为了提高锂电池剩余使用寿命（RUL）的预测精度，提出一种基于长短时分叉记忆网络（LFM-Net）的预测模型。首先，从锂电池老化数据中提取放电容量的衰减趋势作为模型的输入，并采用随机失活技术进行去噪预处理。其次，引入长短时记忆网络和注意力机制，从预处理后的数据中捕捉特征并赋予不同的权重。最后，通过时间分叉机制完成最终的RUL预测。基于NASA和CALCE锂电池数据集进行实验验证，结果表明，LFM-Net在2个数据集上的平均绝对误差最大为0.013 5，均方根误差最大为0.021 3。与常用算法相比，其预测精度更优，验证了该方法的有效性和泛化性。

To improve the prediction accuracy of the remaining useful life of lithium batteries， a prediction model based on the long-short-term forked memory network （LFM-Net） is proposed. First， the degradation trend of the discharge capacity is extracted from the lithium battery aging data as the input for the model， and random dropout technique is applied for denoising preprocessing. Second， long short-term memory networks and attention mechanism are introduced to capture features from the preprocessed data and assign different weights. Finally， the time-fork mechanism is employed to complete the final remaining useful life prediction. Experiments are conducted on the NASA and CALCE lithium battery datasets， and the results show that the maximum mean absolute error of LFM-Net on the two datasets is 0.013 5， and the maximum root mean square error is 0.021 3. Compared to commonly used algorithms， it demonstrates superior prediction accuracy， validating the effectiveness and generalization ability of the proposed method.