Table 7 Prediction accuracy comparison among different methods.

From: Deep recurrent neural networks for water hammer transient prediction and dynamic protection optimization in long distance pipelines

Prediction method

RMSE

MAPE

R²

Computation time

Proposed method (Deep BiLSTM + Attention)

Lowest

Lowest

Highest

Fast

Method of characteristics

Moderate

Moderate-High

Moderate

Slow

Support vector machine

High

High

Low-Moderate

Moderate

BP neural network

Moderate-High

Moderate-High

Moderate

Fast

Single-layer LSTM

Moderate

Moderate

Moderate-High

Fast