Energy heat load forecasting plays a crucial role in the low-energy management of buildings. With the growing demand for energy and increasing environmental pressures, accurately predicting building heat loads can provide reliable data support for energy management. This enables the optimization of energy dispatch plans, improves energy utilization efficiency, and helps achieve the goals of energy conservation and emission reduction. However, traditional forecasting methods often struggle with low accuracy when dealing with complex external factors and are susceptible to inappropriate hyperparameter selection. To address these challenges, this study proposes an innovative energy heat load forecasting algorithm that enhances the Long Short-Term Memory (LSTM) model using an improved Artificial Rabbit Optimization (ARO) technique to boost both prediction accuracy and efficiency. First, the Seasonal and Trend decomposition using Loess (STL) algorithm is employed to decompose energy heat load data into trend, seasonal, and residual components, reducing the impact of data fluctuations on model prediction. Next, the ARO algorithm is improved with a Cauchy Mutation and Adaptive Crossover Strategy (CMACS) to optimize the hyperparameters of the LSTM model. To validate the effectiveness of the proposed model, experiments were conducted using real-world data from Byron Apartments at Nottingham Trent University, UK. Due to the unique living patterns of student apartments, energy consumption in these buildings exhibits significant fluctuations and complexity, making the data highly representative. Experimental results show that the proposed CMACS-ARO-LSTM-Attention (CALA)-STL method achieves a coefficient of determination of 98.30%, significantly outperforming traditional methods. This method provides an efficient and reliable solution for energy heat load forecasting, offering robust data support for the optimized management of building energy systems. It enables precise energy management, thereby reducing energy waste and operational costs.

中文翻译：

---

一种用于优化建筑能源热负荷预测的新型 CALA-STL 算法


能源热负荷预测在建筑物的低能耗管理中起着至关重要的作用。随着能源需求的增长和环境压力的增加，准确预测建筑热负荷可以为能源管理提供可靠的数据支持。这可以优化能源调度计划，提高能源利用效率，有助于实现节能减排的目标。然而，传统的预测方法在处理复杂的外部因素时，往往准确性低，并且容易受到不适当的超参数选择的影响。为了应对这些挑战，本研究提出了一种创新的能量热负荷预测算法，该算法使用改进的人工兔子优化 （ARO） 技术增强了长短期记忆 （LSTM） 模型，以提高预测的准确性和效率。首先，采用黄土季节和趋势分解 （STL） 算法将能源热负荷数据分解为趋势、季节和残差分量，减少数据波动对模型预测的影响;接下来，通过柯西突变和自适应交叉策略 （CMACS） 改进 ARO 算法，以优化 LSTM 模型的超参数。为了验证所提出的模型的有效性，使用来自英国诺丁汉特伦特大学 Byron Apartments 的真实数据进行了实验。由于学生公寓独特的生活模式，这些建筑的能源消耗表现出显著的波动性和复杂性，使数据具有很强的代表性。实验结果表明，所提出的 CMACS-ARO-LSTM-Attention （CALA）-STL 方法实现了 98.30% 的决定系数，显著优于传统方法。 该方法为能源热负荷预测提供了一种高效可靠的解决方案，为建筑能源系统的优化管理提供了强大的数据支持。它可实现精确的能源管理，从而减少能源浪费和运营成本。