![]() ![]() The generated model must be recalibrated for other formations, for horizontal and deviated wells, and when logging while drilling imaging is employed. The methodology depends on image quality, and anomalies in these data increase the error. The accuracy of this method is high enough to be considered during petrophysics evaluations, showing a root-mean-square error of 0.44% and Pearson's correlation coefficient of 0.88. The outcome is equivalent to the curve obtained using a semi-log regression of organic carbon measured in core against density log values. The working data set totalizes 960 wireline log measurements, randomly split into 80% for training and 20% for validation. Implemented machine learning is based on ensemble learning techniques, in this case, an ensemble of decision trees known as random forest. This work aims to teach a machine how to recognize patterns between fractal features in borehole images and their content of total organic carbon. The methodology was applied in La Luna Formation, which has been reported as one of the principal source rocks for Colombia and western Venezuela. Specifically, borehole resistivity imaging, its average resistivity, and gamma rays log are employed to train a regression model. This research presents an alternative approach to computing the content of total organic carbon using wireline logs and machine learning techniques. To allow for a deeper insight into the importance of predictor variables, two case studies in different climate regions are considered in the numerical evaluation. By means of a sensitivity analysis, the paper shows how the value of expert variables is affected by the tilt angle of the PV system. In addition, we investigate the optimal selection of predictor variables for PV power estimation and forecasting. To this end, we use common physical models to create so-called expert variables and test their impact on the performance of single-point and probabilistic models. In this study we quantify the value of predictor variables for PV power estimation and forecasting, assess deficiencies in estimation and forecasting models, and introduce a number of pre-processing steps to improve the overall estimation or forecasting performance. Yet, the importance of predictor variables are consistently ignored in such developments and as a result those models fail to acknowledge the value of including physics-based models. ![]() In particular, machine learning techniques received significant attention in the past decade. This is recognized in literature, where a growing amount of studies deal with the development of PV power estimation and forecasting models. Let’s now do some residual analysis in Python! Residual Analysis In Python Fitting a Holt Winters’ Modelįor this short walkthrough, we will fit the exponential smoothing Holt Winters’ model to the famous US airline passenger dataset.Reliable estimates and forecasts of Photovoltaic (PV) power output form a fundamental basis to support its large-scale integration. Therefore, we want to fail to reject the null hypothesis and the p-values to be greater than 5%. In reality, this is quite easy to adjust for by simply adding or subtracting the bias from the forecasts.įor some context, the null hypothesis of the Ljung–Box test assumes that the residuals are not correlated. The mean of the residuals should be zero, otherwise the forecast will be biased.We can use the Ljung–Box statistical test and a correlogram to determine if the residuals are indeed correlated. If they have any form of correlation, then the model has missed some information that’s in the data. Show very little or no autocorrelation or partial autocorrelation.We can use the residuals to analyse how well our model has captured the characteristics of the data. ![]()
0 Comments
Leave a Reply. |