Energy systems engineer. I build open, tested models of energy systems and use them to answer design questions.

My work sits where physical models meet data: batteries and electrolysers, buildings and multi-energy neighbourhoods, distribution grids and wind turbines, from component physics to optimisation, machine learning and validation against measurements. I am on the Erasmus Mundus DENSYS Master's (Université de Lorraine, now Politecnico di Torino) and hold a first-class BEng in Renewable Energy Engineering from the University of Exeter.

Looking for: a six-month Master's thesis internship from March 2027 (mid-February possible), with a PhD as the next step.

muscat-energy-hub · How much model detail does a sizing study need?
A district-cooled neighbourhood with PV, battery, chiller, chilled-water tank and a PEM electrolyser, sized with a simple and a detailed model of every component. Most details barely move the design; the buildings' thermal mass cuts the optimal tank by 39%. Includes new, tested OMEGAlpes units for a chiller and an electrolyser.
Pyomo HiGHS OMEGAlpes uncertainty analysis

building-digital-twin-kalman · A digital twin kept in step with a real building
Grey-box thermal models of 23 office zones identified from 18 months of open measured data (CU-BEMS), run online as an extended Kalman filter. The synchronised twin forecasts, detects faults and tests demand-response scenarios; a hybrid with gradient boosting beats both physics and machine learning alone.
Kalman filter grey-box models XGBoost demand response

lv-grid-copper-or-flexibility · Copper or flexibility? Long-term LV planning
Three SimBench feeders planned from 2025 to 2045 with full-year power flows: cables and transformers against §14a control, a battery and a tap-changing transformer, compared on cost, life-cycle CO2, metals and regret with WSM, TOPSIS, PROMETHEE II and SMAA-2.
pandapower LCA MCDA SMAA

wind-fleet-instability-monitoring · Turbine loads from SCADA data
Detection of edgewise blade instabilities and their fatigue cost across a fleet, and a machine-learning load emulator for power de-rating, checked against real 20 Hz measurements of a research turbine.
SQL SciPy XGBoost rainflow

hybrid-energy-storage-pybamm · From cell physics to a sized storage system
A PyBaMM model calibrated on experimental data, reduced to a fast equivalent circuit, and used to size a battery and supercapacitor system for frequency reserve, re-checked on a year of measured grid frequency. Exported as an FMI 2.0 FMU.
PyBaMM FMI co-simulation

residential-microgrid-ems · Home energy management, end to end
PV, battery, bidirectional EV and heat pump: MILP control, annual sizing, forecast-driven MPC, coupling with AC load flow on a feeder, and automated tests of a battery controller.
Pyomo XGBoost pandapower

• ptx-methanol-surrogate: physics-constrained machine-learning surrogates of a power-to-methanol reactor, trained on noisy operating logs simulated with a kinetic plant model.
• pv-pem-electrolyser-degradation: a year of PV-driven PEM electrolyser performance and ageing in Oman and the UK; code for an Energies Editor's Choice article.
• energy-time-series-transfer-learning: household PV and load forecasting with LSTM and Transformer models, transfer learning, and EV and heat-pump detection from meter data.
• AEM_Case_based: electrochemical and thermal model of an AEM water electrolyser, calibrated on polarisation and EIS data (team project; my part was the electrochemical model).

• Al-Mandhari, M., Ghosh, A. (2025). Modelling solar intermittency effects on PEM electrolyser performance and degradation: a comparison of Oman and UK. Energies 18, 6131. doi · Editor's Choice
• Al-Mandhari, M. et al. (2026). Floating photovoltaic-powered green hydrogen for decarbonization of the energy-consuming sectors in the United Kingdom. Energies 19, 2931. doi
• Sanders, C. et al., incl. Al-Mandhari, M. (2026). From pasture to platform: integrating agrivoltaics and green hydrogen for the Penzance-London railway corridor. Solar Energy 318, 115064. doi
• Hussain, S.N., Al-Mandhari, M. et al. (2026). GIS-driven regional assessment for sustainable data center siting in the United Kingdom. Land 15, 516. doi

More on ORCID.

Modelling and optimisation: Python, Pyomo, OMEGAlpes, HiGHS, Gurobi, PyBaMM, pandapower, SimBench, MATLAB/Simulink, FMI
Decision support: multi-criteria analysis (WSM, TOPSIS, PROMETHEE, SMAA), life-cycle assessment
Data and machine learning: pandas, SQL, scikit-learn, XGBoost, PyTorch, TensorFlow, uncertainty and sensitivity analysis
Energy tools: pvlib, PVsyst, HOMER Pro, QGIS, openLCA, GaBi

LinkedIn · ORCID · e66840u@etu.univ-lorraine.fr

Arabic (native) · English (professional) · French (A2, learning)