Methods for disaggregation and reaggregation of gas consumption Essay

In a different application, disaggregation of natural gas consumption data has attracted attentions in the recent years. A statistical Time Series Disaggregation (TSD) method is proposed for disaggregation and reaggregation of gas consumption data with constant time step where temperature is the independent variable. Parameters of the model are calculated by maximum likelihood estimation [14]. The method is applied to LF consumption time series of 1000 consumers all connected to a single pipeline with known daily gas flow. It is observed that sum of daily values of disaggregated HF consumption time series is different from the total daily gas flow of the feed pipeline of the gas network which violates flow conservation. A regression based TSD method called Time Series Reconstruction (TSR) is developed to extract daily gas consumption from monthly data in order to forecast gas consumption in the coming days [15]. Daily values of the disaggregated time series within each month must sum to the recorded monthly consumption of that month. The TSR method violates this condition.
Forecasting the data behavior requires establishing a relationship between the data and the independent variables that dominate its behavior [16][17]. It is possible to forecast HF values of the LF time series using the TSD method since TSD relates LF and HF dependent time series to the independent HF variables [15]. There are four reasons why a new disaggregation method is required
1) Although gas flows continuously through the gas meter but flow is recorded irregularly (once per day, 15 days, 30 days, 36 days, etc.) which differs from one gas meter to another. These are the related LF time series with different time steps which should be disaggregated to HF (daily) data.
2) NILM uses features (load signatures) of the appliances to disaggregate the total consumption at the meter to the appliancespecific consumption. The only feature we have for the gas consumption disaggregation is daily temperature which highly influences gas consumption and is very different to the appliancespecific characteristics. Therefore, a disaggregation method different from NILM is required.
3) The existing TSD methods are suitable for single time series with constant time step whereas gas consumption data are multiple related time series with variable time steps.
4) The TSD methods presented for the gas data in the recent years violate TSD basic constraints.
Many attempts have been made to solve the problem but a complete, straightforward and comprehensive solution is not proposed for the natural gas industry. For this purpose, we present a TSD method which disaggregates mixed LF and HF data by introducing a universal TSD for many related time series with variable time steps. The rest of the paper is organized as follows. General structure of the problem is sketched in Section II. The proposed TSD method is formulated in Section III. Compatibility of LF and HF data and test cases are studied in Sections IV and V, respectively. The proposed TSD method is applied to a gas network in Section VI. Concluding remarks are drawn in Section VII.