Reports

An Analysis of Alternative Tariff Systems for Taiwan

An abstract from an extensive study

commissioned by Taiwan Power Company

The UPLAN modeling system is used to simulate the Taiwan power system in order to compare alternative grid pricing and tariff methods.

The crucial features critical to the Tariff Analysis are UPLAN’s ability to simulate a wide range of scenarios regarding restructuring, market design, market uncertainties, and business and investment strategies. The key models are the Volatility Model and the Merchant Plant Model that are linked to (and rely on outputs from) the OPF and Multi-Market models. The model uses a specially developed database that captures the technical parameters of generator and transmission elements that constitute the Taiwan network. Also factored in are financial data on these assets, as well as uncertain market drivers such as fuel prices, load growth and hydro conditions.

The principal UPLAN reports used for the analyses (2002 through 2006):  

• power flows on different grid elements;
• hourly grid-wide and location-specific electricity prices (or marginal costs), transmission constraints (congestion) and costs for resolving constraints; 
• projected system thermal losses and related costs; 
• hourly prices and costs for certain ancillary services based on simulated generator bidding with arbitrage among the different markets; and
• financially rational generator additions (not announced) for longer-term simulations to evaluate potential stranded costs for existing generators.

Embedded Costs - Five different methods were used to determine tariffs to recover embedded transmission ownership costs. Case 1 applies grid-wide $/MWh postage stamp charge to loads based on their MWh of load, and grid administration costs are recovered by a second $/MWh charge to loads (also in case 3). Case 2 uses a $/kW “license plate” charge to loads in each of four zones based on annual peak kW of load. Grid administration costs recovered on a system-wide basis. In Case 3, recover 50% from loads on a system-wide basis, based on average monthly peak load over the four summer months June-September, the other 50% of transmission ownership costs are recovered from generators based on their kW of capacity. In Case 4 a “line usage” method., based on the change in power flow over the element calculated to be caused by the particular user, divided by the overall power flow on that element due to the particular user plus all other users combined. “Users” are defined as seller-buyer (generator-load) combinations, rather than as loads alone. Four alternative methods are used to calculate power flow changes: modulus (absolute flow change), zero counterflow, dominant flow, and net flow. Case 5. Like Case 4, except that (1) “users” are defined in terms of load location and magnitude but not as specific seller-buyer (generator-load) combinations, (2) a particular user’s fractional share of a given grid element’s cost is calculated as the user’s marginal impact on power flows over that element, and (3) power flow impacts are calculated using shift factors. Grid administration charges are the same as in Cases 2 and 4 and 5.

Grid Administration Cost Recovery - These are assumed recoverable on a grid-wide (not zonal) basis, using a per-MWh or per-peak kW postage stamp charge. Total “Access” Charges - A user’s total “access” charge includes the charges to recover costs of transmission and distribution ownership plus grid administration. 

In summary, “Access Tariff” Cases 1 through 3 use straightforward license plate allocation of transmission ownership costs and give the most uniform and predictable overall rates. The zonal transmission charges in Case 2 may send desirable signals regarding zonal costs, but may also excessively penalize users in zones with transmission assets that substantially benefit users in other zones. Cases 4 and 5 use line usage methods to allocate transmission ownership costs, resulting in highly varied charge rates among users. Because it is based on load locations rather than generator-load combinations, the line usage method used in Case 5 is more computationally feasible than the method used in Case 4, produces charge rates that are somewhat more uniform and predictable, and is guaranteed to produce charges that sum to exactly 100% of costs being allocated.

Under projected “most likely” conditions for 2002-2006, no congestion (constraints on transmission interfaces) was simulated to occur for the Taiwan grid. Several likely conditions have been examined. For purposes of illustrating congestion costs and charges, a scenario was constructed based on projected year 2002 conditions, but with two nuclear units and one fossil unit in the northern zone (zone 1) assumed to be on outage during the peak load months of July and August. Combined with the double (two-line) transmission outage contingency incorporated into the simulated total transfer capacity (TTC) for transmission links between the different zones, this represents a very low probability combination of unfavorable circumstances.

The first of three methods illustrated for charging users for congestion is based on zonal electricity prices. Under the simulated congestion scenario, zonal electricity prices were the same across the different zones during 1,412 July/August hours, but in the remaining 76 hours the zonal prices were substantially higher in the northern zone (zone 1) due to transmission constraints that created a need to re-dispatch higher cost generation in that zone. Two alternative approaches to allocating congestion costs without use of zonal electricity prices are simply to divide the cost of generator re-dispatch among (1) all users (loads) system-wide or (2) all users in the load pocket zone. Typically, these re-dispatch costs would be allocated on a pro rata “postage stamp” basis, in proportion to each user’s MW of load in an hour in which constraints and re-dispatch occur.

Hourly transmission losses were projected based on AC power flow simulations taking into account the electric properties of the grid, generator dispatch, and hourly consumption at the numerous load buses. Typically, losses costs would be included in grid users’ hourly electricity prices. This billing might be based on general loss factors periodically calculated for a limited number of different grid conditions, using power flow simulations. Then, when final accounting shows that actual hour by hour losses were somewhat different than this, there will be billing adjustments. Bilateral transactions may be given the option of making up losses in kind, via increased generation.

Ancillary services (AS) considered in the study are regulation up, regulation down, spinning reserve and non-spinning reserve. These services were simulated 2002-2006 through marginal cost-based methods.  AS costs were also calculated using an “average cost” method. For illustration, costs of these four AS under market-based and cost-based pricing as allocated on a postage stamp basis. Two other generator-based AS, reactive supply (voltage regulation) and black start, depend on particular generator characteristics and locations, so that their costs are not well represented by simulating hourly markets for these AS.

Generators owned by an incumbent utility may experience reduced operating incomes due to decline in market share and/or prices in newly deregulated markets with open transmission access. These investment costs are unrecoverable, or “stranded.” Two basic methods for calculating stranded costs are as follows.

• A generator’s net operating income can be compared to the income level needed to recoup the investment cost (depreciation plus return). If the income level falls short, there are stranded costs.
• A generator’s net operating income after some precipitating event  (such as deregulation or loss of a customer) can be compared to the net operating income in the absence of the event. The difference represents stranded costs caused by the event, and stranded costs calculated in this manner may be quite different from those calculated under the first approach.

To estimate the reduced net income projected for existing Taiwan Power generators, two UPLAN scenarios were tested for horizons from 2006 through 2025: (1) assuming that all generators run and sell power based on bids indicative of their marginal costs, and (2) alternatively assuming that output from certain IPP generators must be taken under existing arrangements (“IPP must-take”). The “must-take” scenario results in a projected NPV of net income for existing Taiwan Power generators is shown to be 43% lower, due to a combination of lost market share and lower prices. The report identifies specific generators that potentially suffer stranded costs.