Multi-Microgrid Vessim Test Environment

Note: This README was created with assistance from an LLM.

A comprehensive Vessim simulation environment that models three interconnected microgrids across different geographical locations to demonstrate distributed renewable energy systems with intelligent load balancing and dynamic computing load management.

Overview

This test environment simulates a distributed energy system consisting of three microgrids located in Berlin (Germany), San Francisco (USA), and Sydney (Australia). Each microgrid operates with computing systems, solar generators, and energy storage, coordinated through a global load balancing controller that optimizes energy consumption based on renewable energy availability and battery state of charge.

System Architecture

Core Components

3 Microgrids: Each with computing systems, solar generators, and energy storage
Global Load Balancing Controller: Distributes computing load across microgrids based on renewable energy availability
Solar Power Generation: Uses real-world solar data from Solcast 2022 dataset
Energy Storage: SimpleBattery components for each microgrid
Dynamic Computing Loads: AdjustableSignal-based computing nodes that can be controlled in real-time
Comprehensive Monitoring: Data collection and analysis with modular visualization

Microgrid Details

Berlin Microgrid (Germany)

Location: Berlin, Germany
Computing Load: 900W total (3 computing nodes)
- Server 1: 400W
- Server 2: 300W
- Server 3: 200W
Solar Generation: 3 solar panels (7.5kW total capacity)
- Solar Panel 1: 3.0kW
- Solar Panel 2: 2.5kW
- Solar Panel 3: 2.0kW
Energy Storage: 3.5kWh SimpleBattery
- Initial SOC: 65%
- Minimum SOC: 10%

San Francisco Microgrid (USA)

Location: San Francisco, USA
Computing Load: 1,400W total (4 computing nodes)
- Server 1: 500W
- Server 2: 400W
- Server 3: 300W
- Server 4: 200W
Solar Generation: 3 solar panels (9.0kW total capacity)
- Solar Panel 1: 3.5kW
- Solar Panel 2: 3.0kW
- Solar Panel 3: 2.5kW
Energy Storage: 4.5kWh SimpleBattery
- Initial SOC: 75%
- Minimum SOC: 15%

Sydney Microgrid (Australia)

Location: Sydney, Australia
Computing Load: 750W total (3 computing nodes)
- Server 1: 350W
- Server 2: 250W
- Server 3: 150W
Solar Generation: 3 solar panels (10.5kW total capacity)
- Solar Panel 1: 4.0kW
- Solar Panel 2: 3.5kW
- Solar Panel 3: 3.0kW
Energy Storage: 5.5kWh SimpleBattery
- Initial SOC: 70%
- Minimum SOC: 10%

Load Balancing Controller

MultiMicrogridLoadBalancer

The MultiMicrogridLoadBalancer controller implements intelligent load distribution across microgrids based on renewable energy availability and storage capacity.

Key Features:

Real-time Monitoring: Tracks solar power generation and battery state of charge across all microgrids
Availability Scoring: Calculates availability score for each microgrid (weighted 70% solar power, 30% battery SOC)
Dynamic Load Adjustment: Scales computing load from 20% to 150% of base consumption
Cross-Microgrid Coordination: Global controller that manages all microgrids simultaneously

Load Balancing Algorithm:

Solar Power Assessment: Normalizes solar generation to 0-1 scale (based on 5kW maximum)
Battery State Evaluation: Incorporates current battery state of charge
Availability Calculation: Combines solar (70%) and battery (30%) factors
Load Distribution: Adjusts computing loads proportionally across microgrids

Technical Implementation

Custom Signal System

The simulation uses a custom AdjustableSignal class to enable dynamic load control:

class AdjustableSignal(vs.Signal):
    def __init__(self, initial_value: float):
        super().__init__()
        self._value = initial_value

    def now(self, at=None):
        return self._value

    def set_value(self, value: float):
        self._value = value

This allows the load balancer to dynamically modify computing loads based on microgrid power availability.

Controller Architecture

The global load balancer is attached to the Berlin microgrid but coordinates all microgrids through a shared config dictionary.

Simulation Configuration

Environment Setup

Simulation Start: June 15, 2022 00:00:00
Duration: 48 hours
Step Size: 300 seconds (5 minutes)
Data Source: Solcast 2022 Global Dataset
Load Balancing Frequency: Every simulation step (5 minutes)

Analysis and Visualization

Modular Analysis Structure

The simulation provides separate analysis cells for each microgrid and a composite system view:

Individual Microgrid Analysis

Each microgrid has its own dedicated analysis cell with:

Computing Load: Real-time power consumption
Battery State of Charge: Storage utilization over time
Solar vs Computing Load: Renewable energy matching
Grid Power: Grid dependency and energy exchange
Summary Statistics: Key performance metrics

Composite System Analysis

Comprehensive system-wide analysis including:

Load Distribution: Computing load across all microgrids
Battery Utilization Comparison: Storage performance across locations
Grid Dependency Comparison: Grid power consumption patterns
Renewable Energy Usage: Percentage comparison across microgrids
Load Balancing Effectiveness: Controller performance metrics

Key Metrics

Performance Indicators:

Solar-to-Load Ratio: How well solar generation matches computing demand
Grid Dependency: Energy drawn from the grid (kWh)
Renewable Energy Export: Excess energy fed back to grid (kWh)
Battery Utilization: Dynamic usage patterns (standard deviation)
Load Variability: Coefficient of variation in computing loads
Peak Solar Utilization: Maximum solar power relative to average load

System-wide Metrics:

Total Grid Energy Drawn: Combined grid dependency across all microgrids
Total Renewable Energy Exported: Combined renewable energy export
Average Solar-to-Load Ratio: System-wide renewable energy utilization
Net Grid Dependency: Overall grid reliance after renewable exports

Simulation Outputs

Data Files

The simulation generates three CSV files in the results/ directory:

berlin-microgrid.csv: Berlin microgrid data
san_francisco-microgrid.csv: San Francisco microgrid data
sydney-microgrid.csv: Sydney microgrid data

Data Columns

Each CSV file contains:

Timestamp: Simulation time (datetime)
p_delta: Grid power consumption/production (W)
e_delta: Energy delta (Wh)
storage.soc: Battery state of charge
Solar columns: Individual solar panel outputs (berlin_solar_1, berlin_solar_2, etc.)
Server columns: Individual computing node loads (berlin_server_1, berlin_server_2, etc.)

Usage

Prerequisites

pip install -r requirements.txt

Running the Simulation

python multi-microgrid.py

Expected Output

Console output showing simulation progress
CSV files with detailed results in results/ directory
Interactive plots for each microgrid (if running in Jupyter)
Comprehensive system analysis and comparison charts
Detailed load balancing effectiveness metrics

Jupyter Notebook Structure

The simulation is organized into modular cells:

Setup and Configuration: Environment and microgrid setup
Simulation Execution: 48-hour simulation run
Berlin Microgrid Analysis: Individual Berlin performance
San Francisco Microgrid Analysis: Individual SF performance
Sydney Microgrid Analysis: Individual Sydney performance
Composite System Analysis: System-wide performance and metrics

Extensibility

The system is designed for easy extension:

Add New Microgrids: Follow the existing pattern with location-specific solar data
Modify Controller Logic: Implement different optimization strategies
Integrate Additional Sources: Add wind, hydro, or other renewable sources
Implement Advanced Storage: Replace SimpleBattery with more sophisticated models
Add Real-time APIs: Integrate with actual power management systems

Future Enhancements

Predictive Control: Machine learning-based load forecasting
Real-time Integration: API connections to actual microgrid systems
Advanced Storage Models: More sophisticated battery and storage modeling
Weather Integration: Real-time weather data for improved solar forecasting
Carbon Intensity: Integration with grid carbon intensity data for carbon-aware computing

birnbaum/multi-microgrid