Li Ying Wang

Forecasting Groundwater Level Based on Relevance Vector Machine

Relevance Vector Machine (RVM) is a novel kernel method based on sparse Bayesian, which has many advantages such as its kernel functions without the restriction of Mercer’s conditions, and the relevance vectors are automatically determined and have fewer parameters. In this paper, the RVM model is applied to forecasting groundwater level. The experimental results show the final RVM model achieved is sparser, the prediction precision is higher and the prediction values are in better agreement with the real values. It can be concluded that this technique can be seen as a very promising option to solve nonlinear problems such as forecasting groundwater level.

Forecasting Groundwater Level Based on Wavelet Network Model Combined with Genetic Algorithm

This paper proposed an improved wavelet network model (WNM) which combined with genetic algorithm (GA) to forecast groundwater level, GA is used to determine the weights and parameters of WNM, which can avoid encountering the curse of dimensionality and overcome the shortage in the responding speed and learning ability brought about by the traditional models. Compared to WNM, our results show that the GA-WNM predictor can reduce significantly both relative mean errors and root mean squared errors of predicted groundwater level. We demonstrate the feasibility of applying GA-WNM in groundwater level prediction and prove that GA-WNM is applicable and performs well for groundwater data analysis.

Rolling Bearing Fault Diagnosis Based on Wavelet Packet Feature Entropy-MFSVM

On the basis of wavelet packet-characteristic entropy(WP-CE) and multiclass fuzzy support vector machine(MFSVM), the author proposes a new fault diagnosis method of vibrating of hearings，in which three layers wavelet packet decomposition of the acquired vibrating signals of hearings is performed and the wavelet packet-characteristic entropy is extracted，the eigenvector of wavelet packet of the vibrating signals is constructed，and taking this eigenvector as fault sample multiclass fuzzy support vector machine is trained to implement the intelligent fault diagnosis. The simulation result from the proposed method is effective and feasible.

Rolling Bearing Fault Diagnosis Based on Wavelet Packet- Neural Network Characteristic Entropy

On the basis of neural network based on wavelet packet-characteristic entropy(WP-CE) the author proposes a new fault diagnosis method of vibrating of hearings, in which three layers wavelet packet decomposition of the acquired vibrating signals of hearings is performed and the wavelet packet-characteristic entropy is extracted, the eigenvector of wavelet packet of the vibrating signals is constructed，and taking this eigenvector as fault sample the three layers BP neural network is trained to implement the intelligent fault diagnosis. The simulation result from the proposed method is effective and feasible.

Water Quality Evaluation Based on Multiclass Fuzzy Support Vector Machine

It has been a more complex problem for water quality assessment. And its aim is to well and truly evaluate its degree of pollution for bodies of water, which will be easy to provide some principled projects and criterions for water resource’s protection and their integration application. So, a water quality assessment method based on Multiclass Fuzzy Support Vector Machine is put forward. and a two-step cross-validation was used to search for the best combination of parameters to obtain an optimal training model. The test results show that the method proposed in this paper has an excellent performance on correct ratio compared to BP. It indicated that the performance of the proposed model is practically feasible in the application of water quality assessment.

Adaptive Power Function Based on Data Fitting Method

For the hydropower station in the turbine spiral case pressure measurement of the actual traffic volume issue, a data fitting method with an adaptive power function is proposed, a large number of practical examples of the calculations show that the proposed method can be quitely closer to the actual data and its fitting error has been less than other fitting function. So the method is promising in practice.

Application of ANN Trained with GA in Energy Characteristics of Hydraulic Turbine

The learning algorithm of artificial neural network (ANN) trained with genetic algorithm (GA) are introduced, based on the operation data of hydropower station, the network model of energy characteristics is established based on GA-ANN, the relationship curve between head H and output N is gained under some efficiency. The results show that the algorithm is better than BP neural network and avoid the limitations of BP neural network, the results can be used in the optimal operation of hydropower, and it has a practical significance. The results show the new model has a great importance in hydraulic unit study. It could be generalized into other all efficiency prediction, and it offers a new way in water conservancy and at the meantime a new method for the study of ANN and GA.

Pressure Fluctuation Based on Cascade Correlation Algorithm in Draft Tube

The cascade-correlation(CC) is presented as a neural network growing technique which allows one to gradually build network architecture without the need to redefine the number of neurons to be used in a feed forward. In view of the actual situation that the corresponding space curved surface which expresses pressure fluctuation in draft tube is too complex to be analyzed, considering the pressure fluctuation in draft tube, the network model is established based on CC algorithm and it is applied to hydropower station. Comparing with BP neural network, the experimental results show the prediction precision of the final model is higher and the prediction values are in better agreement with the real values.

Application of Cascade-Correlation Algorithm in Vibration Characteristics of Hydro Turbine

The cascade correlation algorithm that is CC algorithms, CC network structure and CC network weights learning algorithm are introduced, based on the operation data of Wanjiazhai hydropower station, considering the pressure fluctuation, the network model of vibration characteristics is established based on CC algorithm, and the applications of CC and BP algorithm in vibration characteristics of turbine are compared. The results show that the CC algorithm is better than BP neural network, the results can be used in the optimal operation of hydropower, and it has a practical significance.

Fault Diagnosis of Power Transformer Based on DDAG-SVM

Support vector machine (SVM) is a novel machine learning method based on statistical learning theory. SVM is powerful for the problem with small sampling, nonlinear and high dimension. A decision directed acyclic graph(DDAG) based on SVM classifier is applied to fault diagnosis of power transformer. We optimize the structure of a decision directed acyclic graph by putting SVM with higher generalization ability at the upper nodes of the decision tree. The test results show that the classifier has an excellent performance on training speed and reliability.