Single-Product Real-Time Energy Market

• Single-product real-time energy market is defined as a design where one homogeneous energy commodity is traded using time-indexed prices integrated with operational constraints.
• Various models merge day-ahead and real-time stages using methods like OPF, stochastic, and rolling-horizon formulations to handle congestion, storage, and renewable uncertainty.
• Empirical and computational studies demonstrate that leveraging market flexibility and advanced pricing mechanisms can reduce energy costs and emissions while aligning participant incentives.

A single-product real-time energy market is a market design in which the traded commodity in real time is electrical energy or active power, while congestion, ramping, storage state of charge, reserve activation, emissions, and other operational features enter through constraints, dual variables, or settlement rules rather than through separate real-time products. In the literature, this concept appears in hourly retailer–consumer pricing, two-stage day-ahead and real-time settlement, OPF-based nodal dispatch, continuous frequency-based price discovery, and industrial consumption scheduling against time-varying grid prices and emission factors (Jia et al., 2016, Dvorkin et al., 2019, Liu et al., 2021, Burmeister, 2024, Liu et al., 10 Apr 2026).

1. Concept and scope

In its narrowest formulation, the single product is simply electricity with a time-indexed price. A flexible industrial scheduling model states this explicitly: one homogeneous electricity product is characterized by a price $p_t$ and an emission factor $\gamma_t$, and the manufacturer schedules operations to minimize makespan, energy cost, and emissions against those time-varying series (Burmeister, 2024). In retailer-centric formulations, the same single commodity is represented as hourly electricity sold through a day-ahead hourly price vector $\pi \in \mathbb{R}^{24}$, with real-time consumption adjusting to that vector (Jia et al., 2016). In competitive wholesale formulations, the commodity is energy traded first day-ahead and then in real time as deviations, with no explicit separate product for reserves or ancillary services (Dvorkin et al., 2019).

The scope of the concept is broader than a single institutional template. In OPF-based models, the single product is active power $p_i$ at each node, with locational marginal prices obtained from dual variables of network-constrained dispatch (Liu et al., 2021). In rolling-window storage models, the market remains energy-only even though ramping and state-of-charge create strong temporal coupling (Chen et al., 2022). In aggregator and prosumer platforms, the single product is again energy, sometimes with personalized prices or internal peer-to-peer trades, but without separate reserve or capacity products in the real-time layer (Shomalzadeh et al., 2021, Bjarghov et al., 2019).

A recurrent theme is that “single-product” refers to the commodity, not necessarily to the settlement geometry. Energy-only models may still be single-node or nodal, uniform-price or nonuniform-price, deterministic or stochastic, and centralized or distributed.

2. Temporal architectures and participants

A common architecture is a two-stage settlement system. In a competitive stochastic formulation, agents choose day-ahead quantities $p$ and $d$, then adjust in real time through recourse variables $r_\omega$ and $l_\omega$ after renewable uncertainty $\xi_\omega$ is realized (Dvorkin et al., 2019). A storage market mechanism extends this to a multi-interval setting: generators bid in day ahead and real time, while storage bids cycle depths in day ahead and charge-discharge power in real time for last-minute adjustments (Bansal et al., 2024). In aggregator-oriented real-time platforms, the upper layer sets prices and the lower layer consists of prosumers with renewable generation, elastic demand, and horizon-wide energy constraints (Shomalzadeh et al., 2021).

The temporal meaning of “real-time” varies sharply across the literature. One line of work treats day-ahead hourly prices as the key signal, with consumers adjusting real-time consumption hour by hour (Jia et al., 2016). Another studies the CAISO real-time market at 15-minute resolution within a model predictive control loop for 100 prosumers (Travacca et al., 2018). Community reserve-activation models use a 5-minute real-time stage on top of hourly day-ahead and intraday decisions (Bjarghov et al., 2019). An agent-based spot-and-balancing model simulates the physical system at one-minute resolution while preserving a distinct spot and balancing structure (Kühnlenz et al., 2016). At the fastest end, continuous-time or sub-minute formulations derive the real-time price trajectory directly from frequency measurements (Liu et al., 10 Apr 2026).

The participant set is equally varied. It includes conventional generators, retailers, industrial flexible loads, storage operators, DER aggregators, prosumers, balancing service providers, and distribution system operators. What unifies these models is that each participant ultimately buys, sells, or shifts one energy commodity, even when the surrounding optimization is multi-interval and network-constrained.

3. Clearing models and price formation

The canonical clearing problem is either a welfare maximization or a cost minimization with power balance. In the retail Stackelberg model, the retailer sets a day-ahead hourly price $\pi$, consumers respond with aggregate demand $\gamma_t$0, and the retailer trades off retail profit and consumer surplus on a concave Pareto frontier (Jia et al., 2016). In a two-stage competitive wholesale model, the system-wide balance in each scenario is $\gamma_t$1, with the corresponding dual variable interpreted as the real-time energy price (Dvorkin et al., 2019).

In network-aware real-time markets, price formation is typically OPF-based. Under DC-OPF, the nodal price vector is

$\gamma_t$2

where $\gamma_t$3 is the system marginal energy price and $\gamma_t$4 is the congestion component induced by binding line constraints (Liu et al., 2021). A balancing-market OPF for unbundled electricity markets yields an analogous real-time spot price decomposition into system lambda, marginal losses, and congestion terms, while co-optimizing incremental and decremental energy, reserve deployment, replacement reserves, and bilateral curtailment (Masoud et al., 2016). These formulations are still single-product: the commodity is energy, and the different shadow-price components only describe how network physics modifies its marginal value.

A different but related price-formation logic appears in frequency-based real-time markets. There, the real-time market is represented as a dynamic price-discovery process coupled to swing-equation dynamics, leading to

$\gamma_t$5

so that the real-time price can be reconstructed solely from frequency measurements and known inertia and damping parameters (Liu et al., 10 Apr 2026). This preserves the single-product interpretation: frequency deviations reveal the instantaneous marginal value of balancing active power.

4. Flexibility, storage, and demand shaping

Single-product real-time markets derive much of their economic content from temporal flexibility. In green flexible production, a manufacturer solves a Green Flexible Job Shop Scheduling Problem with operations assigned to machines and start times, while the energy cost $\gamma_t$6 and emissions $\gamma_t$7 of each operation depend on when it is scheduled (Burmeister, 2024). The market signal is exogenous and time-varying, and production scheduling converts that signal into a consumption pattern $\gamma_t$8.

In dynamic retail pricing, thermostatic demand produces an affine aggregate response $\gamma_t$9. The matrix $\pi \in \mathbb{R}^{24}$0 captures both own-price elasticity and temporal cross-effects, so the single energy product is already temporally coupled even before storage is introduced (Jia et al., 2016). When local storage is added under net metering, the HVAC control problem and the arbitrage problem decouple: the affine demand structure is preserved, while the battery exploits inter-temporal price differences (Jia et al., 2016).

Storage intensifies this coupling. In a two-stage market mechanism with storage, day-ahead bids are not merely power quantities; storage bids cycle depths via a Rainflow-based degradation model, and real-time bids are charge-discharge power deviations around that day-ahead baseline (Bansal et al., 2024). In rolling-window real-time dispatch, storage, conventional generation, and DER aggregators are all dispatched in an energy-only optimization, but ramping and state-of-charge shadow prices alter the effective marginal value of energy through Temporal Locational Marginal Pricing (TLMP) (Chen et al., 2022). That formulation shows that single-product does not mean temporally memoryless: one energy commodity can inherit ramp and state-of-charge scarcity through the dual system.

Flexible-demand service models push the same idea further. Rate-constrained energy services specify a delivery window, total energy $\pi \in \mathbb{R}^{24}$1, and maximum per-slot delivery rate $\pi \in \mathbb{R}^{24}$2, and the supplier chooses day-ahead purchases and real-time purchases $\pi \in \mathbb{R}^{24}$3 to satisfy those services while exploiting renewable supply (Nayyar et al., 2014). A plausible implication is that many “energy-only” market designs are best understood not as memoryless spot transactions but as temporally constrained allocation problems for a single commodity.

5. Settlement, equilibrium, and incentive problems

The equilibrium properties of single-product real-time markets depend heavily on information structure and settlement design. Under symmetric information in a two-stage stochastic market, the competitive equilibrium coincides with the centralized social optimum; under asymmetric probability beliefs about renewable generation, prices, day-ahead dispatch, social welfare, and even tâtonnement convergence become highly sensitive to the level of information asymmetry (Dvorkin et al., 2019). Information sharing therefore has economic, operational, and computational value in an energy-only market with renewable uncertainty.

A second issue is uplift and incentive compatibility for inter-temporal resources. In rolling-window real-time dispatch with storage, the paper on TLMP shows that uniform pricing mechanisms require discriminative out-of-the-market uplifts, making settlements under locational marginal pricing discriminative (Chen et al., 2022). TLMP augments LMP with nonuniform shadow prices of ramping and state-of-charge; under TLMP, price-taking participants have truthful bidding incentives, and lost opportunity cost is removed without out-of-market uplift (Chen et al., 2022). This is a direct rebuttal of the misconception that an energy-only market must settle all participants at a single uniform price to remain a single-product market.

A third issue is strategic manipulation across settlement stages. In a two-stage market with prosumers, active prosumers can buy energy in the day-ahead market and sell energy in the real-time market for balancing real-time energy deviations. The paper shows that consumers’ incentives for demand under-reporting vanish when the day-ahead market scales, but prosumers’ incentives remain lower bounded by a positive gain that depends only on the real-time market generation stack and their shares over it (Koumpis et al., 24 Jun 2026). To restore incentive compatibility under existing informational constraints, the day-ahead operator implements a leave-one-out contrastive scoring rule-based penalty that incentivizes truthful demand reporting and keeps charges small for honest participation (Koumpis et al., 24 Jun 2026).

At the aggregator–prosumer layer, similar tensions appear in a different form. A bilevel real-time market with an aggregator setting personalized buy and sell prices for prosumers is nonconvex in its original form, but for the linear–quadratic single-product model considered, a convex quadratic reformulation captures a subset of the global optima of the bilevel problem (Shomalzadeh et al., 2021). The central point is not merely computational: settlement rules and optimization structure jointly determine whether the energy-only market remains implementable in real time.

6. Empirical performance and computational methods

Empirical and computational studies show that single-product real-time markets can produce nontrivial operating gains, but the magnitude depends on the source of flexibility. In flexible industrial production with German hourly prices and emission intensities from 1 February to 30 June 2022, even modest flexibility of 5–20% longer makespan can yield 5–12% electricity cost savings on average, and with extreme flexibility up to 75% longer makespan, cost reductions of about 20–50% are possible; emission savings of 8–12% with 20–50% extended makespan are also feasible on average (Burmeister, 2024).

In community-based peer-to-peer balancing with reserve activation, introducing internal intraday and real-time trading increases participation of 19 and 44 % in reserve markets, because internal balancing energy relaxes the effective real-time constraints of storage-backed peers (Bjarghov et al., 2019). By contrast, a CAISO real-time MPC study with 100 prosumers reports that the supplementary cost over the day is $\pi \in \mathbb{R}^{24}$40.90) when $\pi \in \mathbb{R}^{24}$5, indicating only marginal gains in that illustrative single-product energy-only setting (Travacca et al., 2018).

Agent-based validation against Nord Pool data indicates that even simple spot-plus-balancing models capture relevant real-time phenomena. The model reports an average spot price of about $\pi \in \mathbb{R}^{24}$6 €/MWh versus about $\pi \in \mathbb{R}^{24}$7 €/MWh in Nord Pool data, average regulation of $\pi \in \mathbb{R}^{24}$8 versus $\pi \in \mathbb{R}^{24}$9, and intra-hour regulation around $p_i$0 hours/day versus $p_i$1 hours/day (Kühnlenz et al., 2016). This suggests that energy-only real-time behavior is strongly shaped by intrahour variability even when the day-ahead schedule is hourly.

On the computational side, OPF learning has become a parallel research track. A graph neural network framework for learning real-time nodal prices exploits the locality property of prices, uses physics-aware regularization, and demonstrates scalability and topology adaptivity on IEEE 118-bus and 2383-bus systems with substantially fewer parameters than fully connected baselines (Liu et al., 2021). A plausible implication is that the practical barrier to high-frequency energy-only pricing is increasingly computational architecture rather than the absence of a principled price definition.

7. Assumptions, misconceptions, and frontier directions

Several assumptions recur. Flexible production models often assume perfect foresight of hourly price and emission-factor trajectories, price-taking behavior, unlimited grid supply at posted prices, and linear pricing without demand charges (Burmeister, 2024). Competitive equilibrium models often assume convex and compact feasible sets, price-taking agents, and no network constraints (Dvorkin et al., 2019). Distribution-level real-time incentive schemes assume convex DER costs, linearized network models, and price-taking DER responses to posted incentive signals (Guo et al., 2022). These assumptions make the single-product abstraction analytically clean, but they also delimit its realism.

Two misconceptions are especially persistent. The first is that single-product implies a single uniform price. In OPF-based real-time markets the single product is energy, yet the price is a locational marginal price vector; in storage pricing the product remains energy while TLMP adds nonuniform shadow prices of ramping and state-of-charge to LMP (Liu et al., 2021, Chen et al., 2022). The second is that “real-time” denotes a unique settlement interval. The literature spans hourly real-time consumption under day-ahead hourly prices, 15-minute real-time energy, one-minute physical balancing, and continuous-time frequency-based price discovery (Jia et al., 2016, Travacca et al., 2018, Kühnlenz et al., 2016, Liu et al., 10 Apr 2026).

Current frontier directions combine robustness, network awareness, and online incentives. A two-stage framework for aggregated DERs uses distributionally robust optimization in day ahead and a bi-level time-varying optimization in real time to design online incentive signals that trade off the real-time imbalance penalty for DSOs and the costs of individual DER-owners while satisfying voltage regulation requirements (Guo et al., 2022). Other extensions proposed in the literature include stochastic and rolling-horizon versions of industrial scheduling, hierarchical market structures with network constraints, explicit treatment of demand charges and nonlinear tariffs, and richer probabilistic information sharing in two-stage markets (Burmeister, 2024, Dvorkin et al., 2019).

Taken together, these results define the single-product real-time energy market not as a minimalist market with few constraints, but as a market in which one commodity—energy—remains central while temporal coupling, network physics, strategic behavior, and settlement design determine how that commodity is priced, dispatched, and hedged across time.