Required stack
필요 기술
Pythonscikit-learnpandasnumpyPuLPOR-ToolsJupyter NotebookMachine LearningAIPredictive MaintenanceRandom ForestSVMDecision TreeData AnalysisFeature EngineeringHyperparameter TuningCross-validationOptimization AlgorithmsMonte Carlo SimulationElectrical EngineeringData Science
Project brief
프로젝트 내용
am currently working on my Master’s thesis in Electrical Engineering (Power Systems) and I am looking for an experienced freelancer to assist with the technical implementation.
The project focuses on developing an AI-based predictive maintenance framework for distribution transformers, combining machine learning models with optimization-based maintenance scheduling.
Scope of Work:
1. Machine Learning Models
Implement and refine:
Random Forest (partially completed)
Support Vector Machine (RBF kernel)
Decision Tree (for benchmarking if needed)
Perform:
Hyperparameter tuning (GridSearchCV or RandomizedSearchCV)
5-fold cross-validation
Model evaluation (Accuracy, F1-score, AUC)
2. Data Handling and Feature Engineering
Work with BRAVO dataset (15,000+ samples, 16 features)
Handle:
Missing data
Class imbalance (SMOTE already applied)
Feature engineering:
Load Stress
Maintenance Overdue
Failure Risk Score
Criticality Index
3. Synthetic Data Generation
Generate additional fault scenarios using Monte Carlo simulation
Based on DGA gas ratios (IEC standards)
Validate synthetic data (e.g., KS test)
4. Optimization and Simulation
Develop maintenance scheduling models:
Baseline approach
Greedy algorithm
Integer Programming (PuLP or similar)
Integrate machine learning predictions into scheduling decisions
5. Analysis and Results
Compare AI-based maintenance with traditional methods
Perform cost-benefit analysis (target improvement: 10–15%)
Conduct sensitivity analysis
Technical Requirements:
Python (mandatory)
scikit-learn, pandas, numpy
Optimization libraries (PuLP or OR-Tools)
Jupyter Notebook
Experience in predictive maintenance or engineering datasets is preferred
Deliverables:
Clean, well-documented Python code
Model outputs and evaluation metrics
Simulation results (graphs and comparisons)
Brief explanation of methodology for thesis use
The project focuses on developing an AI-based predictive maintenance framework for distribution transformers, combining machine learning models with optimization-based maintenance scheduling.
Scope of Work:
1. Machine Learning Models
Implement and refine:
Random Forest (partially completed)
Support Vector Machine (RBF kernel)
Decision Tree (for benchmarking if needed)
Perform:
Hyperparameter tuning (GridSearchCV or RandomizedSearchCV)
5-fold cross-validation
Model evaluation (Accuracy, F1-score, AUC)
2. Data Handling and Feature Engineering
Work with BRAVO dataset (15,000+ samples, 16 features)
Handle:
Missing data
Class imbalance (SMOTE already applied)
Feature engineering:
Load Stress
Maintenance Overdue
Failure Risk Score
Criticality Index
3. Synthetic Data Generation
Generate additional fault scenarios using Monte Carlo simulation
Based on DGA gas ratios (IEC standards)
Validate synthetic data (e.g., KS test)
4. Optimization and Simulation
Develop maintenance scheduling models:
Baseline approach
Greedy algorithm
Integer Programming (PuLP or similar)
Integrate machine learning predictions into scheduling decisions
5. Analysis and Results
Compare AI-based maintenance with traditional methods
Perform cost-benefit analysis (target improvement: 10–15%)
Conduct sensitivity analysis
Technical Requirements:
Python (mandatory)
scikit-learn, pandas, numpy
Optimization libraries (PuLP or OR-Tools)
Jupyter Notebook
Experience in predictive maintenance or engineering datasets is preferred
Deliverables:
Clean, well-documented Python code
Model outputs and evaluation metrics
Simulation results (graphs and comparisons)
Brief explanation of methodology for thesis use