In this approach also, model regions are considered as single nodes and intra-regional variability of RE resources are replaced by averages. Theoretically, spatial resolution can be increased by considering many sub-regions, but computational complexity and data unavailability restrict this. Apart from this, in planning models physical aspects of power flow through electricity grid are often considered as a basic transportation problem. Thus, these models cannot consider network congestion or spatial RE resource variation while calculating inter-regional transmission capacity.

Figure 2.5 illustrates high granular annual average solar global horizontal irradiance (GHI) (kWh/ m²/ Day) data (10 km spatial resolution) corresponding to North Indian states [97]. GHI values are further aggregated to 1⁰ * 1⁰ grid cells, and states respectively in

4PV generation data is taken from PVWatts online calculator. The hourly solar radiation is from the typical meteorological year (TMY) data.

5Single node model

ave_ghi 4.07 - 4.61 4.61 - 4.84 4.84 - 5.00 5.00 - 5.12 5.12 - 5.22 5.22 - 5.33 5.33 - 5.45 5.45 - 5.55 5.55 - 5.64 5.64 - 6.04

Figure 2.5GHI (kWh/ m²/ Day) in North India at 0.1 degree resolution [97]

ave_ghi 4.64 - 4.78 4.78 - 4.96 4.96 - 5.04 5.04 - 5.09 5.09 - 5.15 5.15- 5.23 5.23 - 5.34 5.34 - 5.44 5.44 - 5.58 5.58 - 5.77

(a)Aggregation to 1 degree resolution (Mean)

ave_ghi 4.911 5.015 5.019 5.035 5.040 5.086 5.130 5.137 5.521

(b)Aggregated to States (Mean) Figure 2.6GHI values of Figure 2.5 aggregated to 1 degree resolution grid cells, and states of North India

2.3 Power System Flexibility and Planning Models 25 Figure 2.6. Sub-figure 2.6b clearly illustrates that state level regional definition leads to loss of intra-regional spatial variability of solar generation potential. Unless intra-regional variability is considered at suitable spatial granular level (as illustrated in Figure 2.6a), coarse spatial definition will lead to inaccurate representation of regional RE generation and capacity potential.

Operational Constraints in Energy System Models

Energy system models often use linear programming technique with linearized objective function and constraints. The nonlinear functions are represented as a stepped sequence of linear functions. Though it leads to loss of granularity, it is often a rational choice to keep model within a computational limit. Core focus of energy system models is not to optimize daily generator scheduling or dispatch; rather their strength lies in chronological investment planning. Dispatch of generators in these models is only restricted by a user defined availability factor corresponding to each time slice. System operational constraints on generators, transmission line, etc. are often ignored, or their representation becomes unrealistic due to limited temporal and spatial resolution. Due to the lack of operational constraints, renewable and conventional power plants are treated equivalently (differences are cost, capacity factor, efficiency, and time slice definition,etc.). Therefore, variability from RE power plants is often not traceable. These models do not provide any facility to maintain flexibility or reliability metrics for reliable portfolio calculation [48, 98, 99].

Modeling in absence of operational constraints in low temporal and spatial resolution can lead to overestimation of system capability to assimilate renewable energy, underestimation of operational costs and undervaluation of the requirement for flexible resources [15, 16, 18, 100, 21]. Therefore, it is not guaranteed that the system’s resources planned by these models could be operated in a flexible and reliable manner in real time.