基于数据驱动的分布式低碳能源站状态预测方法

A data-driven method for state prediction of distributed low-carbon energy stations

分布式低碳能源站(distributed low-carbon energy station,DLCES)能提高能源利用效率和可再生能源消纳率,准确预测DLCES的未来运行状态能保障其安全可靠运行。为此,提出一种基于数据驱动的分布式低碳能源站状态预测方法。首先,分析DLCES结构与运行状态,利用关键状态量和偏移量变化将运行状态划分为正常、恢复、临界及紧急状态;其次,构建深度长短期记忆(long-short term memory,LSTM)模型,并利用改进粒子群算法进行超参数优化,提升预测模型性能;最后,利用测试集数据对柯西变异的粒子群算法(Cauchy mutation particle swarm optimization,CMPSO)和LSTM相结合的模型进行预测仿真,将其与RNN、LSTM及BP神经网络预测结果对比分析。结果表明:CMPSO-LSTM模型能提高预测效果,更具实际意义。

Distributed low-carbon energy stations(DLCES)can improve energy utilization efficiency and renewable energy consumption rates.Accurate prediction of the future operating status of DLCES can ensure its safe and reliable operation.Therefore,a data-driven prediction method for the status of DLCES is proposed.Firstly,the structure and operating status of DLCES are analyzed,and the operating status is divided into normal,recovery,critical,and emergency states using key state variables and deviations.Secondly,a deep long-short term memory(LSTM)model is constructed,and an improved particle swarm optimization algorithm is used for hyper-parameter optimization to improve the performance of the prediction model.Finally,the CMPSO-LSTM model is simulated using test sets data,and the results are compared with those of RNN,LSTM,and BP neural networks.The results show that the CMPSO-LSTM model can improve prediction results and has more practical significance.