Source-linked AI summary
Integration of an Energy Management Tool and Digital Twin for Coordination and Control of Multi-vector Smart Energy Systems
Edward O'Dwyer, Indranil Pan, Richard Charlesworth, Sarah Butler, Nilay Shah
TL;DR
Urban energy systems need flexible tools to coordinate increasingly interconnected assets while addressing environmental, economic, and resilience objectives. The paper develops SEMS, an open-source framework combining machine-learning forecasting, model-predictive control, subsystem coordination, and digital-twin testing. Greenwich case studies illustrate trade-offs, renewable-energy utilisation, and coordinated operation across heating and electric-vehicle systems.
Problem
Urban electrification and interconnected energy demands require a reconfigurable framework for forecasting and optimisation across different energy sectors and system constraints.
Method
SEMS combines machine-learning forecasts, grey-box predictive models, MPC-based subsystem control, high-level coordination, and digital-twin scenario testing.
Results
Greenwich case studies demonstrated environmental-financial trade-offs, increased local renewable-energy utilisation with EV charging loads, and coordination of interconnected energy assets.
Takeaways & Limitations
A flexible open-source tool can support transferable management of multi-vector urban energy systems and scenario-based decisions for local-government objectives.
Takeaways & Limitations
Forecast-generation approaches are not described in detail in the case studies, and SEMS is designed generically rather than fitted to one specific context.
Abstract
from arXiv · showhide
As Internet of Things (IoT) technologies enable greater communication between energy assets in smart cities, the operational coordination of various energy networks in a city or district becomes more viable. Suitable tools are needed that can harness advanced control and machine learning techniques to achieve environmental, economic and resilience objectives. In this paper, an energy management tool is presented that can offer optimal control, scheduling, forecasting and coordination services to energy assets across a district, enabling optimal decisions under user-defined objectives. The tool presented here can coordinate different sub-systems in a district to avoid the violation of high-level system constraints and is designed in a generic fashion to enable transferable use across different energy sectors. The work demonstrates the potential for a single open-source optimisation framework to be applied across multiple energy vectors, providing local government the opportunity to manage different assets in a coordinated fashion. This is shown through case studies that integrate low-carbon communal heating for social housing with electric vehicle charge-point management to achieve high-level system constraints and local government objectives in the borough of Greenwich, London. The paper illustrates the theoretical methodology, the software architecture and the digital twin-based testing environment underpinning the proposed approach.
1. Introduction
Urban energy systems face rising decarbonisation and infrastructure challenges as heating and transport electrify, while IoT, machine learning, and advanced control create opportunities for coordinated management. The paper introduces SEMS, an open-source, modular framework combining forecasting, MPC, coordination, and digital-twin testing for transferable multi-vector energy management.
- Cities consume approximately 60-80% of global energy and produce a similar share of global CO2 emissions.
- Electrification of heating and transport increases resilience demands on power infrastructure while creating opportunities for cross-vector synergies.IoT-enabled sensing, communication, and computational intelligence are presented as tools for real-time operational energy management.
- Existing energy-management approaches motivate a single reconfigurable framework combining forecasting, optimisation, common models, data acquisition, and new functions without major redevelopment.
- SEMS combines model-predictive control and machine-learning forecasting to provide optimal operational decisions for interconnected energy assets.
- SEMS decomposes district problems into subsystem trajectories, then uses high-level coordination to address constraint violations through updated receding-horizon set-points.A digital twin supports testing before implementation and adaptive evaluation alongside real-world operation.
- The open-source framework is designed for multi-vector applications, extensibility, transferability, and scenario-based digital-twin analysis across different energy sectors.
2. Modular control and scheduling framework
SEMS is a generic, modular, open-source framework that combines forecasting, optimisation and advanced control for diverse energy-system contexts. It optimises subsystems individually while coordinating them at district level to respect shared resource constraints.
- Framework architecture: SEMS is designed as a generic and modular framework whose subsystems represent supplies, demands and energy stores across multiple resources.Demand can be represented by static time series or state-space models, supporting adaptation to different district contexts.
- Framework architecture: The open-source architecture uses interchangeable modules for model identification, forecasting, control and coordination, allowing alternative algorithms without major code changes.The SEMS core is written in Python 3.6 and is intended to evolve as new applications arise.
- Control and scheduling: Each subsystem updates model states, set-points and forecasts before solving an optimisation problem that sends optimal inputs and outputs over the following horizon.The framework supports data-driven state-space model generation using ridge regression and formulates optimisation constraints with Pyomo.
- Forecast generation: Forecasting options include classification and clustering for repetitive daily profiles and regression-based artificial neural networks when common patterns are not expected.The classification approach uses representative cluster prototypes, while the ANN approach projects future outputs from learned feature-output relationships.
- Control and scheduling: Model predictive control provides constrained operational decisions for assets such as heat networks, storage systems and electric-vehicle chargers.The framework is intended to address implementation challenges involving interconnected systems, system dynamics and non-proprietary software.
- Coordination layer: District coordination addresses coincident subsystem demand peaks by evaluating shared resource requirements and tightening subsystem limits when a specified threshold is exceeded.This coordination layer is particularly relevant as heating and transport become more electrified.
3. Integration of the SEMS and system simulation
The paper integrates a digital-twin simulation environment with SEMS to test and coordinate multi-subsystem district energy strategies before deployment. The environment models energy sources, conversion, storage, demand, and external influences, while SEMS uses coordinated optimisation to manage shared constraints.
- Simulation environment: A digital twin combines online measurements with simulated system behaviour to support operational decisions, performance analysis, fault identification, and uncertainty testing.The underlying simulation environment replicates real-system behaviour and supports scenario testing before installation.
- Coordination algorithm: The coordination algorithm iteratively reduces subsystem power limits when predicted aggregate demand exceeds the system capacity threshold, then reformulates and resolves each subproblem.The procedure updates subsystem and total-power predictions after each reformulation until the constraint condition is satisfied or the iteration limit is reached.
- Simulation environment: Off-line simulations enable software-interface testing, rigorous evaluation of control and operational strategies, and validation of proposed technological interventions.This supports in-silico refinement of strategies before installation in districts with diverse subsystem configurations and stakeholder priorities.
- Simulation environment: The simulation environment combines energy-source, conversion, storage, demand, and external-influence models to represent district energy-system behaviour.Sources include gas, grid electricity, wind, and solar; conversion technologies include boilers, heat pumps, PV panels, chillers, and CHP systems.
- System architecture: The architecture connects physical assets and controllers with independently containerised cloud services, including databases, SEMS modules, and the digital twin.Services communicate through API calls, while the data store receives asset and environmental variables and SEMS algorithms use them to compute controller set-points.
4. Implementation: The SEMS in the Sharing Cities project
The Greenwich case studies use a digital twin with SEMS to model heating and EV subsystems, optimise their operation, and coordinate them against district objectives and power constraints. Results show trade-offs between cost, emissions, renewable utilisation, and grid-constraint violations.
- System configuration: The Greenwich digital twin represents upgraded communal heating, EV charging, and solar PV subsystems for SEMS testing.The heating model includes heat pumps, boilers, and thermal storage; the electrical subsystem includes 300 EVs and 270 kWp of PV generation.
- Heating optimisation: SEMS balances financial cost and CO2 emissions in heating operation by switching between boiler- and heat-pump-dominated strategies using forecasts.It can preheat the network and thermal store before price spikes, while objective weights determine the cost–emissions balance.
- Heating optimisation: A Pareto frontier shows that higher spending can achieve greater emission reductions, with diminishing returns.Five 14-day simulations were repeated with different financial and environmental objective weights.
- Heating optimisation: The thermal store charges to 60°C during an overnight price dip and discharges during the day when electricity prices are higher.This illustrates how storage shifts energy use across time to improve the selected objective.
- EV charging optimisation: SEMS delays and reallocates EV charging toward the solar-generation peak, increasing PV utilisation and the renewable share of EV consumption.The reallocation also reduces grid consumption, producing reported cost and CO2 savings over the 14-day period.
- Coordinated operation: Coordinated subsystem management reduced the summed magnitudes of district power-supply violations by approximately 97%.The imposed constraint increased first-subsystem CO2 emissions by 20.8% and energy cost by 8.9%, reflecting additional boiler operation.
5. Conclusions
The paper presents SEMS as a modular tool for coordinating interconnected urban energy assets, combining predictive control with district-level constraints. Greenwich case studies illustrate trade-offs between environmental and financial objectives, renewable-energy utilisation, and coordinated operation.
- SEMS combines machine-learning forecasts, model fitting, flexible model predictive control, and a coordination layer for interconnected energy assets.The coordination layer imposes higher district-level constraints while subsystems can be considered independently.
- Greenwich case studies demonstrate SEMS applications across interconnected urban energy systems.The cases use current interventions from the Sharing Cities project as their basis.
- The heating case illustrates trade-offs between environmental and financial objectives for decision makers.
- A further case study examines increasing local renewable-energy utilisation while incorporating electric-vehicle charging loads.
- A final example demonstrates SEMS coordination capabilities, while its flexible open architecture supports later integration of algorithms and energy sub-systems.