Forecasting-enabled microgrid optimization for adaptive renewable energy planning in remote coastal communities

Remote coastal communities in the Sundarbans of Bangladesh depend on an unreliable distribution grid, yet microgrid planning studies for such sites usually treat demand and renewable generation as fixed inputs and rarely show what forecast quality means for operation. This study examines whether a grid-connected hybrid microgrid can be both least-cost and resilient for 150 households and three community buildings in Dhangmari, and what operational value hourly forecasting adds to such a design. Long-term NASA POWER resource data and a survey-based demand of 355.3 kWh/day were used in HOMER Pro to compare four configurations, followed by sensitivity analyses of resource, economic, grid-outage and battery-cost parameters and a screening of multi-day outages. Seven machine learning and deep learning models were then benchmarked for hourly load and renewable power forecasting under a chronological, leakage-free protocol, and their errors were converted into battery reserve requirements. The optimal system, combining 102 kW of photovoltaics, 45 kW of wind turbines and 150 kWh of battery storage with the grid, achieved a net present cost of $71,559, an energy cost of $0.0254/kWh and a renewable fraction of 73.4%. Unmet load stayed below 0.08% for 240–560 grid failures per year, and critical loads were fully served during 72-hour outages unless wind and solar generation were both lost. Renewable power was predicted with a coefficient of determination of 0.998 when weather at the forecast hour was known, falling to about 0.94 with 20% weather errors, whereas load accuracy (about 0.81) was bounded by the random variability of the demand profile; the best model, a stacking ensemble, reduced the hourly battery reserve by about one third relative to linear regression. The framework provides a reproducible basis for planning resilient, forecast-aware microgrids in climate-exposed coastal settlements.

Authors

Institutions

Publication Details

Journal
Next Energy
Published
2026-11-14
DOI
https://doi.org/10.1016/j.nxener.2026.101198
Primary Topic
Microgrid Control and Optimization
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Forecasting-enabled microgrid optimization for adaptive renewable energy planning in remote coastal communities

Md. Masudur Rahman, Prapti Das, Md. Feroz Ali, Diponkor Kundu et al.
Next Energy
Microgrid Control and Optimization
article

Forecasting-enabled microgrid optimization for adaptive renewable energy planning in remote coastal communities

Md. Masudur Rahman, Prapti Das, Md. Feroz Ali, Diponkor Kundu, Md. Galib Hasan, Taib Hasan, Md Alamgir Hossain, S M Shakil, A.H.M. Iftekharul Ferdous
article en

Abstract

Remote coastal communities in the Sundarbans of Bangladesh depend on an unreliable distribution grid, yet microgrid planning studies for such sites usually treat demand and renewable generation as fixed inputs and rarely show what forecast quality means for operation. This study examines whether a grid-connected hybrid microgrid can be both least-cost and resilient for 150 households and three community buildings in Dhangmari, and what operational value hourly forecasting adds to such a design. Long-term NASA POWER resource data and a survey-based demand of 355.3 kWh/day were used in HOMER Pro to compare four configurations, followed by sensitivity analyses of resource, economic, grid-outage and battery-cost parameters and a screening of multi-day outages. Seven machine learning and deep learning models were then benchmarked for hourly load and renewable power forecasting under a chronological, leakage-free protocol, and their errors were converted into battery reserve requirements. The optimal system, combining 102 kW of photovoltaics, 45 kW of wind turbines and 150 kWh of battery storage with the grid, achieved a net present cost of $71,559, an energy cost of $0.0254/kWh and a renewable fraction of 73.4%. Unmet load stayed below 0.08% for 240–560 grid failures per year, and critical loads were fully served during 72-hour outages unless wind and solar generation were both lost. Renewable power was predicted with a coefficient of determination of 0.998 when weather at the forecast hour was known, falling to about 0.94 with 20% weather errors, whereas load accuracy (about 0.81) was bounded by the random variability of the demand profile; the best model, a stacking ensemble, reduced the hourly battery reserve by about one third relative to linear regression. The framework provides a reproducible basis for planning resilient, forecast-aware microgrids in climate-exposed coastal settlements.

Next EnergyVol. 14
Pabna University of Science and Technology (BD), Dhaka University of Engineering & Technology (BD)
Microgrid Control and Optimization
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon