A Novel Electrical Load Forecasting Model Using a Deep Learning Approach

The estimate of electricity appeal in modernistic years is becoming progressively relevant thanks to market-free trade and, thus, the initiation of sustainable assets. To satisfy the demands, leading intelligent models are built to form sure explicit power forecasts for multi-time prospects. The load forecasting of electric Power is a crucial process in devising the electric industry and operating electric power systems. Short-term forecasts are adopted to program the power generation and transmission of electricity. Medium-term forecasts are meant to line up the fuel purchases. This necessitates the implementation of the productive determination of algorithms could be a fundamental feature of smart grids and an efficient tool for determining ambiguity for better cost and energy ability decisions like slate the origination, authenticity, power escalation of the system, and monetary smart grid activities. This work introduces a model for the evaluation of the utilization of electricity, which can accurately forecast 68subsequently estimated from minimum to maximum duration with significant improvement in the accuracy of forecasting through advanced deep learning techniques. The analyzes or findings also can provide interesting results for energy consumption with parameters like forecasting efficiency and error with duration of data monitoring algorithms namely (LSTM)-long short-term memory (RNN) - recurrent neural networks and multi-layer perceptron algorithms (MLP). These algorithms furnish the most interesting results with respective to the duration of data. Mainly, MLP and RNN proved to produce favorable results for 24-hour data. Similarly, LSTM has proved better for 15-day data and monthly data with consistency in terms of errors, squared, and mean square. To anticipate data ranging from day to month, the minimal Forecasting error was attained by adopting MLP with R2(0.91). On hour-based data, R2of LSTM holds effective for half-monthly and monthly data with (0.88 and 0.93), RMSE (89.54 and 84.98), MAPE (3.51 and 2.47). RNN has been proven to attain the moderate outputs comparatively. MLP for half-monthly and monthly in terms of R2(0.81 and 0.92), RMSE (90.72 and 85.78) and MAPE (4.25 and 4.01). The result of LSTM acknowledges the enhanced attainment and substantial achievements of electrical load forecasting. © 2024 by Apple Academic Press, Inc.