在能源低碳化的背景下,利用风光可再生能源制氢是大势所趋,然而新能源的波动特性导致电网出现弃风、弃光问题。为解决这一问题,更好地实现新能源功率预测,实现宽功率波动下的高效制氢,基于长短期记忆(Long Short-Term Memory,LSTM)神经网络,提出了一种面向新能源制氢的短期功率预测方法。首先通过数值天气预报中风速、风向、气温、相对湿度和气压等各类气象因素数据,测算出新能量功率和气象因素之间的互相关系数,以便对多维天气数据作降维分析。然后基于新能源出力波动情况及趋势,利用90%置信区间刻画新能源出力的波动性,对预测训练数据起到初步筛选作用,保证预测精度。最后采用LSTM神经网络预测模型,对筛选后的数值结果加以训练并建立天气参数和新能源发电功率间的映射关系。以某光伏电站的实测发电功率数据为例进行验证分析,新能源制氢的短期功率预测方法具有较高的预测精度。将预测结果应用于质子交换膜(Proton Exchange Membrane,PEM)电解槽制氢,可为新能源制氢系统的优化控制提供依据。

In the context of low-carbon energy, producing hydrogen from wind and solar renewable energy has become an inevitable trend. However, the fluctuation nature of new energy often leads to the curtailment of wind and solar power in the grid. To address this issue, improve the accuracy of new energy power forecasting, and enable efficient hydrogen production under wide power fluctuations, this study proposes a short-term power prediction method for new-energy-based hydrogen production using a Long Short-Term Memory (LSTM) neural network. First, based on the numerical weather prediction data, including wind speed, wind direction, air temperature, relative humidity and air pressure, the correlation coefficient between the new energy power output and meteorological factors are calculated to support the dimensionality reduction of multi-dimensional weather data. Then, considering the fluctuation and trend of new energy output, the 90% confidence interval is used to characterize the fluctuation of new energy output, serving as a preliminary screening processs for the training data to ensure the prediction accuracy. Finally, the LSTM neural network prediction model is used to train the selected numerical data and establish a mapping relationship between weather parameters and new energy power generation output. Using measured photovoltaic power output data as a case study, the proposed method demonstrates high prediction accuracy. When applied to hydrogen production via proton exchange membrane (PEM) electrolyzers, the results can provide a reference for the optimal control and operation of new-energy-based hydrogen production system.