Q1: In the proposed methodology, it is used chi-square as the distance estimate. Is it any advantage of this?
Q3: Can you please explain what exactly the developed model architecture which is used to do the wind generation forecast? Is it a framework combined with LSTM and Random tree?
Artificial intelligence in power system a survey of Japanese industry Enhanced Wind Generation Forecast Using Robust Ensemble Learning Emin Akhmedli CONTENTS Introduction Preliminaries Proposed Methodology Numerical Studies Conclusion INTRODUCTION Demand estimates refer to the future needs of a product or service. It is necessary to follow a process to obtain a crystalline graph of demand to identify the customer’s need to maintain their position in the market. Since the last era, the steel industry in Bangladesh has been the fastest growing industry in the local market. The industry has been able to produce large quantities of steel to suit local and international markets, but producing large quantities of steel without proper estimation can cause various problems. Demand forecasting is used to support a number of basic business estimates, including turnover, total revenue, revenues, capital utilization, opportunity estimation and control plans, scope scaling, transportation and distribution plans, and more. Any kind of misconduct evaluation can lead to decay or shortage of raw materials. This can lead to high production or even low production. All of these cases destroy the entire supply chain and total revenue, resulting in opportunity cost. Again, the overall industry setup depends on this demand such as the quantity of raw materials, labor and location. There are two categories of forecasting methods statistical methods- Statistical prediction refers to the use of data based on historical data to predict what will happen in the future. This can be done on any quantitative data: stock market results, sales, GDP, home sales, etc. In this example, I will focus on the avalanche data set here. artificial intelligence-based methods- The AI-based assessment solution uses a set of machine learning algorithms to optimize estimates. The system selects the model that best suits the specific business metric you are evaluating. PRELIMINARIES Wind speed measurement is usually done using a wind cup anemometer. Air The cup anemometer consists of a vertical axis and three windshields that capture air. Device Generates voltage related to wind speed or electronically records the number of revolutions per unit Time. Typically, anemometers are attached to wind vans to determine wind direction . Other types of anemometers include ultrasonic anemometers, which are a. Identify the changes in the phase of Sound, and laser anemometers use light that is coherently scattered from the air. Hot-wire anemometer Find the wind speed based on the small temperature difference between the stars placed between them Wind and wind downward (shadow). Non-mechanical purpose anemometers are less sensitive to icing. However, cup anemometers are commonly used. Special designs can be used anywhere in the Arctic and with electrically heated windows and roofs. Spiral anemometers are also commonly used in practice . Air measurement method Mounting the anemometer on the speed mast at the possible point of the wind turbine. PROPOSED METHODOLOGY Some method in wind generation Long-term wind variation information is required to obtain this. Improving wind power forecasting. However, it is possible to collect measurement data at locations Expensive, especially in national research A robust ensemble learning approach has been proposed for large-dimensional wind energy estimation. ,To verify the effect, we used an additional four machine learning models based on our proposed methodology., For training and optimization of the k-near neighborhood algorithm, Murkowski metric analysis was used., The chi-square distance was used to estimate the matching k-nnc goodness between the theoretically expected and observed values. The results show good accuracy for daily, monthly, seasonal and annual wind energy forecasts. NUMERIC WEATHER PREDICATION Based on the processing of meteorological information with state-of-the-art meteorology for forecasting Wind speed and power, many physical models have been industrialized. This model Industrialized based on a number of dynamics, including barrier avoidance, local roughness Surface change, terrain effect, acceleration or descent, level of local wind speed within the air Farm, fan power curve and wind farm design. RANDOM FORECAST REGRESSION RF is a collective approach based on multiple single decisions Trees (identical specimens). In the regression task, assessment in a single decision The tree has an average target value of all cases associated with a single leaf node. The final estimate is the average value of all n single decision trees. ENSEMBLE METHOLOGY Utilization is a good alternative to the well-known machine learning algorithms Clothing model. By using different types of predictors and finally Estimating their output values, combining with the accuracy of the classification And the regression can be improved by reducing the required calculation Time. Unlike sophisticated machine learning algorithms, ensemble Methods require less tuning and specialized domain knowledge. Hybrid methods Several hybrid models have been established to estimate wind power [96]. The nature of Mixed model (i) a combination of statistical and physical methods, (ii) small-medium range Approach, (iii) Alternative statistical approaches in combination. The goal of the hybrid approach Its goal is to influence each method and achieve the best estimated performance worldwide. For example, In autoregressive models with exogenous inputs (ARX), more than a few statistical approaches Better results were found between online measurements and weather forecasts. In general, system participants can be compared to a history of actual available estimates Output data to track the wind performance of measuring instruments FORECAST METHOD OF THE YEARs ENSEMBLE SYSTEM OF FORECASTING GENERATION It is develop with the help of Embedded layer of representative markov chain model There are two characteristics Deep – it is a distributive approximation using deep neural network Generative- for taking x to generate z then taking z to sequence of generate y1 this is process of generative. NUMERIC STUDIES This paper studies the daily indications of local wind speed with a time resolution of 15 minutes, which is a multi-stage instruction consisting of 96 steps. We use the WRF model to implement 72h weather simulation covering different opening regions and different horizontal resolutions in Ningxia, China. Simulation suggests that none of the different ensemble members are the best members, which means that it is difficult to predict the exact single-value NWP simulation in real applications; This is an evolution of the pattern uncertainty of numerical simulations. EFFORTS TO REDUCE THE UNCERTAINTIES OF NUMERICAL SIMULATIONS Since uncertainties are not completely eliminated from numerical sample beginnings and / or errors, there are inevitable errors in the estimation processes. Consequently, minimizing errors from model uncertainty is critical to improving reference performance, such as the numerical wind speed indicators studied in this paper. CONCLUSION In the conclude of this ppt in this paper use of two wind fields for statistical analysis and forecasting would be greatly improved Newly created effective forecasting methods and reliability of results. The result of A summary of the study is given below. Original ensemble methods developed to investigate specific implementations, Less errors of power generation estimates for two wind farms compared to single Ways. Accuracy measurement is the best integration system for ensemble methods for Energy of 30 A new, native integrator has been developed for estimates, called "without bulk average is a basic integration system for aggregate methods for accuracy measurement Integrator developed for estimates "Weighted Average as an Integrator“ Ensemble with three methods in the ensemble "(method code INT_AVE). New, original "additional expert correction" minimizes power generation errors Forecast for two wind fields. Deep Neural Network LSTM Best Single Method, MLP The latter is better when using SVR, KNNR and less suitable for both the physical model Wind farms. Hybrid methods that use nMAE and nRMSE have worse accuracy measurements Comparison of aggregate methods for two wind fields Thank you