The continuous penetration of renewable energy resources (RES) into the energy mix and the transition of the traditional electric grid towards a more intelligent, flexible and interactive system, has brought electrical load forecasting to the foreground of smart grid planning and operation. Predicting the electric load is a challenging task due to its high volatility and uncertainty, either when it refers to the distribution system or to a single household. In this paper, a novel methodology is introduced which leverages the advantages of the state-of-the-art deep learning algorithms and specifically the Convolution Neural Nets (CNN). The main feature of the proposed methodology is the exploitation of the statistical properties of each time series dataset, so as to optimize the hyper-parameters of the neural network and in addition transform the given dataset into a form that allows maximum exploitation of the CNN algorithm’s advantages. The proposed algorithm is compared with the LSTM (Long Short Term Memory) technique which is the state of the art solution for electric load forecasting. The evaluation of the algorithms was conducted by employing three open-source, publicly available datasets. The experimental results show strong evidence of the effectiveness of the proposed methodology.

The massive penetration of Distributed Energy Resources (DER) and high speed power electronics, even at the distribution grid edge, increases the management and operational complexity of the network while necessitates predictive maintenance, control and failure prediction mechanisms in near real time. A distribution subsystem comprises a CPS (Cyber Physical System) where the Cyber part is enabled by a low latency IIoT network and computational resources at the edge (distribution grid subsystem controller) while the physical system is represented (Digital Twin) by a hybrid model of the grid: a data driven machine learning representation of a limited number of nodes, along with a discrete-time model-based transient state estimator which captures the near real time physical operation of the grid. In case of early warning measurements acquired through the IIoT, machine learning edge computations can detect the severity of the warning, identify the network area of interest while ignite the execution of the transient state estimation. The transient state estimation predicts the incoming failure in near real time while the timing window allows possible preventive actuation.

The extreme ultraviolet (EUV) photolithography process is a cornerstone of semiconductor manufacturing and operates under demanding precision standards realized via nanometer-level overlay (OVL) error modeling. This procedure allows the machine to anticipate and correct OVL errors before impacting the wafer, thereby facilitating near-optimal image exposure while simultaneously minimizing the overall OVL error. Such models are usually high dimensional and exhibit rigorous statistical phenomena such as collinearities that play a crucial role in the process of tuning their parameters. Ordinary least squares (OLS) is the most widely used method for parameters tuning of overlay models, but in most cases it fails to compensate for such phenomena. In this paper, we propose the usage of ridge regression, a widely known machine learning (ML) algorithm especially suitable for datasets that exhibit high multicollinearity. The proposed method was applied in perturbed data from a 300 mm wafer fab, and the results show reduced residuals when ridge regression is applied instead of OLS.

Extreme Ultraviolet (EUV) photolithography process, is a cornerstone of semiconductor manufacturing and operates under demanding precision standards realized via nanometer-level overlay (OVL) error modeling. This procedure allows the machine to anticipate and correct OVL errors before impacting the wafer, thereby facilitating near-optimal image exposure while simultaneously minimizing the overall OVL error. Such models are usually high dimensional and exhibit rigorous statistical phenomena such as collinearities that play a crucial role in the process of tuning their parameters. Ordinary Least Squares (OLS) is the most widely used method for parameters tuning of Overlay models but in most cases it fails to compensate for such phenomena. In this paper we propose the usage of Ridge Regression, a widely known Machine Learning (ML) algorithm especially suitable for datasets that exhibit high multicollinearity. The proposed method has been applied in perturbed data from a 300 mm wafer fab and the results show reduced residuals when ridge regression is applied instead of OLS.

In the photolithography process of integrated circuit (IC) manufacturing, overlay (OL) control is a key factor for successful exposure. Overlay control is achieved by creating models that can estimate the expected overlay error, so that this can be corrected before the wavefront reaches the wafer surface. Such models consist of basis functions, influenced by application-controllable variables such as process settings, tool characteristics, and field-related factors with their corresponding model parameters. The process of tuning the model parameters involves several time-consuming sensor measurements on markers distributed across the wafer surface and significantly impacts the throughput performance of the exposure system. For this reason, a strategic selection of wafer markers is necessary. In this paper, we propose a methodology to improve the overlay modeling process by exploiting Surface Reconstruction (SR). SR is used as an intermediate step, during the parameter estimation process, to generate additional data from a strategically selected set of markers that are spatially uniform and provide maximum information gain. The proposed method reconstructs the wafer surface by incorporating spatially interpolated estimates derived from the physical insights of existing measurements. This augmented data set, comprised of measured and synthetic overlay data, serves as a comprehensive input for the parameters’ tuning process leading to more accurate overlay modeling. The proposed method is evaluated using real-industry data from a semiconductor process of 300mm diameter wafers. The results demonstrate a significant reduction in the overlay residuals in both x and y directions.

In photolithographic processes, nanometer-level-precision wavefront-aberration models enable the machine to be able to meet the overlay (OVL) drift and critical dimension (CD) specifications. Software control algorithms take as input these models and correct any expected wavefront imperfections before reaching the wafer. In such way, a near-optimal image is exposed on the wafer surface. Optimizing the parameters of these models, however, involves several time costly sensor measurements which reduce the throughput performance of the machine in terms of exposed wafers per hour. In that case, photolithography machines come across the trade-off between throughput and quality. Therefore, one of the most common optimal experimental design (OED) problems in photolithography machines (and not only) is how to choose the minimum amount of sensor measurements that will provide the maximum amount of information. Additionally, each sensor measurement corresponds to a point on the wafer surface and therefore we must measure uniformly around the wafer surface as well. In order to solve this problem, we propose a sensor mark selection algorithm which exploits genetic algorithms. The proposed solution first selects a pool of points that qualify as candidates to be selected in order to meet the uniformity constraint. Then, the point that provides the maximum amount of information, quantified by the Fisher-based criteria of G-, D-, and A-optimality, is selected and added to the measurement scheme. This process, however, is considered “greedy”, and for this reason, genetic algorithms (GA) are exploited to further improve the solution. By repeating in parallel the “greedy” part several times, we obtain an initial population that will be the input to our GA. This meta-heuristic approach outperforms the “greedy” approach significantly. The proposed solution is applied in a real life semiconductors industry use case and achieves interesting industry as well as academical results.