Scholarlink

This study evaluates the performance of four time-series fore-casting models—ARIMA, SARIMA, LSTM, and Neural Prophet—in predicting UK domestic electricity demand. The analysis incorporates time-series data enriched with weather-related features such as temperature and humidity. The models were assessed based on their prediction accuracy and ability to capture trends. Neural Prophet emerged as the most accurate model across all metrics, including a Mean Absolute Error (MAE) score of 0.104, a Mean Absolute Percentage Error (MAPE) score of 30.251, a Mean Squared Error (MSE) score of 0.037, a Root Mean Squared Error (RMSE) score of 0.193. It demonstrated a high level of accuracy and a solid capability to capture trends in the dataset, although it struggled with predicting the most significant outliers. LSTM, while ranking second in performance metrics, had notable difficulties with outlier prediction. This suggests that it may not handle irregular data points as effectively as Neural Prophet and could benefit from further refinement to improve its trend-capturing capabilities. In comparison, traditional models such as ARIMA and SARIMA displayed lower accuracy. ARIMA could not effectively account for seasonal trends, and SARIMA, despite its improvements over ARIMA, still lagged behind LSTM and Neural Prophet. The findings indicate that Neural Prophet, with its excellent trend-capturing ability and superior handling of outliers, emerges as the most effective model for improving prediction accuracy in energy billing. These insights offer valuable guidance for UK energy suppliers to develop more accurate billing systems, potentially enhancing current EPC-based methods significantly.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Forecasting Electricity Demand for UK Consumers: A Time Series Machine Learning Approach

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Kamran Ali,
  • Essa Q. Shahra,
  • Shadi Basurra,
  • Moad Idrissi

摘要

This study evaluates the performance of four time-series fore-casting models—ARIMA, SARIMA, LSTM, and Neural Prophet—in predicting UK domestic electricity demand. The analysis incorporates time-series data enriched with weather-related features such as temperature and humidity. The models were assessed based on their prediction accuracy and ability to capture trends. Neural Prophet emerged as the most accurate model across all metrics, including a Mean Absolute Error (MAE) score of 0.104, a Mean Absolute Percentage Error (MAPE) score of 30.251, a Mean Squared Error (MSE) score of 0.037, a Root Mean Squared Error (RMSE) score of 0.193. It demonstrated a high level of accuracy and a solid capability to capture trends in the dataset, although it struggled with predicting the most significant outliers. LSTM, while ranking second in performance metrics, had notable difficulties with outlier prediction. This suggests that it may not handle irregular data points as effectively as Neural Prophet and could benefit from further refinement to improve its trend-capturing capabilities. In comparison, traditional models such as ARIMA and SARIMA displayed lower accuracy. ARIMA could not effectively account for seasonal trends, and SARIMA, despite its improvements over ARIMA, still lagged behind LSTM and Neural Prophet. The findings indicate that Neural Prophet, with its excellent trend-capturing ability and superior handling of outliers, emerges as the most effective model for improving prediction accuracy in energy billing. These insights offer valuable guidance for UK energy suppliers to develop more accurate billing systems, potentially enhancing current EPC-based methods significantly.