基于时间序列分析的油中产气数据趋势测度

Prediction of the Trends in the Data of Dissolved Gas Production in Transformer Oil Based on Time Series Analysis

基于数理统计学方法研究变压器绝缘油中溶解气体监测数据时间序列的变化规律,并构建基于差分自回归移动平均法的产气变化趋势测度模型。模型首先将经置信检验的监测数据序列视为一个随机序列,并对历史监测数据进行平稳化预处理,以滤除异常干扰测值;随后进行模型识别与解析,并引入最小信号准则评价、定阶,实现绝缘油产气变化测值预测。通过实际案例验证,模型可由时间序列中的历史值及当前值量化趋势测度值,达到绝缘油中产气趋势分析所需的置信度。

The variation of monitoring data＇s time series of dissolved gas in transformer insulation oil is studied based on mathematical statistics. And the variation trend measurement model for gas production based on autoregressive integrated moving average（ARIMA） is built. Firstly, the model regards the monitoring data sequence which is through confidence test as a random sequence, and conducts smooth preprocessing to historical monitoring data to filter abnormal interference measured value. Then, model identification and analysis is conducted, and the Akaike information criterion（AIC） is introduced for model evaluation and model order to achieve prediction of measurement value of changes in gas production in insulation oil. It is validated through actual case that the model can be used to quantify the trend measure value by the historical values and current values in the time series, and achieve the confidence coefficient that is required for trend analysis of gas production in insulation oil.