Co-Optimizing Energy and Ancillary Services with PLEXOS Under Centralized Dispatch Systems Market Clearing Model

Introduction

Centralized Dispatch System (CDS) is a market clearing model that is typically used by the market operators in North America. Under CDS market clearing, the offers of the individual producers are collected by the market operators. The operators then clear the market and determines the dispatch level for each generation facility to ensure that the generation-demand balance is met in every period, while also keeping the total system cost at minimum. In other words, the system operator defines the unit commitment.

It is well known that due to the demand fluctuations and the intermittency of the renewable energy resources, the actual total system generation level in real time operation is never matches the pre-determined cleared volumes. To ensure that the generation-demand balance can be always maintained, the system operators procure ancillary (in this case frequency regulation) services. Such services in traditional power grids are also provided by the conventional generation plants.

It is obvious, that the provision of ancillary services (reserving certain generation capacity) limits the amount of the power that the generation plant can deliver to the energy market. It has historically been proven, that the clearing the energy and ancillary service market mutually (by co-optimisation) benefits both the system operator and the service providers.

However, it must be remembered, that some operators allow the producers to submit different bids (offer size in MW and corresponding offer price) for the ancillary service (AS) provision, while in other markets, the service providers can only deliver energy bids.

As expected, an identical system under the two market set-up leads to different energy and reserve prices, which also impact the potential revenues of the service providers.

This article discusses the impact of the two bidding strategies (both energy only, and energy and AS bids together) on the system prices, market conditions and expected revenues for service provision.

System Configuration

In PLEXOS, to analyse the co-optimisation of the energy and AS markets (in the presence and absence of separate AS service bids) a simplified network model from [1] was used that was modified to fit better for the purpose of the assessment. The system consists of 4 generator units with the following specifications (maximum power generation, maximum reserve provision, short run marginal cost (SMRC) for energy generation and variably operation and maintained cost (VO&M) associated with reserve capacity provision):

·         Gen 1: Pmax =250 MW, Reserve (max)=0 MW, SRMC=2 $/MWh

·         Gen 2: Pmax =230 MW, Reserve (max)=160 MW, SRMC=17 $/MWh.

·         Reserve(VO&M-Gen2)=0 $/MWh

·         Gen 3: Pmax =240 MW, Reserve (max)=190 MW, SRMC=20 $/MWh.

·         Reserve(VO&M-Gen3)=5 $/MWh

·         Gen 4: Pmax =250 MW, Reserve (max)=150 MW, SRMC=28 $/MWh.

·         Reserve(VO&M-Gen4)=7 $/MWh

The total system demand varied between 300 and 720 MW with a reserve up-regulation requirement of 250 MW, allowing representation of diverse scenarios.

In PLEXOS a reserve object was created with “Raise” type (to represent up-regulation), each generator was linked to the reserve objective with their respective maximum reserve capacity (as above) and offer prices (when separate AS bids were allowed). Two simulation models were created, one with only energy bids and the other one with energy and separate AS bids.

Since PLEXOS, by default, co-optimizes energy and AS markets no additional settings were required.

Linear Programming Files in PLEXOS

Every PLEXOS simulation model generates linear programming (LP) files that represent the system objective functions, and the corresponding equality and inequality constraints of the system. These files are then passed to the solver to calculate the system states the results. Since the LP files are accessible, comparing them shows how PLEXOS handles the energy and AS market co-optimization.

Figure 1 shows the LP files for 300MW demand case as an example. The left-hand side shows the energy bid only case, while the right-hand side corresponds to the energy+AS bids scenario.

In the case of energy only bids, PLEXOS tries to minimize the total system cost, considering the SRMCs of the Machines. The system demand of 300MW, along with the generator ratings and total reserve requirements of 250MW are listed at the bottom of the LP file, under section “Bounds”.

The system constraints are given in section “Subject to”. Here, the “ResDef_Spinning” formula specifies that the 250 MW of reserve must be met by the combination of reserve provisions from Machines 2, 3, and 4. Furthermore, for each generator, there is a constrains (let us use unit 2 as example) that the combined generator output (GenLoad_Gen_2) and reserve provision (GenRaise_Gen_2) cannot exceed the respective generation rating (<= 230).

This confirms that, by default, any reserve provision requirement is taken into account when optimizing the system for minimum cost.

Comparing the LP file on the left to that on the right-hand side (when energy and separate AS bids are considered) the objective functions is expanded by the AS provision costs (VO&Ms) of the respective units to ensure that, in overall, the system runs at lowest cost considering both energy and ancillary service bids.

Energy Bids Only Scenario

The first case investigates the market conditions under energy only bids scenario. In this case the respective VO&M costs of reserve provision were ignored. The Machine generation outputs and reserve provision levels are shown on Figure 2 and 3, while the Figure 4 shows the corresponding energy and reserve prices in the network, and Figure 5 shows the net profits by Machines. Each figure provides the results as a function of the system demand (300-720 MW).

Machine 1 generates energy at its full capacity irrespective of the demand. This is logical as Unit 1 has the lowest SRMC (2 $/MWh) and does not contribute to the reserve provision.

Between demand levels of 300-480MW, Machine 2 (second cheapest) generates between 50-230 MW (up to its rated capacity). At this demand level, Machine 2 is the marginal unit for energy generation, setting the energy price at its SRMC of 17 $/MWh.

Since the reserve provision has no dedicated price tag attached, the reserve provision cost is not part of the objective function in this scenario. Accordingly, PLEXOS is free to use which Machine provides the reserve, as it will not cost anything to the objective function. Up to a demand of 380 MW, machines 2 and 4 provide the reserve. Machine 4 provides its maximum 150MW reserve capacity. Beyond 380MW, Machines 2 increases generation output further (marginal Machine) and decreases reserve provision, which is picked by Machine 3.

From 480MW of demand, when both Machine and 2 runs at its full load, Machine 3 becomes the marginal unit for energy setting energy price at its SRMC of 20 $/MWh. Up to 620MW, Machine 3 increases generation output.

Since between 300-620MW, the overall reserve capacity is plentiful with Machine 2, 3 and 4, the reserve constraint is not binding, therefore the reserve price is 0 $/MWh. It is possible to deliver extra MW of reserve without any limitation.

Beyond 620MW of load, however, since Machine 3 reached an output of 140 MW (and provides 100MW reserve) Machine 4 starts generating. It is now the marginal unit, setting energy price to its SRMC of 28 $/MWh.

Since beyond 620MW of demand, Machine 3 and 4 are providing their maximum reserve capacity, in the case if additional reserve is required the following happens. To deliver an extra 1MW of system reserve, Machine 3 reduced its generation output by 1MW which is picked by Machine 4, so that the system can have 1MW extra reserve (above the 250MW). Machine 4, which picks up the extra MW generation does it for a cost of 28$, but the reduction in Machine 3 output reduces the system cost by 20$, according their SRMCs. Hence the cost of reserve is 28-20=8$/MWh. Consider this as a shadow price for reserve provision.

Figure 4 shows the respective changes of energy and reserve prices.

The net profits of the Machines are easy to determine. For example, at demand of 650MW, Machine 1, 2, 3, and 4 generates 250MW, 230MW, 140MW and 30MW, respectively. The energy price is set by Machine 4 at 28$/MWh. Considering the SRMCs, at this price level, Machine 1 makes 28-2=26$, Machine 2 makes 28-17=11$, Machine 3 makes 28-20=8$ profit on each MW delivered. Machine 4 as marginal unit does not make profit on energy. Machine 3 and 4 makes profit on their reserve provision at 8$/MW. The overall profit variation as a function of the system demand is shown in Figure 5.

Separate Energy + AS Bids Scenarios

The second scenario considers that energy and ancillary service bids are both submitted. In this case the cost of AS provision is included in the objective function (as the LP files already confirmed) which results in different market conditions. Generation output and reserve provision levels for the Machines are shown on Figure 6 and 7. Figure 8 shows the resulting market conditions (energy and reserve prices) and the Figure 9 shows the corresponding net profits of the four Machines.

Again, since the demand is always higher than Machine 1 rating, who does not contribute to the reserve, it is operating at full load, irrespective of the demand level.

Between 300-320MW of load, Machine 2 is the marginal unit for energy, setting the price to its SRMC of 17$/MWh. Since Machine 2 provides reserve for free (~VO&M=0$/MWh) it is contributing to the reserve with is maximum reserve capacity of 160MW, while the rest of reserve is covered by the second cheapest reserve producer, Machine 3. Accordingly, Machine 3 reserve VO&M sets the price for reserve as 5 $/MWh.

Between 320-470MW, Machine 2 is kept at 70MW generation (to provide maximum reserve of 160MW) and Machine 3 starts generating, setting the energy price to its SRMC of 20$/MWh. This set-up is indeed the optimum lowest system cost option).

Let us consider the demand of 330MW, as an example. Machine 3 generating 10MW in this case. Machine 2 makes 3$ profit on each MW generated (70*3=240$) while also receiving the AS revenue of 160*5=800$. If Machine 2 was to generate extra 10MW instead, it would need to reduce its reserve (from 160 to 150MW) which would be picked by Machine 3. In this case Machine 2 would receive for the extra 10MW generation 10*3=30$, while on the other hand would lose 10*5$=50$ on having 10MW less reserve provision. Accordingly, increasing Machine 3 output over 320MW of demand is indeed the optimum solution.

At 470MW of demand, Machine 2 and 3 generates 70MW and 150MW, and provide reserve of 160MW and 90MW. Neither one can provide more reserve of generate more energy at this point. Above 470MW of demand, Machine 3 as marginal unit for energy increases generation, hence, reduces its reserve provision which is picked by Machine 4. Machine 4 is the marginal provider of reserve, setting the reserve price to its VO&M of 7$/MWh.

It is interesting to see what happens with the energy price. Producing the extra 1MW with generator 3 costs 20$, but this production reduces the reserve provision, by the same 1MW, saving 5$. This reserve provision is picked by unit 4 at a cost of 7$. Accordingly, the energy price is 20-5+7=22$/MWh, which is not matching any individual Machines SRMC. This reflects the impact of the network (in this case reserve provision) constraints on the energy price. In this case the 22$/MWh is the shadow price (of having the reserve constraint present).

When demand raise above 560MW, interesting to note that Machine 2 becomes the marginal energy provider again, but the energy price is set at 24$/MWh. This is because increasing Machine 2 output, cost 17$, and it is reducing its reserve contribution, but is provides reserve at 0$/MWh. The reserve deficit is picked by Machine 4 at 7$/MWh, hence the energy price is 17-0+7=24$/MWh. If instead unit 4 started to generate the energy price would have been 28$ instead of 24$, while the reserve price was unchanged, leading to a sub-optimal solution.

Lastly, above 620MW, similarly to energy only bids case, Machine 4 generates energy as marginal unit setting the energy price of the system to its SRMC of 28$/MWh. Since generator 4 already provides its maximum reserve capacity, unit 2 (the cheapest reserve provider) would produce any extra MW of reserve. This would be done by reducing its output (saving 17$) which would be picked by generator 4 at 28$, leading to a reserve price of 28-17=11$/MWh as Figure 8 shows.

The corresponding profit plots (Figure 9) shows the impact of Machine 3 becoming marginal unit at 480MW, requiring it to reduce itst reserve, hence leading to decreased profit up to 560MW. Being marginal for energy at this demand range does not yield to profit, while the reduced reserve revenues, indeed decrease the overall profit Machine 3 achieves.

Comparing the net profits of the Machines under both scenarios, as expected, the generally higher profit levels can be obtained when ancillary service-related bids are allowed, so that service provision related cost can be priced better.

Summary

This article explored the co-optimization of energy and ancillary service (AS) markets within a Centralized Dispatch System (CDS) using PLEXOS simulation under two distinct market-clearing methods: energy only bids and energy plus ancillary services separate bids. The examination provided insights into how different bidding strategies affect system prices, potential revenues for service providers, and the overall system efficiency.

Impact on System Prices: The simulation results indicate that separate AS bids lead to different energy and reserve prices, affecting the financial outcomes for service providers. The scenario, which includes costs for AS provision, generally results in higher reserve prices and more accurate compensation for reserve providers.

Generator Utilization and Market Efficiency: Under the separate energy and AS bid scenario, generators capable of providing reserves do so more efficiently, reflecting the true cost of reserve provision in the market prices. This leads to an optimized use of resources, where generators are dispatched and reserves are utilized in a manner that reflects their true costs and capabilities.

Financial Outcomes for Service Providers: The comparison of net profits across scenarios reveals that generators potentially achieve higher profits when allowed to submit separate bids for energy and AS. This is particularly evident for units that serve as marginal providers of energy or reserves, as their contributions are better valued in the market.

Conclusion

The co-optimization of energy and ancillary service markets, especially under a system that allows for separate AS bids, enhances market efficiency and ensures a more accurate reflection of the costs and values of energy and reserve provisions. The findings underscore the importance of market design in accommodating the complexities of modern power systems, highlighting the benefits of flexibility in bidding strategies for both system operators and service providers. As the energy landscape evolves, embracing co-optimization strategies in market clearing processes will be crucial in maintaining system reliability, efficiency, and fairness in compensating service providers.

Reference

[1] Daniel Kirschen, Goran Strbac (2004) Fundamentals of Power System Economics. John Wiley & Sons, Ltd.