Search

Large-scale heat pumps are a promising technology for the decarbonisation of heat supplied in buildings and industries, provided they operate as expected. However, common faults like fouling and unplanned downtime periods can significantly affect their performance and availability. This could limit the widespread adoption of large-scale heat pumps over other heating technologies such as gas and electric boilers. Approaches described in the literature to optimize the operation of large-scale heat pumps often lack validation under real-world conditions and do not account for performance degradation due to faults. This work demonstrates a step towards utilizing digital twins to improve the energy performance of a commercial large-scale heat pump affected by fouling. A framework was proposed based on the real-time adaptation of digital twins, where a simulation model was calibrated online based on measurements from the heat pump in operation, which was then used for set point optimization. This enabled to determine optimal intermediate pressure set points in the heat pump operating under varying levels of fouling over time. The framework was tested on different periods of heat pump operation spread over ten calendar months. The results showed that the use of online calibration rather than a single calibration decreased performance estimation errors between 3 and 17 percentage points. Moreover, the set points determined by the online-calibrated model, along with a simpler polynomial model derived from it, showed improvements in the heat pump performance by up to 3%, depending on the level of fouling. The findings of this study demonstrated the potential to extend the proposed framework using digital twins to enhance the energy efficiency of large-scale heat pumps. • Use of a model calibrated online to simulate a two-stage ammonia heat pump. • Online calibration enabled to assess fouling growth and mitigation in real-time. • Intermediate pressure set points were optimized through the online calibrated model. • Online calibration improved performance estimations by 3 to 17 percentage points. • Model-based fouling monitoring was beneficial for the set point optimization. [ABSTRACT FROM AUTHOR]

Grid-Connected Wind Turbine is a comprehensive dynamic system comprising cross-coupling characteristics in various time scales. This paper aims to construct a real-time semi-physical co-simulation of Variable-Speed-Variable-Pitch Wind Turbines (VSVP-WTs) equipped with a permanent magnet synchronous generator (PMSG). As for the deployment configuration, software such as OpenFAST and LabVIEW are used for system modeling, while the hardware, mainly NI devices, is adopted for real-time operation and strategy implementation. The physical behaviors of energy conversion are discussed separately, especially on aeroelastic dynamics and electronic power switching. Additionally, operational strategy for the whole production is established through an embedded industrial controller, which also serves as a bi-directional interaction path. An Ethernet network-based communication interface is constructed as an information channel in which a discrete model is extracted and discussed. The synchronization and stability among hardware are guaranteed with a proper period. Several cases demonstrate detailed dynamics of the whole WT, brought on by command adjustment, environmental factors, and power system faults. The proposed platform would be considered a critical stage in digital twin application of engineering. • An aero-elastic–electrical hybrid real-time platform is constructed for PMSG-WT. • OpenFAST-NI HIL co-simulation experimental is implemented via stable communication. • A combination of mechanical–electrical operational strategy is established for power regulation. • Accuracy is guaranteed through experimental cases with three kinds of power scenarios. [ABSTRACT FROM AUTHOR]

The soaring potential of edge computing leads to the emergence of cloud–edge collaboration. This hierarchy enables the deployment of artificial intelligence models in the cyber–physical venue. This paper presents Dual Digital Twin, the next level of digital twin, in the presence of two levels of communication availability, for battery system real-time monitoring and control in electric vehicles. To implement the dual digital twin concept, an online adaptive model reduction problem is formulated with time scale differences induced by the time sensitivity property of industrial applications and limitations of infrastructure. To minimize the model reduction error and battery system control penalty, the online adaptive battery reduced order model framework is proposed, consisting of the gated recurrent unit neural network to construct battery internal states given Internet of things sensor measurements, and incremental learning techniques to facilitate the update of the reduced-order model given data stream. Moreover, we design the physics-informed update of the neural network using the Lyapunov stability theorem to enhance the synchronization with the physical battery behavior. A Li-ion battery and single particle digital twin model with electrolyte and thermal dynamics are utilized in the simulation to justify the effectiveness of the proposed framework. Numerical results demonstrate 1.70% average reduced-order model prediction error and 43.3% accuracy improvement with the novel physics-informed online adaptive framework. The method is also robust concerning varying environmental factors and noise. • Introduce the dual digital twin concept and its realization in EV batteries. • Online adaptive model reduction problem with intrinsic time scale difference in AIoT system. • Cloud–edge collaborated AIoT online adaptive battery ROM framework. • Physics-informed ROM incremental learning algorithm to track the real system dynamics. [ABSTRACT FROM AUTHOR]

Renewable-dominated power grids will require industry to run their processes in accordance with the availability of energy. At the same time, digitalization introduces new possibilities to leverage the untapped optimization potential to provide this flexibility. Mathematical optimization methods such as mixed-integer linear programming (MILP) are widely used to predict optimal operation plans for industrial systems. MILP models are difficult to adapt, but the viability of the predicted plans relies on accurate underlying models of the actual behavior. New automation paradigms, such as the digital twin (DT), can overcome these current drawbacks. In this work, we present the implementation and experimental evaluation of several micro-services on a standardized five-dimensional DT platform that automate MILP model adaption and operation optimization. These micro-services guarantee that, (1) deviations between the physical entity and its virtual entity models are detected, (2) the models are adapted accordingly, (3) subsequently linearized to suit the MILP approach and (4) used for live operational optimization. These novel services and DT workflows that orchestrate them were experimentally tested with a packed bed thermal energy storage (PBTES) test rig that acts as a physical entity. A waste heat recovery use case in steel production is used as the evaluation scenario. While the model error of a static simulation model would increase to 60% over 7 days of operation, the model error remains well below 25% as a result of successful model adaption. The prediction error of the optimization model remains in a typical magnitude of 10 to 20% during the evaluation period, despite the degradation of the PBTES power. • Digital-twin (DT) based approach for adaptive energy system modeling and optimization. • Novel DT services for automated simulation model adaption and linearization. • Use case of waste heat recovery in green steel production operation planning via MILP. • Experimental live operation on a packed bed thermal energy storage test rig. [ABSTRACT FROM AUTHOR]

An accurate estimation of dynamic loads and structural dynamic responses is deemed an indispensable prerequisite for developing trustworthy digital twin (DT) models of dynamically excited structures. Traditional joint load–response estimation (JLRE) algorithms necessitate a full-rank feedthrough matrix, implying that accelerometers are required at all degrees of freedom with input loads. If an unknown input is a bending moment or torque, the rotational acceleration must be measured, which is usually difficult, if not impractical, in real applications. Therefore, the applications of the traditional JLRE in wind turbines (WTs) are rather limited due to the complex excitation mechanism. However, the unified linear input and state estimator (ULISE) algorithm presented in this paper eliminates this constraint. This paper experimentally tested this algorithm on an operating 1:50 scaled WT model. The influences from blade rotation were considered. A comprehensive structural health monitoring (SHM) sensing system was deployed on the WT tower to measure the tower's dynamic responses. A camera system, which integrated digital image correlation (DIC) technology and binocular stereo vision technology, was also adopted. The unknown excitations and unmonitored dynamic responses of the WT tower were estimated simultaneously using the ULISE algorithm. The reconstructed strain and acceleration responses achieved high accuracy, with a normalized root mean square error less than 9%. The blade rotation had a notable impact on the WT tower, primarily manifested as a bending moment whose dominant frequency was corresponding to the blade's rotational speed. Such joint estimations could function as the virtual sensing of unknown inputs and responses in the DT model of the WT tower, which will enable DT-based online remote structural condition assessment and preventive maintenance in the future. • Feasibility of digital twin technology for wind turbine towers is experimentally verified. • Joint load-response estimation for an operating wind turbine. • Excitation mechanism of blades on the wind turbine tower is analyzed. • Feasibility of using computer vision technology for wind turbine monitoring is discussed. [ABSTRACT FROM AUTHOR]

Sustainable use of water and energy sources is a crucial challenge for smart and resilient urban water infrastructure. At this aim, this study presents a methodology for implementing and selecting pumps-as-turbines (PaTs) in water distribution networks (WDNs) to minimise leakage losses and maximise energy recovery. A novel dynamic control algorithm for PaTs is proposed, which optimises the pressure dissipation while granting a minimum pressure head in the network. The digital hydraulic model of a WDN, which incorporates the developed control strategy, outputs the available pressure heads and flow rates of each PaT for an extended period simulation. The entire simulated data series are fed into a new algorithm to select the best machine to be installed to achieve a reliable solution under realistic and varying conditions. Finally, this methodology is applied to a real test case, improving the sustainability and energy efficiency of a mountainous WDN by enhancing the water-energy nexus. • Novel algorithm for dynamic pressure control in water supply systems simulations. • Innovative algorithm for optimal pumps-as-turbines selection. • Sustainable and efficient water-energy nexus in water distribution systems. • Methodology applied on a digital twin of an existing water supply system. [ABSTRACT FROM AUTHOR]

Lithium-ion cells are subject to degradation due to a multitude of cell-internal aging effects, which can significantly influence the economics of battery energy storage systems (BESS). Since the rate of degradation depends on external stress factors such as the state-of-charge, charge/discharge-rate, and depth of cycle, it can be directly influenced through the operation strategy. In this contribution, we propose a model predictive control (MPC) framework for designing aging aware operation strategies. By simulating the entire BESS lifetime on a digital twin, different aging aware optimization models can be benchmarked and the optimal value for aging cost can be determined. In a case study, the application of generating profit through arbitrage trading on the EPEX SPOT intraday electricity market is investigated. For that, a linearized model for the calendar and cyclic capacity loss of a lithium iron phosphate cell is presented. The results show that using the MPC framework to determine the optimal aging cost can significantly increase the lifetime profitability of a BESS, compared to the prevalent approach of selecting aging cost based on the cost of the battery system. Furthermore, the lifetime profit from energy arbitrage can be increased by an additional 24.9% when using the linearized calendar degradation model and by 29.3% when using both the linearized calendar and cyclic degradation model, compared to an energy throughput based aging cost model. By examining price data from 2019 to 2022, the case study demonstrates that the recent increases in prices and price fluctuations on wholesale electricity markets have led to a substantial increase of the achievable lifetime profit. • Open-source framework for designing aging aware operation strategies. • Increased lifetime profit by determining the optimal aging cost. • Additional 29.3% higher lifetime profit through linearized degradation models. • Case study focussed on energy arbitrage on the intraday electricity market. • Recent electricity price volatility caused substantial increase in lifetime profit. [ABSTRACT FROM AUTHOR]

Low voltage distribution networks deliver power to the last mile of the network, but are often legacy assets from a time when low carbon technologies, e.g., electrified heat, storage, and electric vehicles, were not envisaged. Furthermore, exploiting emerging data from distribution networks to provide decision support for adapting planning and operational strategies with system transitions presents a challenge. To overcome these challenges, this paper proposes a novel application of digital twins based reinforcement learning to improve decision making by a distribution system operator, with key metrics of predictability, responsiveness, interoperability, and automation. The power system states, i.e., network configurations, technological combinations, and load patterns, are captured via a convolutional neural network, chosen for its pattern recognition capability with high-dimensional inputs. The convolutional neural networks are iteratively trained through the fitted Q-iteration algorithm, as a batch mode reinforcement learning, to adapt the planning and operational decisions with the dynamic system transitions. Case studies demonstrate the effectiveness of the proposed model by reducing 50% of the investment cost when the system transitions towards the winter and maintaining the power loss and loss of load within 5% compared to the benchmark optimisation. Doubled power consumption was observed in winter under future energy scenarios due to the electrification of heat. The trained model can accurately adapt optimal decisions according to the system changes while reducing the computational time of solving optimisation problems, for a range of scales of distribution systems, demonstrating its potential for scalable deployment by a system operator. • Key features of network configurations, technology installations, and load patterns are digitally represented. • A novel digital twin-based distribution network model to adapt planning and operational decisions with dynamic state transitions. • Informed decisions to minimise the investment cost, power loss, loss of load, and renewable curtailment. • Synthesising scalable and computational efficient distribution networks. [ABSTRACT FROM AUTHOR]

This work develops digital entities of a commercial Fresnel Solar Collector (FSC) installed in an absorption cooling plant. The objective is to create and validate models that describe the FSC dynamics across its whole operation range during the day and the night. Thus, the temperatures range between operation temperature of 180 ° C and almost ambient temperature due to overnight heat losses. In the same sense, the flow range between zero to 13 m 3 / h. The idea is that the digital twin will aid start-up and shut-down optimization and control design reliability. The paper employs two modeling approaches, then evaluates their twinning/adaptation time and performance validation. One model uses phenomenological modeling through Partial Differential Equations (PDE) and parameters identification, and another uses a data-driven technique with Adaptive Neuro-Fuzzy Inference Systems (ANFIS). The available measurement data sets comprise 25 days of operation with a sampling time of 20 s which, after outlier removal, filtering and treatment, resulted in 108416 samples. The validation considers six separate operating days. Results show that both models can twinning/adapt considering measured data. The models present pretty good results and are suitable for control and optimization. Besides, this is the first paper considering the FSC mirror defocus action on dynamic modeling and validation. • This work validates neuro-fuzzy (NF) and differential (PDE) models with massive data. • The models generally represent the process day and night. • The models are fairly accurate and precise—worst-case MAPE of 2.49%. • The NF model has ten times faster simulation time than the PDE model. • The validated dynamic models are the first accounting with mirror's focus action. [ABSTRACT FROM AUTHOR]

• Places the Phase Change Material in the indoor building structure. • The DTs model of the Phase Change Wall (PCW) structure is constructed. • The PCW structure is used to build a thermal energy storage. The purpose of this work is to explore the role of the safe and optimal scheduling of thermal energy storage systems in intelligent buildings in promoting sustainable economic development under Digital Twins (DTs) technology. Phase Change Material (PCM) has high energy density, constant temperature storage, small footprint, and long service life. Here, PCM is first placed in the indoor building structure, and the DTs technology is introduced. In the development of intelligent buildings, the data generated by the energy storage system of intelligent buildings in the real space can be mapped to the virtual space in real time for simultaneous analysis. In addition, the PCM wall structure and thermal network DTs model are designed for the intelligent building. In addition, the PCW structure is used to build a thermal energy storage and dispatch model of the smart thermoelectric building based on DTs. Finally, the model is evaluated and analyzed experimentally. The analysis of system optimization power under different schemes indicates that the scheduling operation strategy of thermal energy storage of building walls can avoid overcharging or over-discharging batteries in the microgrid and reduce battery power consumption. Besides, the building wall energy storage capacity is always in the range of 0.2 ∼ 0.8 on the all-weather scale. Moreover, the model constructed here achieves significantly lower economic costs, environmental costs, and energy costs and a better energy-saving effect than the existing model. The model built here can serve as experimental reference for further digital energy storage in intelligent buildings and comprehensive energy utilization because of its superior safety performance and lower consumption. [ABSTRACT FROM AUTHOR]

Large-scale grid-connected lithium-ion batteries are increasingly being deployed to support renewable energy roll-out on the power grid. These battery systems consist of thousands of individual cells and various ancillary systems for monitoring and control. Although many studies have focused on the behaviour of single lithium-ion cells, the impact of system design choices and ancillary system controls on long-term degradation and efficiency of these systems, containing thousands of cells, has rarely been considered in detail. Here, we simulate a 1 MWh grid battery system consisting of 18,900 individual cells, each represented by a separate electrochemical model, as well as the thermal management system and power electronic converters. Simulations of the impact of cell-to-cell variability, thermal effects, and degradation effects were run for up to 10,000 cycles and 10 years. It is shown that electrical contact resistances and cell-to-cell variations in initial capacity and resistance have a smaller effect on performance than previously thought. Instead, the variation in degradation rate of individual cells dominates the system behaviour over the lifetime. The importance of careful thermal management system control is demonstrated, with proportional control improving overall efficiency by 5%-pts over on–off methods, also increasing the total usable energy of the battery by 5%-pts after 10 years. • Simulation study of 1 MWh grid battery system with 18900 cells cycled for 10 years. • All cells modelled with individual electrochemical models including degradation. • Cells thermally and electrical coupled and supported by ancillary systems. • Exploration of cell variability, interconnection resistance, and thermal management. • Results show key influence of thermal management design and control on lifetime. [ABSTRACT FROM AUTHOR]

We describe the integration of Bayesian decision theory in a digital twin framework to provide a process optimization tool for a biomass boiler. Our application is the Atikokan Generating Station, a 200 MW, biomass-fired tower boiler operated by Ontario Power Generation in Ontario, Canada. For this analysis, we use all available prior information as well as data from the biomass plant and from science-based models. Our objective is to determine a single-valued operational setpoint for the boiler that satisfies a set of objectives/constraints while accounting for uncertainty in boiler operations and in boiler measurements. This setpoint is then continuously updated at the frequency required by plant operations to provide dynamic control. This process of decision-making under uncertainty is a form of artificial intelligence and provides a formal methodology for making optimized decisions in complex systems. Our methodology consists of defining the decision space where all possible solutions reside, identifying the probability of outcomes given that a specific decision was made, creating a decision/cost model that relates the quantities of interest (QOIs) in the physical system (e.g. gross power output) to the decision QOIs (e.g. dollars), identifying the utility (the value to the user) of each outcome, and maximizing the expected utility (i.e. the decision). Once the decision (operational setpoint) is computed, we can predict all of the QOIs (boiler efficiency, O 2 concentration at the outlet, etc.) at the decision point from the science-based model using the parameter distributions computed as part of the Atikokan Digital Twin. • Digital twin of biomass boiler incorporates formal Bayesian decision theory to provide optimized operating-parameter set points which account for uncertainty. • Cost model translates the physics quantity to the decision quantity (often cost). • Utility maps the cost model to the value of the decision to the decision maker. • Optimal decision results from maximizing the expected utility. • Digital twin computes 790 outputs and their distributions at the optimal decision. [ABSTRACT FROM AUTHOR]

This paper describes a methodology based on the digital-twin concept applied to the condition monitoring of three-phase power electronics converters. The proposed solution does not need additional hardware circuits or sensor calibration. The digital replica acts as a permanent watchdog, working in parallel with the converter control system and providing in real-time the signals to be monitored. An application for a Three Leg Neutral Point Clamped (3L-NPC) converter is demonstrated including theoretical analysis for the digital model, practical considerations about the real-time execution of the virtual representation, and experimental verification. The combination of two optimization algorithms (Particle Swarm Optimization (PSO) and a genetic algorithm (GA)) working in series is employed to estimate the circuit parameters of interest, specifically the degradation of the output LC filter. The results shown in this paper serve as a first step for attaining a low-cost and robust parallel digital-twin model as a basis for its application to the concepts of monitoring and reliability. • Power electronics provide an excellent solution due to their energy efficiency conversion. • A real-time Digital Twin system is proposed as a condition monitoring solution. • The parameters estimation is performed through a combination of optimization algorithms. • The whole system has been validated in simulation and on a real converter. [ABSTRACT FROM AUTHOR]

With the growing percentage of the intermittent renewable power generation, the energy system is under increasing pressure in balancing the supply and the demand. As a major part of urban energy consumptions, buildings can provide considerable regulation flexibility for the energy system by actively managing their energy demands. For the building energy flexibility (BEF) provided by thermostatically controlled loads (TCL), its dispatch performance is vulnerable to the building thermal parameter errors, and in some cases, occupants need to provide the critical information related to the indoor temperature state and the occupancy state to the energy management system outside buildings, which decreases the availability of the BEF and raises privacy concerns. For these issues in the BEF utilization, this paper proposes a data-model fusion dispatch strategy based on the digital twin (DT). The proposed strategy is capable of parameter fault tolerance and privacy protection by combining the model-free advantage of the data-driven method with the analytical optimization advantage of the model-driven method. Firstly, a DT-based BEF dispatch framework is proposed. Secondly, the building DT is established by combining the building thermal dynamics (BTD) data-driven model and the TCL operation mechanism model. And the building response deduction is carried out based on the DT. Finally, under the rolling optimization framework, the data-model fusion dispatch strategy is devised by uniting the DT deduction and the optimization constructed by the BTD mechanism model, in which the multi-dimensional modeling of the BTD is carried out from the state dimension and the energy dimension. The simulation results show that the optimization result can reach 98.4% of the ideal result under the scenario with 15% parameter random error, and 98.3% of the ideal result under the scenario with 15% random state noise injection. • Data-model fusion is realized by uniting the digital twin deduction and the mechanism. • The building thermal dynamics are modeled from both the state & the energy dimensions. • The fusion dispatch is capable of parameter fault tolerance and privacy protection. • Optimized result can reach 98.4% of ideal situation under 15% random parameter errors. • Optimized result can reach 98.3% of ideal situation under 15% random state noises. [ABSTRACT FROM AUTHOR]

• Multiple utility generation incorporating cold energy recovery system is proposed. • Individual component performance degradation is considered. • A digital twin approach is applied for real-time and life-cycle optimization. • Cold energy recovery unit yields 0.72 % improvement of average daily PESR. • Digital twin driven-optimization achieves 1.37 % average daily PESR advancement. Natural gas is expected to be the dominant fossil fuel in the coming decades. Improving the sustainability of natural gas usage is imperative to achieving a low-carbon society. This study proposes a combined cooling, heating, and power incorporating cold energy recovery (CCHP-CER) system to utilize both heat and cold energies of liquified natural gas (LNG) in a cascade way. The system is comprised of four subsystems, namely, gas turbine, water-lithium bromide absorption chiller, hot water heat exchanger, and cold energy recovery unit. A digital twin approach is applied to this system for real-time and life-cycle operational optimization. The cascade forward neural network (CFNN) is employed to construct the virtual representation while a parameter-free intelligent algorithm is adopted to seek the optimal operating parameters. Key results from this study revealed that incorporating the cold energy recovery (CER) unit produces additional electricity and cooling effect, bringing a 0.72 % improvement in the average daily primary energy saving rate (PESR). The digital twin-based optimization process updates the optimal operation parameters in time when the system suffers degradation. Consequently, the degradation performance is alleviated by the living parameters. Compared to static model-based optimization, the digital twin-based optimization improves the daily PESR by 2.23 %, 0.35 %, and 1.53 % during respective winter, summer, and transition days, particularly when the compressor and turbine of the gas turbine suffer degraded efficiency of −2 %. [ABSTRACT FROM AUTHOR]

• A novel 3D multi-physics field digital twin model for PEMFCs is proposed. • The computational fluid dynamic technique is integrated in the digital twin model. • The model is demonstrated within twenty randomly selected working conditions. • The proposed model can predict PEMFC physics field characteristics in 0.913 s. In times of the commercialization process of proton exchange membrane fuel cells (PEMFCs), a full knowledge of in-situ state in PEMFCs is of critical significance to the in-situ operational process and the evaluation of material stage and potential damage. The conventional experimental observation and in-situ prediction models can only obtain very limited information while the computational fluid dynamics approach takes too long time to get the detailed information. To reach a full knowledge of PEMFC real-time state, a novel 3D multi-physics digital twin model for PEMFCs is proposed based on the proper orthogonal decomposition (POD) method. In the model, firstly, for one kind of PEMFC, 139 ex-situ snapshots are designed and simulated based on the three-dimensional two-phase non-isothermal numerical model with the assumption of liquid pressure continuity in the whole membrane electrode assembly. Then the modes of each field in snapshots are extracted by singular value decomposition method using Jacobi algorithm. Finally, the coefficients in the POD prediction equation are obtained by using the multivariate adaptive regression splines. The digital twin results of voltage, temperature, membrane water content and liquid water saturation fields are exhibited and analyzed. Results suggest that for the studied PEMFC, the digital twin technique can capture the global values and the local distribution characteristics of each above physical fields well in 0.913 s. The mean global deviations of the above four fields of 20 groups of random conditions within wide current density and operational condition ranges are 5.7 %, 1.3 %, 8.9 % and 12.0 % respectively. Even though the practical results can only be applied for the studied PEMFC, the proposed methodology has its general application range. [ABSTRACT FROM AUTHOR]

• A model using graph neural network is proposed to predict state points and performance metrics. • A configuration representation method based on thermodynamic graph is developed. • GNN can extract structure features from different graphs of three SCO 2 power systems. • GNN achieves excellent accuracy and efficiency and outperforms classical data-driven models. Considering the increasing energy consumption and greenhouse gas emissions, the Supercritical CO 2 (S-CO 2) power system has attracted more and more attention. Due to the expensive computation resource and time cost, data-based solutions for performance prediction are urged. The surrogate model by machine learning is a promising alternative, but it only focuses on the objective functions and ignores the importance of topological structures and physical states of cycles. Aiming at providing a comprehensive model to predict physical states as well as thermodynamic characteristics, a deep learning method based on graph neural network (GNN) are devised in this paper. With the modeling calculation results as training dataset, a well-trained model can accurately reconstruct the physical states consisting of temperature, pressure, enthalpy, entropy (relative error of most samples <5 %) and exergy as well as thermal and exergy efficiency (relative error of most samples <5 %). Moreover, this model shows superior performance compared with traditional machine learning models including Regression Tree, Artificial Neural Network (ANN), Support Vector Machine (SVM) and Gaussian Process Regression (GPR). Finally, the comparison between different training sizes demonstrate the model can help reduce sampling costs for complex systems. Overall, the presented deep learning model can provide a reliable and competitive choice for the digital twin of S-CO 2 power system and other power systems. [ABSTRACT FROM AUTHOR]

By constructing digital twins (DT) of an integrated energy system (IES), one can benefit from DT's predictive capabilities to improve coordinations among various energy converters, hence enhancing energy efficiency, cost savings and carbon emission reduction. This paper is motivated by the fact that practical IESs suffer from multiple uncertainty sources, and complicated surrounding environment. To address this problem, a novel DT-based day-ahead scheduling method is proposed. The physical IES is modelled as a multi-vector energy system in its virtual space that interacts with the physical IES to manipulate its operations. A deep neural network is trained to make statistical cost-saving scheduling by learning from both historical forecasting errors and day-ahead forecasts. Case studies of IESs show that the proposed DT-based method is able to reduce the operating cost of IES by 63.5%, comparing to the existing forecast-based scheduling methods. It is also found that both electric vehicles and thermal energy storages play proactive roles in the proposed method, highlighting their importance in future energy system integration and decarbonisation. • A deep learning and digital twins (DT) based scheduling method is proposed. • Physical IES performance is addressed by its virtual replica's predictive capability. • A machine learning method is proposed to address multiple uncertainty challenges. • Historical U.K. datasets are used for performance evaluations. • Results demonstrate the Digital Twin model reduces the extra operating cost by 63.5%. [ABSTRACT FROM AUTHOR]