$1,800.27
Optimized weekly cost
Synthetic tariff; down from $4,169.94 grid-only
Engineering case study
Solar + Battery Microgrid Energy-Management Simulator
Seven-day commercial dispatch benchmark comparing grid, solar, rule-based storage, and mixed-integer optimized storage.
- Executed simulation
- Automated test
Problem statement
Determine how control strategy changes operating cost, peak demand, renewable utilization, battery cycling, and outage performance while maintaining interval-by-interval power and energy balance.
A reproducible single-bus simulator that schedules PV, battery, and grid power under time-of-use prices, demand charges, equipment limits, and optional outages.
$2,369.67
Modeled savings
Optimized case versus grid-only over 168 hours
43.35 kW
Optimized peak import
Versus 101.01 kW in the grid-only case
890.90 kWh
Battery throughput
Active cycling under the simplified wear proxy
Project scope
- 168 hourly intervals using a fixed-seed synthetic commercial load and PV profile
- Grid-only, solar-only, causal rule-based battery, and perfect-forecast MILP cases
- Energy charges, demand charges, export credit, throughput cost, and unserved-energy penalty
- Single AC bus; no feeder voltage, reactive power, transient, or protection model
System or processing architecture
PV, a bidirectional battery inverter, the utility interconnection, and facility load meet at one modeled AC bus. A controller chooses battery and grid setpoints while the validation layer independently checks power balance and state-of-charge recursion.
My engineering contributions
- Defined the sign conventions, SOC dynamics, dispatch limits, and cost model.
- Implemented both a causal self-consumption controller and a SciPy/HiGHS mixed-integer optimizer.
- Added binary charge/discharge exclusivity, outage isolation, peak-demand tracking, and a terminal-SOC constraint.
- Built sensitivity studies, CSV/XLSX exports, plots, and automated engineering checks.
Important design decisions
- Used a binary battery-mode variable so simultaneous charging and discharging are structurally impossible.
- Required final SOC to meet or exceed initial SOC to prevent artificial end-of-horizon savings.
- Kept the rule-based controller causal and the optimizer perfect-forecast so they serve distinct operational and benchmark roles.
Engineering methods
- Hourly AC-bus energy balance
- Efficiency-adjusted battery SOC recursion
- Mixed-integer linear optimization
- Fixed-seed scenario generation
- Sensitivity analysis for PV, storage, tariffs, demand charges, SOC, and outage duration
Selected evidence
Readable results, with context
Verification and testing
- Power-balance and SOC-dynamics residuals remain near numerical precision for all four controllers.
- SOC stays within 16-144 kWh for the 160 kWh battery configuration.
- Automated tests verify balance, bounds, mutual exclusivity, and zero grid exchange during outage intervals.
- Fixed seed 321 and machine-readable run metadata make the published week reproducible.
Honest limitations
- Synthetic hourly data and a short billing window; savings are scenario estimates, not a utility bill forecast.
- The optimizer assumes perfect foresight and is not a deployed real-time controller.
- Fixed efficiency and linear throughput cost do not model electrochemical degradation or temperature.
- No reactive power, network voltage, inverter kVA, interconnection, or hardware behavior is represented.
Tools and technologies
- Python
- SciPy HiGHS MILP
- pandas
- NumPy
- Matplotlib
- Excel
Skills demonstrated
- Microgrid controls
- Optimization
- Battery modeling
- Energy economics
- Scenario analysis
- Model validation
Downloadable proof of work
Review the evidence directly
Continue in Energy & Grid Systems