Research On The Detection Of Complex Non-stationary Power Quality Disturbances

In contemporary society,electrical energy has become the most widely used energy source because of its easy to transfer and convert features.An ideal power system should supply users with the electrical energy in the form of the sinusoidal waveform at the fixed frequency and rated voltage level.However,along with the changes of loads composition in modern power systems,there are a large number of non-linear loads and shock loads in the power system,which pollutes the quality of the power supply systems,causing ranging from the insulation aging and life reduction of electric power equipment to leading to the large-scale blackout of power grids which causes significant economic losses.On the other hand,in order to alleviate energy crisis and reduce the environmental pollution,more and more renewable energy generation have been integrated into the power grid.The decentralization and uncertainty of renewable energy could bring about a series of power quality problems such as voltage deviation,fluctuation or harmonics.To effectively evaluate and manage the phenomenon of power quality disturbances,the detection and analysis of power quality are indispensable since effective power quality data can be provided to both electricity department and users of electricity to help them to formulate the scientific and reasonable power quality control scheme.There are three main problems to solve in the current power quality detection field.Firstly,various types of disturbances include not only single disturbances but also complex disturbances in which several disturbances appear simultaneously,increasing the difficulty of detecting power quality accurately.Secondly,power quality disturbances are mostly non-stationary signals which cannot be analyzed by the traditional Fourier transform method due to the lack of the time-frequency localization capability.Thirdly,many disturbances parameters need to be detected.They include the amplitude,frequency,phase of the stationary disturbances and the starting and the ending points of non-stationary disturbances as well as the active power,reactive power,power factor and so on.Unfortunately,there are relatively few detection algorithms which can detect all these parameters accurately.This paper introduces the strong tracking filter(STF)and empirical wavelet transform(EWT)to solve the mentioned three problems.As a result,the classification and parameters calculations of power quality disturbances can be achieved successfully.The main contents and innovative achievements of the paper are as follows:(1)Propose a voltage sag detection algorithm based on the STF method.By introducing the fading factor to the extended Kalman filter,STF can adjust the Kalman gain online to overcome some disadvantages such as the divergence problem induced by mismatched model or improper initial values settings.STF can not only provide the magnitude and phase jump of sag events,but also accurately indicate the starting and ending points of sags.By assigning a large value to the prediction error covariance matrix at the starting and ending points of sags,the response performance of filter is improved effectively.Eliminating the nonidentity of the root mean square method(recommended by IEC61000-4-30),the proposed method is more suitable for some occasions such as sag severity and site index estimation.At the same time,the proposed method has good anti-interference performance since it gives the accurate detection results in the presence of harmonics and interharmonics,frequency deviation and DC offset.(2)A new algorithm of power quality disturbances classification is presented based on STF and extreme learning machine(ELM).Combining the good capability of detecting transient disturbances of the low order STF and the strong anti-harmonic interference ability of the high order STF,the paper proposes a features extraction method based on dual strong tracking filters.Investigating the detection results of different disturbances including the single and complex disturbances,the paper puts forward 6 distinctive features with good distinguishing ability.For the purpose of overcoming the slow convergence and existence of local minimum in conventional neural network,the paper proposes a classification strategy based on the extreme learning machine.Studies on the confusion matrix of extreme learning machine with insufficient hidden nodes show only a few boundary samples need more hidden nodes to identify while most samples can be classified correctly by a small number of hidden nodes.Therefore,a rule-based extreme learning machine is introduced to achieve the classification with high accuracy and good stability with less hidden nodes.The algorithm has low computational complexity and is suitable for real-time application.(3)Propose a disturbance parameters estimation method based on empirical wavelet transform(EWT)and Hilbert transform(HT).EWT can realize adaptive signal decomposition by overcoming the shortcoming of wavelet transform(WT)and S transform(ST)in which the time window width can't change adaptively with the frequency.With less interference between different frequency components,the detection accuracy of complex disturbances is improved.By combining EWT and HT,the time-frequency representation of power quality disturbance signals can be obtained.Then the disturbance parameters can be further estimated accurately.On the other hand,the definitions of electrical parameters proposed by IEEE Standard 1459-2010 is not suitable under the non-stationary conditions since these definitions are based on the Fourier series.The paper proposes the new definitions of electrical parameters based on EWT and HT,which is suitable for both stationary and transient signals.(4)Design a hardware implementation of power quality disturbances signal acquisition platform based on Field Programmable Gate Array(FPGA).Considering the increase of types of power quality disturbances,especially the occurrence of complex disturbances more frequently than expected,the paper proposes to use FPGA which has parallel operation characteristic and powerful data handling capability as the core of the signal acquisition platform.To take up less computing resources and have higher operating efficiency,the state machine is used to control the data acquiring,storing and transferring.At the same time,the configurability of FPGA allows integrating new power quality detection algorithms to achieve the scalability of system function.