Electricity Market Modelling of Network Investments: Comparison of Zonal and Nodal Approaches

Transcription

1 Electricity Market Modelling of Network Investments: Comparison of Zonal and Nodal Approaches Sadhvi Ganga 1,3 Iain Ferguson MacGill 1,2 School of Electrical Engineering and Telecommunications, University of New South Wales 1 Centre for Energy and Environmental Markets, University of New South Wales 2 TransGrid 3 Sydney, Australia Telephone: s.ganga@student.unsw.edu.au Telephone: i.macgill@unsw.edu.au Abstract Application of electricity market simulation models to forecast long-term market outcomes, amidst considerable uncertainties such as levels of electricity demand, market arrangements and energy policy, has become a widely adopted additional tool for power system engineers and economists in assessment of economically efficient network investment. Modelling of network investments within such models which incorporate linear programming techniques, can require modification of constraint equations which form the bounds of the linear program feasible solution space definition. Formulation of thermal contingency constraint equations can differ between aggregative zonal and explicit nodal modelling approaches. An aggregative zonal approach may include constraint equations which incorporate formulation of static coefficients of market variables calculated exogenously to the electricity market simulation model. In contrast, an explicit nodal approach can enable dynamic endogenous formulation of thermal contingency constraints within the electricity market simulation model. This paper empirically investigates the differences in generation dispatch outcomes between aggregative zonal and explicit nodal modelling approaches for the New South Wales transmission network within the Australian National Electricity Market, between base case and augmented case conditions. The preliminary results indicate that the adopted modelling approach could potentially impact indication of whether a network investment may be economically efficient. Keywords: electricity market, generation dispatch, power system planning, zonal, nodal. 1

2 1 Introduction Economic justification of network investments within restructured electricity industries, by methods involving cost-benefit analysis, can be required to cover the life of the asset, which may span several decades, while assessing its implications for other industry investment and overall economic outcomes [1]. Quantification of market benefits which may arise as a result of the investment, will often involve the application of electricity market simulation models to forecast long-term market outcomes. These models inevitably have significant limitations and involve many underlying assumptions, since the modelling is undertaken in the context of considerable uncertainties such as levels of future electricity demand, market arrangements and energy policy [2]. Nevertheless, the simulated future market outcomes produced by these models may play a critical role in network investment decision-making. In the Australian National Electricity Market (NEM) Regulatory Investment Test for Transmission (RIT-T), market benefits refer to the incremental benefit of a credible option (augmented case) over the base case [1]. Examples of market benefits include generation dispatch cost savings and a reduction of Unserved Energy (USE). The NEM currently operates under a zonal market structure, which involves a commercial simplification of its underlying transmission network, whereby nodes, in theory close in electrical distance, hence with similar price (marginal cost) outcomes, are grouped together to form a smaller number of zones. The NEM currently consists of five zones (New South Wales (NSW), Queensland, South Australia, Victoria and Tasmania). Accordingly, electricity market simulation models of the NEM developed by the Australian Energy Market Operator (AEMO) as part of its preparation of the National Transmission Network Development Plan (NTNDP), have adopted a zonal modelling approach [3]. This approach, however, poses challenges for assessing the market benefits of network investment, both within and between zones. For the 21 NTNDP, some constraint equations incorporating formulation of static coefficients of market variables have been calculated exogenously to the electricity market model by AEMO. An important example of this is the estimation of market variables such as interconnector flows subject to constraints such as thermal contingency requirements. Constraint equations form a fundamental component of the Linear Programming (LP) technique, by defining the bounds of the feasible solution space and ultimately the optimal market dispatch. A network investment (augmented case) has the potential to impact on the magnitude and perhaps the sign of the coefficients of market variables, and extensive, time-consuming, power system analysis studies may be required to discern this impact. In the absence of such power system analysis due to time constraints, the augmented case could be modelled by an increase in the Right Hand Side (RHS) of relevant constraint equations which are impacted by the network investment. By comparison, a nodal electricity market modelling approach represents network nodes explicitly, and can therefore enable dynamic endogenous formulation of thermal contingency constraints. Modelling of a network investment which increases the thermal capacity of a section of network can readily be modelled by increasing the capacity of the relevant links (lines). This empirical investigation examines the differences in generation dispatch outcomes between aggregative zonal and explicit nodal modelling approaches for the NSW transmission network within the NEM (see Figure 1), between base case and augmented case conditions. The key matter of interest is whether the adopted modelling approach could potentially impact indication of whether a network investment may be economically efficient. 2

3 Figure 1: New South Wales transmission network. NOTE: Shows circuits at 22 kv or higher voltage only. 2 Data and Methods A commercially available electricity market simulation software tool, PROPHET, has been used for model development and simulation. PROPHET is a product owned and supported by Intelligent Energy Systems (IES). PROPHET is one of the key tools used by AEMO in its preparation of the NTNDP. The 21 NTNDP dataset and assumptions formed the basis for the NSW Single Node Model (zonal model) and NSW Multi-Node Model (nodal model) development. Some modifications to the 21 NTNDP dataset and assumptions aimed to improve alignment between simulated market outcomes and generator operational limits and historical observations. Hourly market model load traces were developed for each of the five NEM zones. The 211 Jurisdictional Planning Body (JPB) Medium Economic Scenario 1% and 5% Probability of Exceedance (POE) Maximum Demand (MD) and Energy forecasts formed the basis for the load trace development [4] - [8]. Fiscal 21/211 historical data was used as the base year upon which the forecast load traces were developed. Load treatment was on an As generated 1 basis for the zonal model, and on an At Node basis for the nodal model. Diversified load traces at unity power factor were developed for thirty-five NSW High Voltage (HV) Bulk Supply Points (BSP) for the nodal model. The NSW nodal model physical network expansion was limited to meshed HV transmission (5 kv, 33 kv, and 22 kv) and totalled sixty-eight nodes and eighty-two HV transmission lines (or one hundred and six links including transformer modelling) represented by resistance and reactance parameters. The extent of the physical NSW network representation for the zonal model, was a single node in the Sydney West area (the zonal reference node), and the only transmission lines represented were the interconnectors between NSW and the neighbouring interconnected zones of Queensland and Victoria. With respect to intra-zonal transmission constraints, the zonal model maintained representation of N-1 thermal contingency constraints as per the 21 NTNDP definition. These feedback constraints are of the form: 1 Includes generator auxiliary load. 3

4 a 1 x 1 + a 2 x a n x n b (1) where x n is a market variable which is optimised for dispatch, a n is the coefficient of the market variable x n, and b is a pre-calculated constant value, comprising the pre-dispatch values of several components which define the limit. Market variables which can be optimised for dispatch include the output of scheduled generation units and interconnector power flows, and appear on the Left Hand Side (LHS) of (1). The RHS of (1) comprises elements that define the limit for a particular system condition, such as line ratings and on/off status of generation plant. For some constraint equations, formulation of static coefficients a n by AEMO, involved calculation exogenous to the 21 NTNDP PROPHET market model. For the nodal model, the intra-zonal NSW N-1 thermal contingency constraints contained within the 21 NTNDP dataset were removed. Instead, the N-1 thermal contingency constraint definition was dynamically endogenously formulated by the PROPHET N Minus One module for the given NSW intra-zonal network topology, and preand post- contingency line ratings. These constraints are of the form [9]: line i flow + shift factor ij line j flow line i limit (2) where shift factor ij is the proportion of the power flow on line j which is transferred to line i when line j fails. Thus, the formulation of N-1 thermal contingency constraints for the zonal and nodal models are inherently different. Consequently, the LP feasible solution space definition between the two modelling approaches are not identical. The modelled NSW 3 MW intra-zonal augmentation aimed to increase the thermal capacity of the transmission corridor. Modelling of this augmentation for the zonal approach, involved increasing the RHS of relevant Nil outage, 15-minute and N-1 thermal constraint equations only. For the nodal approach, modelling of this augmentation involved increasing the capacity of links being augmented (thereby impacting the dynamic formulation of the N-1 thermal contingency constraints), in addition to increasing the RHS of relevant Nil outage and 15-minute thermal constraint equations. The modelled network investment comprised of increase in the thermal capacity of four links located in the southern area of NSW, commissioned in 214. Time sequential security constrained economic dispatch (SCED) was simulated for both the zonal and nodal models. 3 Results Generation dispatch outcomes between the zonal and nodal modelling approaches varied significantly. The difference in total annual energy generation between the augmented case and the base case for each of the five NEM zones are shown in Figure 2 to Figure 6 (zonal model results are shown on the left and nodal model results are shown on the right) 2. Data in these figures are sorted according to plant fuel type; positive numbers indicate an increase in total annual energy generation post augmentation, while negative numbers indicate a decrease in total annual energy generation post augmentation. 2 Note, in these figures: CCGT = Combined Cycle Gas Turbine OCGT = Open Cycle Gas Turbine. 4

5 New South Wales New South Wales Existing - Black Coal New Plant - Solar Existing - Black Coal Figure 2: Change in Total New South Wales Plant Annual Energy Generation: Augmented Case Base Case. Queensland Queensland Existing - Black Coal Existing - Black Coal Figure 3: Change in Total Queensland Plant Annual Energy Generation: Augmented Case Base Case. South Australia South Australia New Plant - Geothermal Existing - Wind Existing - Brown Coal New Plant - Geothermal Existing - Wind Existing - Brown Coal Figure 4: Change in Total South Australia Plant Annual Energy Generation: Augmented Case Base Case. 5

6 Tasmania Tasmania Existing - Natural gas Figure 5: Change in Total Tasmania Plant Annual Energy Generation: Augmented Case Base Case. 4 Victoria 4 Victoria New Plant - Wind New Plant - Solar New Plant - Geothermal Existing - Wind Existing - Brown Coal New Plant - Wind New Plant - Solar New Plant - Geothermal Existing - Wind Existing - Brown Coal Figure 6: Change in Total Victoria Plant Annual Energy Generation: Augmented Case Base Case. The zonal and nodal modelling approaches both produced a large increase in Existing Black Coal energy generation in NSW for almost all years of the simulation period, post augmentation in 214. However, the nodal model produced a large decrease in Existing Natural Gas energy generation in NSW, approximately offset in magnitude by the increase in Existing Black energy production within the State. For the nodal model, relatively high production cost peaking Existing Natural Gas plant, located close in electrical distance to the area of augmentation, produced less energy subsequent to the network investment. This outcome for the nodal model suggests that, this Existing Natural Gas plant may have been constrained-on prior to the augmentation. Another interesting observation of this outcome produced by the nodal model, is the apparent fuel substitution between Existing Black Coal and Existing Natural Gas plant post NSW intra-zonal augmentation, appears to be mainly contained within the borders of NSW. The change in total annual energy production of plant in Queensland produced by the zonal and nodal models were also different. The result for the nodal model, shown in Figure 3, supports the hypothesis of fuel substitution, post NSW intra-zonal augmentation, being mainly contained within the borders of NSW. This is further supported by the reduction in power flow from Queensland to NSW over the inter-zonal Queensland New South Wales Interconnector (QNI) produced by the nodal model, compared to the results of the zonal model. Figure 7 illustrates this reduction in southwards power flow over QNI produced by the nodal model, post augmentation, for years 214, 22 and

7 QLD to NSW 8 6 QLD to NSW 8 6 QNI Power Flow (MW) Zonal Nodal QNI Power Flow (MW) Zonal Nodal -2-2 NSW to QLD -4 NSW to QLD -4-6 Proportion of Time Flow Exceeded (%) -6 Proportion of Time Flow Exceeded (%) 12 1 QLD to NSW 8 6 QNI Power Flow (MW) Zonal Nodal -2 NSW to QLD -4-6 Proportion of Time Flow Exceeded (c) Figure 7: Modelled Queensland New South Wales Interconnector Power Flow Cumulative Frequency Distribution Post Augmentation for the years: (c) 225. A salient observation of the change in total annual energy of plant in the southern NEM zones, is the significant increase in wind generation in South Australia (see Figure 4) produced by the nodal model compared to the zonal model results. Although South Australia and Tasmania are not directly interconnected with NSW, the differences in the change in total annual energy generation between the zonal and nodal modelling approaches post augmentation is also significant (see Figure 4 and Figure 5). This highlights that the nodal modelling approach for NSW does not result in a change in simulated optimal dispatch outcomes confined only to NSW and the neighbouring interconnected zones of Queensland and Victoria, but impacts the other NEM zones as well. 7

8 Total annual energy generation of plant, post augmentation, have varied substantially between the zonal and nodal modelling approaches. NEM generation dispatch costs are consequently impacted. Table 1 and Table 2 list the total NEM dispatch costs for the zonal and nodal models respectively. Total NEM dispatch costs are lower in magnitude for the zonal model as compared with the nodal model for base case and augmented case conditions. However, for the zonal model, generation dispatch costs were counter-intuitively higher in the augmented case relative to the base case, indicating negative market benefits for the network investment. In contrast, generation dispatch costs in the augmented case were lower relative to the base case for the nodal model, indicating positive market benefits. While still preliminary, these results do demonstrate that the LP feasible solution space definition and the approach of modelling augmentations within the electricity market model can potentially have impact on indication of whether a network investment may be economically efficient Base Case Augmentation Case Change in Dispatch Cost (Base Case - Augmentation Case) Table 1: Total National Electricity Market Dispatch Costs ($ million): Zonal Model Base Case Augmentation Case Change in Dispatch Cost (Base Case - Augmentation Case) Table 2: Total National Electricity Market Dispatch Costs ($ million): Nodal Model. Conclusions These results apply to modelling undertaken with PROPHET. However, the intent is to highlight the impact LP feasible solution space definition and this approach to modelling augmentations within electricity market simulation models, may have on optimal dispatch outcomes. This may well apply to other electricity market simulation software tools as well. This empirical investigation focussed on treatment of line thermal limitations and modelling approach between the base case and the augmentation case. Future work may also consider treatment of other power system constraints, such as voltage, transient and oscillatory stability. The results demonstrate that different decision paths for progression of the network investment assessment can potentially be taken due to the modelling approach. While the nodal model indicates potential benefits to the market as a result of the network investment, the zonal model indicates the opposite. The adopted modelling approach could potentially impact on network investment decision-making, and therefore warrants due consideration. Acknowledgements The authors gratefully acknowledge the support of TransGrid, and thank Enrico Garcia and Can Van. 8

9 References [1] Australian Energy Regulator, Regulatory investment test for transmission (RIT-T) and application guidelines 21. [2] A.M. Foley, B.P. O Gallachoir, J. Hur, R. Baldick and E.J. McKeogh, A strategic review of electricity systems models, ELSEVIER Energy, vol. 35, pp , 21. [3] Australian Energy Market Operator, 21 National Transmission Network Development Plan. [4] TransGrid, Annual Planning Report 211. [5] Powerlink, Annual Planning Report 211. [6] Australian Energy Market Operator, 211 Victorian Annual Planning Report. [7] Australian Energy Market Operator, 211 South Australian Supply and Demand Outlook. [8] Transend, Annual Planning Report 211. [9] Intelligent Energy Systems, PROPHET User Guide, vol. 1, p. 388, April