Nuclear Energy Overview Dataset Analytics and Forecasting Project

This repository contains python code for EDA on nuclear energy overview datasets

Project Overview

Nuclear energy is one of the most reliable and cleanest source of energy that plays a big role in energy transition and climate change mitigation. Public perception, policy decisions, and operational constraints may contribute to the dynamics of its utilization over time despite of its capability to have high capacity factor, means that nuclear energy having a high ratio of actual electricity generated to the maximum possible electricity the plant could have produced if it had operated at full power continuously. This project analyzes this data in order to expose trends and other insights in the nuclear energy sector.

Tools Used

Python: Pandas, Matplotlib, Seaborn, Data Wrangler Extension for VS Code, Sci-kit Learn, XGBoost

Dataset Source

Kaggle: Nuclear Energy Datasets

Data Structure (The description's copied from the Kaggle ReadME file)

Column Name	Description	Datatype
Year	The year for the data entry.	Integer
Month	The month for the data entry.	Object
Nuclear Generating Units, Total Operable Units	The total number of operable nuclear generating units (data not always available).	Object
Nuclear Generating Units, Net Summer Capacity	The net summer capacity of nuclear generating units in million kilowatts.	Float
Nuclear Electricity Net Generation	The net generation of electricity from nuclear power in millions of kilowatt-hours.	Integer
Nuclear Share of Electricity Net Generation	The percentage share of total electricity net generation coming from nuclear power.	Float
Nuclear Generating Units, Capacity Factor	The capacity factor (actual output vs potential output) of nuclear generating units.	Float

Result Summary

Data Cleaning

Removed the Nuclear Generating Units, Total Operable Units column because of too many missing data Not Available.

Forecasting Method

1. The method used for the time series forecasting is by using XGBoost model to predict the Nuclear Electricity Net Generation as the target value. The feature added for the forecasting are lag_7, rolling_mean_1, rolling_std_1, rolling_mean_7, rolling_std_7, rolling_mean_30, rolling_std_30, Month, Year, Quarter, DayOfWeek, and DayOfYear.
2. The target value is also scaled using StandardScaler from Sci-kit Learn module before training.
3. The dataframe is splitted using TimeSeriesSplit as the cross validation method.
4. The Hyperparameter tuning uses GridSearchCV method and found the best parameter for the model: {'learning_rate': 0.01, 'max_depth': 2, 'n_estimators': 2000}.

Findings

Data Statistics Summary

index	Nuclear Generating Units, Net Summer Capacity	Nuclear Electricity Net Generation	Nuclear Share of Electricity Net Generation	Nuclear Generating Units, Capacity Factor
count	614	614	614	614
mean	85.6071	49806.5	17.2166	76.4963
std	23.2862	19647.1	4.17584	16.2811
min	14.533	5697	3.9	34.6
25%	78.7078	31481.5	15.525	61.025
50%	98.533	57362	18.8	79.15
75%	99.628	65169.2	20.1	91.875
max	102.206	74649	22.9	101.6

Correlation Heatmap

Net Summer Capacity of Nuclear Energy

This timeseries data explains the maximum output of nuclear generating units during peak summer conditions. This data basically tells us how much reliable generation we can count on during peak demand periods (usually in summer). The line plot shows an overall increase in the net capacity of nuclear power, primarily during the Major Expansion Era (1973–1990) (highlighted in green), when the number of operating plants rose from 109 to 413. In the subsequent period from 1990 to 2003, there were 59 new plants commissioned and 50 plants shut down, indicating a slowdown in growth caused by the Three Mile Island (1979) and the Chernobyl Accident (1986). The most recent five-year data also shows a decline in net capacity, reflected by a total annual growth rate of –3.74% [1].

Nuclear Electricity Net Generation

The data show a clear upward trend in the total electricity generated by nuclear power plants (NPPs). This increase reflects the growing role of nuclear energy in meeting global electricity demand, supported by improvements in reactor performance and operational efficiency.

Nuclear Share of Electricity and Nuclear Capacity Factor

The Nuclear Share of Electricity and the Nuclear Capacity Factor both show how much nuclear power contributes to overall electricity generation, and they often move in the same direction. The Nuclear Share of Electricity tells us what percentage of total electricity comes from nuclear energy relative to all energy source, while the Capacity Factor shows how much of a plant’s full potential is being used.

Seasonality Impact on Nuclear Electricity Net Generation and Capacity Factor

This plot shows how the total electricity generated by nuclear energy and the capacity factor vary by month. The data indicate that nuclear electricity generation and capacity factor peak during the summer and winter seasons, when electricity demand is high due to the use of air conditioners in summer and heaters in winter. Meanwhile, they reach their lowest levels in spring and autumn, when electricity demand is relatively lower, making these seasons the ideal time for scheduled maintenance outages at nuclear power plants.

Net Electricity Generation Forecasting

The XGBoost model demonstrates good performance, achieving mean squared error (MSE) values of 108.19 on the training data and 108.80 on the test data, indicating consistent predictive accuracy and minimal overfitting.

Feature Importance

The figure below highlights rolling_mean_1 and rolling_mean_7 as the most important features, suggesting that recent fluctuations and short-term averages play a key role in shaping the model’s predictive performance.

References

[1] OECD Nuclear Energy Agency (NEA) & International Atomic Energy Agency (IAEA). Forty Years of Uranium Resources, Production and Demand in Perspective: The Red Book Retrospective. OECD Publishing, 2006.