Bidding Strategies

Bidding strategies are a core concept of ASSUME which describe how agents bid their assets on a market.

Overview

As described in the Exchangeable Bidding Strategy, a Bidding Strategy dictates the bidding behavior of units in different markets, whereby it maps certain states and technical constraints to bidding decisions. In general, there is a distinction between two kinds of strategy classes:

• assume.strategies.portfolio_strategies.UnitOperatorStrategy(), which indicate the use in a UnitsOperator and can be used to provide a Portfolio optimization.

• Strategies used for a single unit of Unit Types, which range from naive (e.g. :py:meth:’assume.strategies.naive_strategies.EnergyNaiveStrategy’)to advanced (e.g. :py:meth:’assume.strategies.advanced_orders.EnergyHeuristicFlexableLinkedStrategy’).

Both types have a function calculate_bids which is called with the information of the market and bids to bid on.

UnitOperatorStrategy

The UnitsOperatorStrategies can be used to adjust the behavior of the UnitsOperator. The default is the assume.strategies.portfolio_strategies.UnitsOperatorDirectStrategy(). It formulates the bids to the market according to the bidding strategy of the each unit individually. This calls calculate_bids` of each unit and returns the aggregated list of all individual bids of all units. This is the default for all UnitsOperators which do not have a separate strategy configured.

Another implementation includes the assume.strategies.portfolio_strategies.UnitsOperatorEnergyHeuristicCournotStrategy() that adds a markup to the marginal cost of each unit of the units operator. The marginal cost is computed with EnergyNaiveStrategy and the markup depends on the total capacity of the unit operator.

UnitStrategies

The ASSUME framework provides multiple options in terms of Bidding Strategy methodologies:

Bidding Strategy Methodology	Description
Naive	Simple bidding strategies for participating in a market mechanism that follows the Merit Order principle. No additional dependencies needed.
Heuristic	Basic methodology to form bids, based on participating in a market mechanism that follows the Merit Order principle. These strategies do not utilise forecasting or consider finer details such as the effects of changing a power plant’s operational state (start-up costs etc.), so the bid volume is of the order of its maximum power capacity (given ramping constraints), and the bid price is set to the marginal cost.
Optimization	Methodology, based on the flexABLE methodology [Qussous et al. 2022], offers more refined strategising compared to the Naive methods. Applicable to power plant and storage units, it incorporates market dynamics to a greater degree by forecasting market prices, as well as accounting for operational history, potential power loss due to heat production, and the fact that a plant can make bids for multiple markets at the same time.
Learning	A reinforcement learning (RL) approach to formulating bids for an Energy-Only Market. Agents perform actions (choose bid price(s) and for storage a direction) informed by observations (including forecasted residual load, forecasted price, marginal cost). Bid volumes are fixed to the maximum possible volume. Based on the reward (profits) from accepted bids, agents learn to optimise bids to maximise profits. This requires to have PyTorch installed.
Interactive	Strategies which let a user handle input through a terminal or other interface.

For each Bidding Strategy methodology there are multiple Bidding Strategy options depending on the product and market that the bid is intended for, as well as the type of unit making the bid.

Accordingly, each Bidding Strategy has an associated ID which takes the form “unit_product_method_comment”, the comment being optional. This “bidding_strategy_id” needs to be entered when defining a unit’s bidding strategy. Each Bidding Strategy and associated ID for each methodology is defined and described further below.

When constructing a units CSV file, the bidding strategies are set using "bidding_*" columns, where the market type the bidding strategy is applied to follows the underscore (the market names need to match with those in the config file).

name	technology	bidding_EOM	bidding_CRM_pos	bidding_CRM_neg	max_power
Naive-Bidding Unit	hydro	powerplant_energy_naive	powerplant_capacity_heuristic_balancing_pos	powerplant_capacity_heuristic_balancing_neg	1000
Advanced-Bidding Unit	hydro	powerplant_energy_heuristic_block	powerplant_capacity_heuristic_balancing_pos	powerplant_capacity_heuristic_balancing_neg	1000

We’ll now take a look at the different Bidding Strategies within each methodology, their associated “bidding_strategy_id”, and for which unit and market type(s) they are valid.

Naive

bidding_strategy_id	For Market Types	Description
powerplant_energy_naive	EOM, CRM_pos, CRM_neg	Basic strategy for merit order markets at a single timepoint (hour). Uses marginal cost as bid price and the maximum feasible power (respecting ramping constraints) as bid volume.
demand_energy_naive	EOM, CRM_pos, CRM_neg	Basic naive strategy for demand units. Can realise a price-inelastic demand by setting a very high bid price and volume equal to demand at the timepoint.
powerplant_energy_naive_otc	OTC	Similar to powerplant_energy_naive but for OTC (bilateral) trades.
demand_energy_naive_otc	OTC	Similar to demand_energy_naive but for OTC (bilateral) trades.
powerplant_energy_naive_profile	EOM, CRM_pos, CRM_neg	Similar to powerplant_energy_naive but submitted as a 24-hour block (Day-Ahead). Bid price is set to the marginal cost at the starting timepoint.
powerplant_energy_naive_redispatch	redispatch	Submits unit info and currently dispatched power for upcoming hours to the redispatch market (includes marginal cost, ramping, and dispatch information).
demand_energy_naive_redispatch	redispatch	Submits unit info and currently dispatched power for upcoming hours to the redispatch market (includes marginal cost, ramping, and dispatch information).
household_energy_naive_redispatch	redispatch	Naive DSM strategy for industry/household units; bids available flexibility on redispatch. Volume equals flexible power at product start; price equals marginal cost at product start.
industry_energy_naive_redispatch	redispatch	Naive DSM strategy for industry/household units; bids available flexibility on redispatch. Volume equals flexible power at product start; price equals marginal cost at product start.
exchange_energy_naive	EOM	Incorporates cross-border trading: submits export (negative volume) and import bids. Exports are treated as demand with very high bid price; imports as supply with a very low bid price to virtually guarantee acceptance.

Naive method API references:

• assume.strategies.naive_strategies.EnergyNaiveStrategy()

• assume.strategies.extended.EnergyNaiveOtcStrategy()

• assume.strategies.naive_strategies.EnergyNaiveProfileStrategy()

• assume.strategies.naive_strategies.EnergyNaiveRedispatchStrategy()

• assume.strategies.naive_strategies.DsmEnergyNaiveRedispatchStrategy()

• assume.strategies.naive_strategies.ExchangeEnergyNaiveStrategy()

Heuristic

bidding_strategy_id	For Market Types	Description
demand_energy_heuristic_elastic	EOM	This bidding strategy is formulated for a demand unit to realise a “price-elastic” demand bid, approximating a marginal utility curve. The elasticity can be set to “linear” or “isoelastic”.
powerplant_energy_heuristic_flexable	EOM	A more refined approach to bidding on the EOM compared to naive. A unit submits both inflexible and flexible bids per hour. The inflexible bid represents the minimum power output, priced at marginal cost plus startup costs, while the flexible bid covers additional power up to the maximum capacity at marginal cost. It incorporates price forecasting and accounts for ramping constraints, operational history, and power loss due to heat production.
powerplant_energy_heuristic_block	EOM	A power plant strategy valid for complex market clearing which bids dict blocks to the market. The bid is for a block of multiple hours instead of being for a single hour. A minimum acceptance ratio (MAR) defines how to handle rejected bids within individual hours of the block. For the inflexible bid, the MAR is set to 1, meaning all bids within the block must be accepted otherwise the whole block bid is rejected. A separate MAR can be set for children (flexible) bids. See the Advanced Orders tutorial: https://assume.readthedocs.io/en/latest/examples/06_advanced_orders_example.html#1.-Basics
powerplant_energy_heuristic_linked	EOM	A power plant strategy which handles block and linked bids on a market with these fields as a dict. The strategy is similar to powerplant_energy_heuristic_block but allows to integrate the flexible bids as linked bids. See the Advanced Orders tutorial: https://assume.readthedocs.io/en/latest/examples/06_advanced_orders_example.html#1.-Basics
powerplant_capacity_heuristic_balancing_neg	CRM_neg	A bid on the negative Capacity or Energy Control Reserve Market (CRM); volume is determined by calculating how much it can reduce power. The capacity price is found by comparing the revenue it could receive if it bid this volume on the EOM; the energy price is the negative of marginal cost.
powerplant_capacity_heuristic_balancing_pos	CRM_pos	A bid on the positive Capacity or Energy CRM; volume is determined by calculating how much it can increase power. The capacity price is found by comparing the revenue it could receive if it bid this volume on the EOM; the energy price is the positive of marginal cost.
storage_energy_heuristic_flexable	EOM	Determines strategy of a Storage unit bidding on the EOM. The unit acts as a generator or load based on the average price forecast. If the current price forecast is greater than the average price, the Storage unit will bid to discharge at a price equal to the average price divided by the discharge efficiency. Otherwise, it will bid to charge at the average price multiplied by the charge efficiency. Calculates ramping constraints for charging and discharging based on theoretical state of charge (SOC), ensuring that power output is feasible. The bid volume is subject to the charge/discharge capacity of the unit.
storage_capacity_heuristic_balancing_neg	CRM_neg	Analogous to storage_energy_heuristic_flexable, but bids either on the negative capacity CRM or energy CRM.
storage_capacity_heuristic_balancing_pos	CRM_pos	Analogous to storage_energy_heuristic_flexable, but bids either on the positive capacity CRM or energy CRM.

Heuristic method API references:

• assume.strategies.naive_strategies.EnergyHeuristicElasticStrategy()

• assume.strategies.advanced_orders.EnergyHeuristicFlexableBlockStrategy()

• assume.strategies.advanced_orders.EnergyHeuristicFlexableLinkedStrategy()

• assume.strategies.flexable.CapacityHeuristicBalancingNegStrategy()

• assume.strategies.flexable.CapacityHeuristicBalancingPosStrategy()

• assume.strategies.flexable_storage.StorageEnergyHeuristicFlexableStrategy()

• assume.strategies.flexable_storage.StorageCapacityHeuristicBalancingNegStrategy()

• assume.strategies.flexable_storage.StorageCapacityHeuristicBalancingPosStrategy()

Optimization

bidding_strategy_id	For Market Types	Description
household_energy_optimization	EOM	An energy strategy of a Household DSM unit. The bid volume is the optimal power requirement of the optimization.
industry_energy_optimization	EOM	An energy strategy of a Industry DSM unit. The bid volume is the optimal power requirement of the optimization.
household_capacity_heuristic_balancing_neg	CRM_neg	A negative capacity strategy of a Household DSM unit. The bid volume is the optimal power requirement of the optimization.
household_capacity_heuristic_balancing_pos	CRM_pos	A positive capacity strategy of a Industry DSM unit. The bid volume is the optimal power requirement of the optimization.
industry_capacity_heuristic_balancing_neg	CRM_neg	A negative capacity strategy of a Household DSM unit. The bid volume is the optimal power requirement of the optimization.
industry_capacity_heuristic_balancing_pos	CRM_pos	A positive capacity strategy of a Industry DSM unit. The bid volume is the optimal power requirement of the optimization.
powerplant_energy_optimization_dmas	EOM	Power plant strategy using forecast optimization and avoided cost calculation used for smart bids coming from the DMAS methodology
storage_energy_optimization_dmas	EOM	Storage strategy using forecasts and avoided cost calculation used for smart bids coming from the DMAS methodology

Optimization method API references:

• assume.strategies.naive_strategies.DsmEnergyOptimizationStrategy()

• assume.strategies.naive_strategies.DsmCapacityHeuristicBalancingPosStrategy()

• assume.strategies.naive_strategies.DsmCapacityHeuristicBalancingNegStrategy()

• assume.strategies.dmas_powerplant.EnergyOptimizationDmasStrategy()

• assume.strategies.dmas_storage.StorageEnergyOptimizationDmasStrategy()

Learning

bidding_strategy_id	For Market Types	Description
powerplant_energy_learning	EOM	A reinforcement learning (RL) approach to formulating bids for a Power Plant in an Energy-Only Market. The agent’s actions are two bid prices: one for the inflexible component (P_min) and another for the flexible component (P_max - P_min) of a unit’s capacity. The bids are informed by 50 observations, which include forecasted residual load, forecasted price, total capacity, and marginal cost, all contributing to decision-making. Noise is added to the action, especially towards the beginning of the learning, to encourage exploration and novelty. The reward is calculated based on profits from executed bids, operational costs, opportunity costs (penalizing underutilized capacity), and a regret term to minimize missed revenue opportunities. This approach encourages full utilization of the unit’s capacity.
storage_energy_learning	EOM	Similar RL approach as learning_eom_powerplant, for a Storage unit. The make-up of the observations is similar to those for learning_eom_powerplant, with an additional observation being the State-of-Charge (SOC) of the storage unit. The agent has 2 actions - a bid price, and a bid direction (to buy, sell or do nothing). The bid volume is subject to the charge/discharge capacity of the unit. The reward is calculated based on profits from executed bids, with fixed costs for charging/discharging incorporated.
powerplant_energy_learning_single_bid	EOM	Reinforcement Learning Strategy with Single-Bid Structure for Energy-Only Markets. This strategy is a simplified variant of the standard EnergyLearningStrategy, which typically submits two separate price bids for inflexible (P_min) and flexible (P_max - P_min) components. Instead, EnergyLearningSingleBidStrategy submits a single bid that always offers the unit’s maximum power, effectively treating the full capacity as inflexible from a bidding perspective.
renewable_energy_learning_single_bid	EOM	Reinforcement Learning Strategy for a renewable unit that enables the agent to learn optimal bidding strategies on an Energy-Only Market.

Learning method API references:

• assume.strategies.learning_strategies.EnergyLearningStrategy()

• assume.strategies.learning_strategies.EnergyLearningSingleBidStrategy()

• assume.strategies.learning_strategies.StorageEnergyLearningStrategy()

• assume.strategies.learning_strategies.RenewableEnergyLearningSingleBidStrategy()

Other

bidding_strategy_id	For Unit Types	For Market Types	Description
powerplant_energy_interactive	Any	Any	The bidding volume and price is manually entered in the terminal.

Miscellaneous method API references:

• assume.strategies.interactive_strategies.EnergyInteractiveStrategy()