Jie Wan

Day-Ahead Prediction of Wind Speed with Deep Feature Learning

Day-ahead prediction of wind speed is a basic and key problem of large-scale wind power penetration. Many current techniques fail to satisfy practical engineering requirements because of wind speeds strong nonlinear features, influenced by many complex factors, and the general models inability to automatically learn features. It is well recognized that wind speed varies in different patterns. In this paper, we propose a deep feature learning (DFL) approach to wind speed forecasting because of its advantages at both multi-layer feature extraction and unsupervised learning. A deep belief network (DBN) model for regression with an architecture of 144 input and 144 output nodes was constructed using a restricted Boltzmann machine (RBM). Day-ahead prediction experiments were then carried out. By comparing the experimental results, it was found that the prediction errors with respect to both size and stability of a DBN model with only three hidden layers were less than those of the other three typical approac...

Ultra-short-term wind speed prediction based on multi-scale predictability analysis

Accurate prediction of ultra-short-term wind speed is important in many applications. Because of the different patterns of wind speed, researchers have indicated that a multi-scale prediction method based on wavelet algorithms could improve forecasting. However, traditional multi-scale methods that directly synthetize results of different-scale components may reduce forecasting accuracy because of error accumulation. In this paper, a multi-scale synthesis strategy that takes into account the predictability of different scales is proposed. As the correlation length of each frequency sub-series is different, prediction models are constructed for each frequency sub-series using different numbers of prediction steps. Finally, based on two wind farms in different regions of China and different basic forecast models, eight experiments are performed using real-world wind speed data. Experimental results show that the proposed multi-scale synthesis method has better performance than traditional multi-scale forecasting methods that use direct synthetizing strategies, indicating its utility in wind engineering applications.

A universal power-law model for wind speed uncertainty

The uncertainty is a significant characteristic of wind speed in wind engineering field. Especially, it has brought much more problems to the grid in safe and efficient utilization of large scale wind power. And there is urgent need of systematic and perfect models that can describe windspeed uncertainty in grid scheduling and controlling. In this paper, a universal power-law model is proposed for properly depicting the uncertainty of both wind speed and wind power. According to the turbulence nature of wind uncertainty, the uncertainty model of wind speed is firstly obtained by using wavelet multi-scale transform algorithm for its tight supporting characteristic, which is more reasonable than the traditional algorithm of getting the mean valve and the variance valve of the time series. And the turbulent intensity model is further improved by a power-law model, which is suitable for much more kinds of turbulence on complex geographical conditions than that proposed in current international IEC standard with the sufficient actual data. In physically speaking, the model improvement with three parameters is consistent with turbulence development mechanism. Moreover, the uncertainty modeling method of wind power is developed based on the universal power-law model, which is not only suitable for the power of single wind turbine, but also suitable for the power of whole wind farm. It’s very importance that the wind speed uncertainty model is extended to model the power uncertainty of wind turbine and farm, in especial its or their power output is usually limited for human adjustment control. It has a certain significance to the real-time dispatch and optimal control of the renewable energy power system.

Experimental Study on Large Turbine HP-Valve’s Excitation Fault Caused by Steam’s Unsteady Flow and its Economic Solution

In order to stabilize the uncertainty of large-scale new energy power’s random fluctuation in the grid, there is an increasing number of large thermal power generating unit needing to do deep variable load operation. However, the pattern of steam inlet on turbine’s part load has a very significant impact on the unit operation condition of safety, stability and economy. In this paper the HP-valve’s body vibration fault of large steam turbines caused by unsteady steam flow under partial arc admission operating at part of their full load is researched, and an economic solution based on analysis and diagnosis of fault mechanism is provided by designing of complex HP-valve opening sequence rules. This solution solves the safety problem of valve vibration and avoids the economic loss by using full arc admission or replacing the valve body equipment directly, which is of great effective and practical verified by units’s actual operating experiment. As a result, it demonstrates that the optimization of HP-valve iadmission mode can not only change the stress state of high pressure rotor and prevent its vibration caused by the force of unbalanced flow to improve the shafting stability of the unit when running with variable load, but also can improve the steam instability and solve the resulting vibration problem in HP-valve. And it is of great engineering practical value to improve high-power thermal power unit depth secure efficient load operation.

Performance monitoring of steam turbine regenerative system based on extreme learning machine

The performance conditions of steam turbine regenerative system have important influence on the safety and economy of the units. It is of great significance to doing the research on the performance monitoring of the regenerative system to ensure the safe and economical operation of the whole coal-fired power plants. In view of the shortcomings of the complexity of traditional performance monitoring methods, this paper presents a method of performance degradation of steam turbine regenerative system based on Extreme Learning Machine (ELM). The training set is constructed by using the actual running data of steam turbine regenerating system in previous year, besides the input variables and output variables are selected in combination with its operating mechanism analysis to train the ELM model. Then, the model was used to forecast respectively two test set that include previous years data sample except training set and next years data sample. The residual line is derived from the discrepancy between the actual values and predicted values of test sets. By comparing the residual line between the previous years test set and next years test set, we can see that the heat transfer performance of regenerators has deteriorated. Compared with the actual performance of regenerators, it is verified that the model can effectively monitor the performance of regenerators. In addition, some comparison experiments between ELM, Artificial Neural Networks (ANN) and Elman Regression Neural Networks (ERNN) are taken to compare the performance of each algorithms above mentioned. Through these comparative tests, it is shown that different methods have a certain effect on the performance degradation monitoring of regenerators, but the ELM algorithm is better sensitive and much faster execution than ANN and ERNN.

Frequent pattern extraction based on data and prior knowledge fusion in gas turbine combustion system

Gas turbine combustion system (GTCS) works in the highly adverse environmental conditions of high temperature and high pressure. Because GTCS initial temperature is too high to be directly measured, exhaust gas temperature (EGT) is an alternative to detect performance of GTCS. However, various interferences, such as working and ambient conditions as well as compressor and combustion efficiencies and so on, have a comprehensive effect on EGT, which causes low detection accuracy to GTCS. So, its necessary to build a model that can eliminate the interferences on EGT so as to monitor performance of GTCS effectively. Feature extraction and model structure are keys to modeling. In general, its based on data-driven or mechanism analysis approaches to build the model. However, the former containing a low information density leads to too many input variables and the overcomplicated model structure. In addition, although the latter has a high information density, modeling of a gas turbine is a difficult task in practice, which still limits the application of model-based approaches. For deficiencies of the existing methods, this paper proposes a method based on data and prior knowledge fusion to build a model called Frequent Pattern model for detecting performance of GTCS. Firstly, its based on prior knowledge to tentatively select the model feature. And its found that Average Exhaust Gas Temperature (AET) contains all the influences of various factors on EGT. Secondly, its determined by data analysis that there is a linear relationship between AET and EGT. Therefore, we conclude that a simple linear model can eliminate the interferences on EGT and characterize the structural differences of different flame tubes, by which performance of the GTCS can be monitored accurately and effectively. Thirdly, a contrast test is done to verify the conclusion. A linear model is built based on the proposed method and a nonlinear model is built only based on data-driven method. By comparison, its concluded that the proposed method can not only simplify the model but also improve the model accuracy.

Short-term solar flare prediction using multi-model integration method

A multi-model integration method is proposed to develop a multi-source and heterogeneous model for short-term solar flare prediction. Different prediction models are constructed on the basis of extracted predictors from a pool of observation databases. The outputs of the base models are normalized first because these established models extract predictors from many data resources using different prediction methods. Then weighted integration of the base models is used to develop a multi-model integrated model (MIM). The weight set that single models assign is optimized by a genetic algorithm. Seven base models and data from Solar and Heliospheric Observatory /Michelson Doppler Imager longitudinal magnetograms are used to construct the MIM, and then its performance is evaluated by cross validation. Experimental results showed that the MIM outperforms any individual model in nearly every data group, and the richer the diversity of the base models, the better the performance of the MIM. Thus, integrating more diversified models, such as an expert system, a statistical model and a physical model, will greatly improve the performance of the MIM.

Multi-scale DBNs regression model and its application in wind speed forecasting

Recently, Deep belief networks(DBNs) have been applied in classification and regression, proved to be superior to general algorithms. But its powerful deep feature extraction ability has not yet been fully played so that a novel algorithm, multi-scale DBNs fusing wavelet transform(WT), is proposed in this paper. Based on the advantages of predicting high frequency components from WT by DBN confirmed, the method realizes the effective combination between WT and DBNs. Through the application of wind speed prediction which is typical time series, the results indicate the new algorithm can enhance the forecasting accuracy of DBNs further. At last, the analyzation of optimization for the multi-scale DBNs algorithm is also been done.

Multi-Scale Amplitude Modulation Effect of Wind Speed Random Fluctuation

Wind power is widely used as a type of clean and renewable energy source in recent years. However, large-scale wind power penetration brings lots of challenges due to the uncertainty of wind speed. As a result, the research of wind speed uncertainty plays an important role in large scale wind power integration. In this paper, the multi-scale amplitude modulation effect of wind speed random fluctuation is found based on actual wind speed data. And the physical mechanism of multi-scale characteristics are presented from the perspective of turbulence. This research has both certain physical academic meaning and engineering application value in wind speed uncertainty research.