Study on reverse deduction method of reservoir inflow based on water level low-pass filtering

Abstract

[Objective] Accurate reservoir inflow data is essential for effective reservoir operation scheduling and hydrological forecasting. However, the inflow process directly derived using the water balance method typically exhibits sawtooth oscillations and contains numerous anomalous negative values. Existing general method still require further improvement in both smoothing effectiveness and applicability. [Methods] Starting with the direct cause of sawtooth oscillations in the reservoir inflow, a low-pass filtering method was used to remove the high-frequency noise in the water level data that affected the calculation of reservoir storage capacity, thereby minimizing its impact on the estimation of inflow. [Results] The result demonstrated that:(1) for the entire flow process, the low-pass filtering method for water level data showed the best performance, with only 3.20% of the calculated inflow data having negative values, an average negative value of-29.76 m³/s, and a smoothing improvement index of 0.77. The moving average method ranked second, and the five-point cubic method performed the worst.(2) For typical flood processes of different magnitudes, the low-pass filtering method reduced the impact of smoothing on the peak flow and the time of peak occurrence, while still achieving a good smoothing improvement. [Conclusion] The result indicate that the reverse deduction method of reservoir inflow based on water level low-pass filtering is characterized by its simplicity in computation and principle. While maintaining a good smoothing improvement, it significantly reduces the anomalous negative values in the inflow process. Additionally, it has a relatively small impact on the peak flow and the time of peak occurrence, which can effectively improve the rationality of the inflow deduction result.

Keywords

reservoir inflow / data smoothing / low-pass filtering / reservoir water level / influencing factors