Skip to main content
Research Q&A
66 main questions · 539+ sub-questions · Linked to thesis evidence
Problem → Framework → Results → Policy
↑↓ Navigate, Esc Clear filters
C8Layer 4 / 10medium~3 min read
What software and computational infrastructure supports the framework?
Python 3.10, PyTorch 2.1, Stable-Baselines3, DVC, Docker. A100 GPUs, ~120h total training. Inference <5 min.
Key Takeaway
The framework uses Python, PyTorch, DVC, Docker; 120 GPU-hours on A100.
Detailed Explanation
The framework operates in a current, reproducible computing stack: **Programming languages**: Python 3.10 (mainly) for a performance-critical part in C++/CUDA for GPU acceleration. The neural networks are built using PyTorch 2.1 (ENSM GRU architecture, MAPPO actor-critic networks), which provides automatic differentiation to calculate the loss gradient for the physics-informed part. It is custom implementation of MAPPO, which is an open-source framework with multi-agent extensions, centralized training-decentralized execution (CTDE) framework. It extends the codebase within Stable-Baselines3. DVC is a data management tool that performs reproducible data pipeline functions for one's dataset and saves versioning and preprocessing transformation. For static/interactive visualisation use Matplotlib, Seaborn, Plotly, and custom D3.js interface for policy stakeholder demonstration. Containerization: Docker containers provide execution environments for different hardware platforms. **Version Control:** Git tracks code changes, GitHub enables teamwork, CI/CD tests code automatically. The training was done on NVIDIA A100 GPUs (40GB VRAM) and the total training time for the complete framework (ENSM 24h, MAPPO 72h, bidirectional integration 24h iterations) was = ~120 hours. The inference time is negligible (<5 min) for generation of policy in a 26-year scenario which enables us to instantly analyze the policy.
Why It Matters
Research significance: Enables reproducibility and replication.
Energy policy relevance: Feasible for Turkey deployment on cloud.
Methodological contribution: Docker + DVC + pinned versions.
Research Graph — Related Questions
C8 What software and computational infrastr…
Sub-questions (10)
C8.1Layer 4 / 10medium~1 min read
What is the Python version?
Python 3.10.
Key Takeaway
Current stack.
Detailed Explanation
Framework uses Python 3.10.
Why It Matters
Research significance: Stack.
Energy policy relevance: N/A.
Methodological contribution: C8.
C8.2Layer 4 / 10medium~1 min read
What is PyTorch 2.1?
Neural network framework.
Key Takeaway
Automatic differentiation.
Why It Matters
Research significance: Framework.
Energy policy relevance: N/A.
Methodological contribution: C8.
C8.3Layer 4 / 10medium~1 min read
What is Stable-Baselines3?
Open-source RL framework.
Key Takeaway
CTDE implementation.
Why It Matters
Research significance: Codebase.
Energy policy relevance: N/A.
Methodological contribution: C8.
C8.4Layer 4 / 10medium~1 min read
What is DVC?
Data version control.
Key Takeaway
Dataset hashes.
Detailed Explanation
DVC tracks datasets with cryptographic hashes for reproducibility.
Why It Matters
Research significance: Reproducibility.
Energy policy relevance: N/A.
Methodological contribution: C8.
C8.5Layer 4 / 10medium~1 min read
What is Docker?
Containerization.
Key Takeaway
Cross-platform.
Detailed Explanation
Docker provides execution environments for different hardware.
Why It Matters
Research significance: Reproducibility.
Energy policy relevance: Deployment.
Methodological contribution: C8.
C8.6Layer 4 / 10medium~1 min read
What is A100 GPU?
NVIDIA A100 40GB VRAM.
Key Takeaway
120h total.
Detailed Explanation
Training on NVIDIA A100 GPUs (40GB VRAM).
Why It Matters
Research significance: Hardware.
Energy policy relevance: N/A.
Methodological contribution: C8.
C8.7Layer 4 / 10medium~2 min read
What is ~120h total training?
ENSM 24h + MAPPO 72h + integration 24h.
Key Takeaway
~$150 cloud.
Why It Matters
Research significance: Cost.
Energy policy relevance: Feasibility.
Methodological contribution: C8.
C8.8Layer 4 / 10medium~2 min read
What is inference <5 min?
26-year scenario generation.
Key Takeaway
0.3s on CPU.
Detailed Explanation
Inference for 26-year scenario: <5 min (0.3s forward pass on CPU).
Why It Matters
Research significance: Speed.
Energy policy relevance: Policy exploration.
Methodological contribution: C10.
C8.9Layer 4 / 10medium~1 min read
What is Git/GitHub?
Version control, CI/CD.
Key Takeaway
Teamwork.
Detailed Explanation
Git tracks code; GitHub enables collaboration and CI/CD.
Why It Matters
Research significance: Workflow.
Energy policy relevance: N/A.
Methodological contribution: C8.
C8.10Layer 4 / 10medium~1 min read
What visualization tools?
Matplotlib, Seaborn, Plotly, D3.js.
Key Takeaway
Stakeholder demos.
Detailed Explanation
Visualization: Matplotlib, Seaborn, Plotly, custom D3.js for policy demos.
Why It Matters
Research significance: Communication.
Energy policy relevance: Stakeholder engagement.
Methodological contribution: C8.