1 Introduction to Computing Intelligence
Computational intelligence is characterized by uncertainty, nonlinearity and time irreversibility, and takes complex problems as the calculation object. It is the product of the development of modern mathematics and computer, and mainly deals with problems that cannot be solved by conventional methods. Traditional artificial intelligence is based on symbol processing mechanism, and its performance in knowledge expression, information processing and solving combinatorial problems is not ideal. Therefore, research new solutions to improve the flexibility and accuracy of artificial intelligence [1]. Fuzzy system has a good application in description and empirical learning; Neural network can learn experience and skills from network data; Evolutionary computing can put forward the best solution to complex problems, with high stability and optimization. Fuzzy system is superior to neural network and evolutionary computation in reasoning ability, while evolutionary computation and neural network are superior to fuzzy system in learning and searching function. Evolutionary computing is superior to neural network in search scope and adaptability, but neural network is superior to evolutionary computing in optimization and learning ability [2]. Computational intelligence includes three parts: fuzzy system, neural network and evolutionary computation. Although the three technologies are different, the collision caused by the combination of them brings new opportunities.
Application of Computational Intelligence in Water Conservancy and Hydropower Engineering
2. 1 Long-term runoff forecast
With the development of modern production and the demand of economic activities, all countries have strengthened the research and detection of weather. Since 1970s, China's meteorological research has developed from short-term numerical weather forecast to medium-term forecast. However, due to the duality of "uncertainty" and "determinacy" of atmospheric movement, its dynamic research methods and statistical methods need to predict atmospheric activities on the one hand, and the calculation method combining statistics and dynamics can be said to be an ideal research method [3]. However, due to the difference of prediction values, the prediction result set is scraped out, and the result set contains real results. Choosing the correct result in the prediction result set needs to smooth the mean value of the set and eliminate the random error among the set members, so as to analyze the real result. In the range of possible errors, scientific decisions are made according to the information provided by the computational intelligence system to highlight the statistical characteristics of computational intelligence. Long-term hydrological forecasting is a relatively new research field in the development of meteorology. With the analysis of many domestic studies, some progress has been made, but the physical mechanism of long-term hydrological process has not been clarified. Combining the regional and non-adiabatic characteristics of long-term hydrology, there are three main problems in long-term hydrological forecasting in China:
(1) The hydrological department used to imitate the forecasting methods of the meteorological department for a long time, but the long-term forecasting of the meteorological department adopted a large time scale and spatial scale, so it could not play an effective role in the long-term forecasting of the hydrological department [4].
(2) Hydrological departments often use statistical methods in long-term forecasting. It is mainly because the hydrological department has little research on synoptic methods, energetics methods and dynamics.
(3) The accuracy of simple statistical methods is low, so at present, the hydrological department collects and supplies a long-term hydrological forecasting system based on physical analysis, combined with meteorological factors with physical significance, using a large number of data and statistical methods. Computational intelligence is a new method for long-term hydrological forecasting. Its neural network is an intelligent bionic model based on connection theory and a nonlinear dynamic system composed of a large number of neurons. It has a high degree of organization, self-handling, adaptability and application. It has some characteristics of biological neural network and can learn by itself. Therefore, it can be applied to various water conservancy and hydropower projects. Relevant literature points out that neural network can provide a new research direction for the study of hydrological and water resources problems, and the linear least square simplex method of three-layer BP network model structure and parameters can meet the needs of hydrological forecasting of water conservancy and hydropower projects. Lanzhou Hydropower Station has used neural network to forecast monthly runoff, and the application results show that neural network has good application effect in hydrological forecasting, and its benefit is higher than that of multiple regression calculation method [5].
2.2 draft tube pressure fluctuation analysis
In the detection and fault diagnosis of mechanical equipment, the vibration signal is often used as reference data to change the frequency and frequency band of the signal, and then these changed values are input into the diagnosis system to get the operation of the equipment. Fast Fourier Transform (FFT) is a widely used signal feature analysis method in modern water conservancy and hydropower engineering. Its main problem is that it can only make an accurate judgment on stationary signals, but it can't effectively analyze its changing law in signals with large fluctuations. Wavelet change can be reflected as the change of mother wavelet at any signal frequency, so as to get the basis function of wavelet change and the corresponding information. Neural network has large-scale parallel, distributed storage and processing, organization, adaptability and self-learning ability, and can deal with various factors and conditions, uncertain and fuzzy information problems at the same time. The time-frequency localization analysis ability of wavelet packet and the spectrum refinement advantage of maximum entropy spectrum estimation are applied to the dynamic characteristic information of hydraulic turbine shafting.
For more information about project/service/procurement bidding, and to improve the winning rate, please click on the bottom of official website Customer Service for free consultation:/#/? source=bdzd