Abstract

Complete characteristic curves of a pump turbine are fundamental for improving the modeling accuracy of the pump turbine in a pump turbine governing system. In view of the difficulty in modeling the “S” characteristic region of the complete characteristic curves in the pump turbine, a novel Autoencoder and partial least squares regression based extreme learning machine model (AE-PLS-ELM) was proposed to describe the pump turbine characteristics. First, a mathematical model was formulated to describe the flow and moment characteristic curves. The improved Suter transformation was employed to transfer the original curves into WH and WM curves. Second, the ELM-Autoencoder technique and the partial least squares regression (PLSR) method were introduced to the architecture of the original ELM network. The ELM-Autoencoder technique was employed to obtain the initial weights of the Autoencoder based extreme learning machine (AE-ELM) model. The PLS method was exploited to avoid the multicollinearity problem of the Moore-Penrose generalized inverse. Lastly, the effectiveness of the proposed AE-PLS-ELM model has been verified using real data from a pumped storage unit in China. The results demonstrated that the AE-PLS-ELM model can obtain better modeling accuracy and generalization performance than the traditional models and, thus, can be exploited as an effective and sufficient approach for the modeling of pump turbine characteristics.

Highlights

  • As the demand for electricity and the requirements for developing a low-carbon economy continues, the driving force for energy development will gradually shift to renewable and clean energy such as photovoltaic power and wind power [1,2,3]

  • As a crucial part of pumped storage units (PSUs), an accurate pump turbine model is the key to the accurate modeling and simulation of pump turbine governing system (PTGS) [10]

  • The crossing, aggregating phenomena, and multi-value problems in the “S” characteristic region of the pump turbine were reduced through the improved Suter transformation

Read more

Summary

Introduction

As the demand for electricity and the requirements for developing a low-carbon economy continues, the driving force for energy development will gradually shift to renewable and clean energy such as photovoltaic power and wind power [1,2,3]. The common methods to estimate the complete curves of pumped turbine mainly include the assistant mesh processing [18], the Suter transformation, its improved versions [19,20], and the 3D surface fitting [14]. According to whether the flow and moment characteristic curves are pre-transformed, the 3D surface fitting technique can be mainly divided into two categories: the first category is based on the original complete curves of the pump turbine. Liu et al [20] first employed a modified Suter transformation to pre-process the complete characteristic curve and proposed an Adaboost-BP neural network ensemble model optimized by particle swarm optimization to describe the WH and WM characteristics of the pump turbine. The flow and moment characteristic curves of pump turbines are transformed into neural network models, which can be used for a real-time simulation.

Nonlinear Modeling of the Pump Turbine
WH and WM characteristic curves
Extreme Learning Machine
The ELM-Autoencoder Technique
Partial Least Squares Regression
The Proposed AE-PLS-ELM Model
Modeling Process of the Pump Turbine Based on AE-PLS-ELM
Numerical Experiments and Analysis
Parameters Setting
Comparative Analysis of the Results
It can be seen
Additional Test Problem
Findings
Conclusions
Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.