Short Term Electric Load Forecasting using ANN in MATLAB
6.5 هزار بار بازدید -
2 سال پیش
-
The video discusses short-term load
The video discusses short-term load forecasting using artificial neural networks (ANN) in MATLAB.
The importance of load forecasting in maintaining a balance between energy production and consumption is highlighted.
Various factors such as weather conditions (rainfall, wind speed, air pressure, humidity, temperature) impact energy consumption.
The video mentions the use of the Levenberg-Marquardt training function for ANN and compares the results with other literature.
The presenter discusses their objectives, including studying different research papers, increasing the number of inputs, and testing the model on weekly and daily load data.
Detailed Description:
Electrical load forecasting is a significant issue and problem in our everyday electric power systems operations and management. It is one of the crucial tasks to be solved in order to maintain economic and sufficient power generation and reliable system operating conditions. However, with the growth and development of complex electrical power systems and a more diversified environment for providing energy services, the problem has now faced many new challenges.
Numerous load forecasting techniques and approaches have been proposed in the past, but researchers have found neural networks-based methods more reliable while producing fewer forecasting errors because of their capacity to adjust appropriately to the hidden characteristics of the connection among the output and input load data. Due to the non-linear characteristics between the changing behaviour of the load patterns and effective-parameters on it and complex relation between load design changes and their parameters, researchers have accepted NN methods more than any other conventional techniques.
Artificial Neural Networks based series short-term load forecasting (STLF) method is broadly utilized in power business. STLF can be performed utilizing other procedures such as regression-models, fuzzy-logic, expert-systems, etc. But applications of ANN have overcome the limitations in the areas of forecasting. Electrical load demands primarily depend on several weather conditions like temperature, humidity, rain, air pressure, wind speed etc.
The major aim of my project was to implement and explore the performance of ANN based approaches for daily and weekly short term load prediction (STLF). A back-propagation feedforward NN configuration was selected for training model on our dataset and compare the outcomes with multiple-linear-regression (MLR) technique. ISO New England hourly load dataset of previous four-and a half-year was utilized in training and testing neural network model. The main focus was given on daily 24-hour load predictions. Hourly load data was taken as the target data while other parameters like hour, dew point, dry bulb, working days, holidays and previous hour, day, week load is taken as input data. Simulations were carried out in MATLAB to get the better accomplishments of our designed network. The outcomes displayed that as compared to the regression technique, the performance of our designed setup was best.
The importance of load forecasting in maintaining a balance between energy production and consumption is highlighted.
Various factors such as weather conditions (rainfall, wind speed, air pressure, humidity, temperature) impact energy consumption.
The video mentions the use of the Levenberg-Marquardt training function for ANN and compares the results with other literature.
The presenter discusses their objectives, including studying different research papers, increasing the number of inputs, and testing the model on weekly and daily load data.
Detailed Description:
Electrical load forecasting is a significant issue and problem in our everyday electric power systems operations and management. It is one of the crucial tasks to be solved in order to maintain economic and sufficient power generation and reliable system operating conditions. However, with the growth and development of complex electrical power systems and a more diversified environment for providing energy services, the problem has now faced many new challenges.
Numerous load forecasting techniques and approaches have been proposed in the past, but researchers have found neural networks-based methods more reliable while producing fewer forecasting errors because of their capacity to adjust appropriately to the hidden characteristics of the connection among the output and input load data. Due to the non-linear characteristics between the changing behaviour of the load patterns and effective-parameters on it and complex relation between load design changes and their parameters, researchers have accepted NN methods more than any other conventional techniques.
Artificial Neural Networks based series short-term load forecasting (STLF) method is broadly utilized in power business. STLF can be performed utilizing other procedures such as regression-models, fuzzy-logic, expert-systems, etc. But applications of ANN have overcome the limitations in the areas of forecasting. Electrical load demands primarily depend on several weather conditions like temperature, humidity, rain, air pressure, wind speed etc.
The major aim of my project was to implement and explore the performance of ANN based approaches for daily and weekly short term load prediction (STLF). A back-propagation feedforward NN configuration was selected for training model on our dataset and compare the outcomes with multiple-linear-regression (MLR) technique. ISO New England hourly load dataset of previous four-and a half-year was utilized in training and testing neural network model. The main focus was given on daily 24-hour load predictions. Hourly load data was taken as the target data while other parameters like hour, dew point, dry bulb, working days, holidays and previous hour, day, week load is taken as input data. Simulations were carried out in MATLAB to get the better accomplishments of our designed network. The outcomes displayed that as compared to the regression technique, the performance of our designed setup was best.
2 سال پیش
در تاریخ 1400/11/04 منتشر شده
است.
6,521
بـار بازدید شده