Research On Condition Monitoring And Fault Diagnosis Of Photovoltaic Array

In response to the series of problems caused by the rapid consumption of traditional energy resources and the growing demand for energy,countries around the world,including China,have placed great prominence on promoting the vigorous development of renewable energy.Thus,the healthy development of solar photovoltaic(PV)generation,which is an important source of energy in the future,is of great requirement.However,issues about low operation efficiency,difficult monitoring and diagnosis,and high cost of manual operation and maintenance are still unsolved for the explosive growing PV systems.In light of the above,this dissertation focuses on the PV array,which is the core component of a PV system.A multi-state thermal model and a PV model with time-varying parameters are established to monitor two critical states,the cell temperature and electrical output characteristics.Then,a differentiated diagnosis method for two prevalent typical faults,i.e.,line-to-line(LL)faults and partial shading(PS),is proposed using the traced maximum power point data and electrical time series characteristics.The main content and contributions are as follows:(1)A multi-state thermal model to characterize the inertial variation of module temperature is developed to address the problem that the cell temperature,which determines the electrical output characteristics of PV,cannot be measured directly due to the module encapsulation.In addition,the temperature is affected by a variety of external and internal material factors.However,the existing steady-state models have limited ability to describe the dynamic characteristics of temperature under changing environmental conditions.In this regard,the heat balance equations are established for the three main layers after analyzing the heat transfer processes between both the module and the surrounding environment and the internal layers.Moreover,different temperatures through the module are modeled as multi-state to avoid the assumption of uniform temperature distribution.Furthermore,the comparison of actual measurements and the estimated outputs can help the timely state correction under certain unforeseen circumstances.The average estimation deviation is around 0.48?C under different seasons and various weather conditions,which verifies the robustness and generality of the proposed model for cell temperature monitoring.Thus,a more accurate temperature for PV modeling is guaranteed compared to the existing empirical equations and one-state dynamic model.(2)A model with time-varying parameters is developed to reliably monitor the PV output states to reduce the false alarms caused by the unstable performance of the static parameter model in changing environments,especially under low irradiance conditions.Firstly,a dualiteration algorithm is presented to extract the parameters using only three points under limited conditions.The anti-interference of the method to noisy data is then enhanced by parameter range constraints,which overcomes the shortcomings of existing parametrization methods with poor practicality.Moreover,the nearest neighbor parameter transfer method is investigated to use the parameter extraction results under the nearest test conditions as the baseline values for calculating the outputs of the unknown conditions.By doing this,the effects of environmental and aging factors on the parameters are taken into account.Additionally,an improved parameter calculation method based on distance weighting is designed to further improve the performance of the one-diode model under low irradiance and high temperature conditions.The relative errors of the proposed method are kept below 1% for different module materials.The results demonstrate its evaluating accuracy and performance stability under varying irradiance and temperature,thus being an effective output assessment for fault diagnosis.(3)In view of the failure of traditional overcurrent protection devices due to the PV current characteristics and poor timeliness of generic diagnosis methods for line connection failure,a fast anomaly detection and accurate diagnosis technique is proposed based on maximum power point tracking(MPPT)data.Firstly,the instantaneous current reduction between two MPPT sampling instants is analyzed to achieve low-cost detection.The time-to-detection is reduced to 0.2 s with the disadvantages of limited application in complex scenarios for existing methods being overcome.Moreover,specific operating points are obtained by the event-triggered I–V curve tracing.By doing so,frequent scanning is substituted to reduce unnecessary power losses.An inflection point is utilized to distinguish LL faults from PS incidents with an accuracy of 99.81% to avoid power failure protection due to misdiagnosis.Particular voltages are set according to the diagnosed fault type and evaluated mismatch level to differentiate the currents in normal strings and fault ones as much as possible.Hence,the faulty string can be located quickly and effectively.Multiple scenarios,including low irradiance conditions,low mismatch levels,with and without blocking diodes,are simulated to validate the proposed fault detection and diagnosis technique.The performance is also compared with existing techniques to demonstrate its superiority in timeliness and accuracy to enhance the efficiency,reliability,and safety of PV systems.(4)Considering that the actual operation and maintenance of PV arrays have an insufficient response to static shading and variable shading discrepancies due to the lack of shading classification,a diagnostic method based on electrical time series characteristics is further investigated for these two common types of shading.Firstly,the false alarm caused by uncertainty and external disturbance can be effectively prevented by applying the presented preprocessing in view of the non-timely needs of shadings.Then,the characteristics of the generated shadow location and area variation are revealed,and the distribution of the periods with abnormal power loss is utilized for preliminary classification.Specifically,a statistical threshold was set for the output voltage to distinguish static shading from the other permanent faults.Moreover,the suspected variable shaded period could be marked according to the periodicity owing to the dominant role of diffuse radiation under variable shading,thus the repeating alarms can be effectively avoided.The volatility and periodicity of the time series are utilized to achieve realtime monitoring instead of current–voltage measurements.The effectiveness of the proposed methodology to differentiate PS patterns is fully validated through the conducted experimental tests in order to design distinguished maintenance strategies.