Machine Learning and Optimization Model Development for Northern Community Energy Planning

Many Arctic communities are currently transitioning away from diesel-based electricity to renewable energy systems, a challenge which requires resource assessment of suitable generation technologies as well as community energy planning.While power estimation models are readily available for wind and solar PV, few accessible models exist for hydrokinetic power prediction, a river-based technology that is currently being piloted in Alaska.To improve hydrokinetic resource assessment, Chapter 2 in this paper develops a predictive model through a machine learning framework to remotely estimate stream velocity.A Random Forest model was found to outperform the traditional Manning equation approach by estimating velocity on a small dataset with a mean absolute percent error of 22%, a 52% improvement in prediction accuracy over the Manning equation, which reported a mean absolute percent error of over 46%.A more generalizable model that was trained on a larger dataset and included additional, novel geometric input variables was found to predict stream velocity with a mean absolute percent error of 24%.This model demonstrated advantages over existing models that either required in-situ data collection or were not compatible with smaller streams suitable for community-level energy planning.Chapter 3 in this thesis develops a new, multi-phase, mixed integer linear programming generation expansion planning model, to plan a comprehensive energy system for a case study community that meets both thermal and electricity demand through a single electric load profile.It was found that a renewably-powered system consisting of wind, solar, and battery storage units had an annualized cost of $17,849, using a 100 kW commercially available wind turbine.This was found to be economically superior by saving up to $3,457 per year when compared to a system with diesel, which would cost $21,306 if fuel prices v NomenclatureThe following nomenclature is used.For ease of reading, the nomenclature is separated by those introduced in Chapter 2 and Chapter 3. 1.1 Nomenclature introduced in Chapter 2 Symbols A Cross-sectional area of turbine blade (m 2 ) Cp Power coefficient of turbine l bends Number of obtuse angle bends in stream within a 2500m radius of the site m5000,up; m2500,up; m500,up Surface slope measured upstream from site across distances of 5000m, 2500m, and 500m, respectively n Manning's roughness parameter n bends Number of bends in stream within a 2500m radius of the site P Power generated by turbine (kW) R Hydraulic radius (m) S Surface slope of stream site bend t Indicates if site is situated in an acute stream bend site bend l Indicates if site is situated in an obtuse stream bend t bends Number of acute angle bends in stream within a 2500m radius of the site V Stream velocity (m/s) z5000,dn; z2500,dn; z500,dn Elevation at points 5000m, 2500m, and 500m downstream from site (m) z5000,up; z2500,up; z500,up, z0,up Elevation at points 5000m, 2500m, 500m, and 0m upstream from site (m) Abbreviations ADCP Acoustic doppler current profiler DEM Digital elevation model MAE Mean absolute error MAPE Mean absolute percent error ML Machine learning NWIS National Water Information System RF Random Forest RMSE Root mean squared error vi SHAP Shapley Additive Explanations USGS Unites States Geological Survey VMT Velocity Mapping Toolbox Greek Letters ρ Water density (1000 kg/m 3 ) 1.2 Nomenclature introduced in Chapter 3 Symbols B Set for building type (cabin, small hotel, microbrewery) CCrp Annualized capital costs of technology, r, in phase, p ($/kW) cerp Number of new capacity additions made for each technology, r, and in each investment phase, p Cp Specific heat capacity of water at 50 o C (0.001162 kWh/kg K) crp Total capacity of each technology, r, installed in phase, p (kW) D Number of days per month FCrp Fixed O&M costs ($/kW) grhp Power produced per generation technology, r, in hour, h, in phase, p (kW) H Set for hour index {0, 1, …, 23} Lhp Electric demand profile (kW) Mer Maximum limit of expansion of technology, r (kW) MPrh Marginal production of technology, r, in hour, h P Set for investment phases {1,2,3} P_brewhp Electric load for microbrewery equipment in hour, h, phase, p (kW) P_elechbp Electric load for non-thermal needs in hour, h, building, b, phase, p (kW) P_HWhbp Electric load for hot water heater in hour, h, building, b, phase, p (kW) P_max Charge/discharge rate limit of the BESS (kW) P_SHhbp Electric load for hydronic space heating system in hour, h, building, b, phase, p (kW) Qhbp Space heating thermal demand in hour, h, building, b, phase, p (kW) R Set for power generation technology type {PV, WT, HK, DG, BESS} Sgr Unit size of technology (kW) sochp The current state of charge of the BESS in hour, h, phase, p (kWh) VCrp Variable O&M costs ($/kWh) V ̇_HWhbp Demand for hot water in hour, h, building, b, phase, p (m 3 /s) V ̇_SHhbp Required volumetric flowrate of hot spring water for hydronic space heating system in hour, h, building, b, phase, p (m 3 /s) vii

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.