Ready to Work Together?
Stochastic Energy Trading
Intelligent Dispatch Control for Optimal Energy Arbitrage
The increasing integration of intermittent renewable energy sources into electricity grids is driving a growing demand for flexible energy storage solutions. While such renewables bring environmental benefits, they also contribute to greater volatility in electricity prices. Energy arbitrage offers a means of addressing this challenge by purchasing electricity when prices are low and selling it when prices are high. In the context of energy storage systems, this involves charging during periods of low prices and discharging when prices are elevated. The primary objective is to reduce operational costs or enhance revenues for storage operators.
Effective energy arbitrage requires a control strategy capable of optimising the dispatch of energy storage systems to maximise overall performance. Conventional optimisation methods are generally suited to short-term operations and quickly become impractical when applied to larger-scale systems or extended time horizons. This limitation creates an opportunity for more advanced and scalable approaches.
Intractābilis has developed an advanced dispatch control framework that combines deep reinforcement learning with large language model-powered time series forecasting. This approach enhances the ability of reinforcement learning agents to operate effectively in highly variable real-world market conditions. The solution models an energy arbitrage environment for grid-connected storage systems, explicitly accounting for charge and discharge efficiencies. Deep reinforcement learning is employed for dispatch decision-making, while time series forecasting predicts future electricity prices to inform those decisions.
Our results demonstrate that large language model-based forecasting can significantly improve the performance of price-driven arbitrage strategies, enabling more efficient utilisation of energy storage systems and better financial returns. This innovation represents a step forward in bridging the gap between advanced artificial intelligence techniques and the operational demands of modern energy markets.
Optimisation Framework for Real-Time Energy Market and Generator Scheduling
The electricity market sets production and consumption schedules and their prices. Its operations rely on economic dispatch and optimal power flow, both driven by forecasts of demand and the output of renewable technologies. When these forecasts are inaccurate, dispatch decisions can be wrong, which undermines market performance and revenue.
Intractābilis has developed an integrated learning and optimisation framework that addresses these challenges by combining real-time market operations with non-real-time generator scheduling. This unified approach ensures that operational decisions remain economically sound under dynamic market conditions.
The framework is designed to accurately model real-time market processes and generator scheduling practices, incorporating them into a feedback loop that refines unknown parameters within dispatch and power flow formulations. This adaptive capability significantly reduces additional costs borne by demand participants, mitigating the financial impact of generator ramping requirements and line congestion caused by inaccurate load predictions.
This solution also reduces additional charges for demand participants. It does so by limiting costs associated with generator ramping and network congestion that arise from errors in load forecasts.
Quantum Monte Carlo Methods for Real-Time Imbalance Energy Pricing
Renewable energy sources such as photovoltaic arrays and wind farms provide the advantage of low carbon emissions. Their output, however, is inherently dependent on weather conditions making their production forecasting particularly challenging. This variability complicates the task of matching electricity generation to consumption in real time. Even minor forecasting errors can lead to significant system imbalances, compelling Transmission System Operators to undertake costly balancing interventions.
Intractābilis has developed an optimisation framework for real-time imbalance pricing based on Quantum Monte Carlo (QMC) simulation techniques. QMC enables the probabilistic exploration of potential system states with greater sampling efficiency than classical Monte Carlo Tree Search, particularly in high-dimensional and highly non-linear contexts. In a quantum-inspired setting, these algorithms are executed on classical high-performance computing platforms. As quantum-hardware matures, the same approach can be deployed on quantum processors potentially delivering exponential speed in pricing decision generation.
Our system model integrates three core components. The first is an advanced ensemble forecasting system based on Constant Variable Selection Network architectures, designed to predict system imbalance trajectories with high accuracy. The second is a behavioural model represented as a cluster of virtual batteries, to capture demand response effects in reaction to price changes. The third is a quantum-enhanced optimisation loop, in which QMC evaluates alternative price sequences within a simulated environment and identifies those that minimise expected imbalances while preserving grid stability. Our adaptive and resilient pricing mechanism is prepared to meet the operational challenges of an electricity market with high renewable energy sources integration.
© 2025 Intractābilis