Spatio-Temporal Feature Fusion-Based Hybrid GAT-CNN-LSTM Model for Enhanced Short-Term Power Load Forecasting

Electrical load forecasting is of vital importance in intelligent power management and has been a hot spot in industrial Internet application field. Due to the complex patterns and dynamics of the data, accurate short-term load forecasting is still a challenging task. Currently, many tasks use deep neural networks for power load forecasting, and most use recurrent neural network as the basic architecture, including Long Short-Term Memory (LSTM), Sequence to Sequence (Seq2Seq), etc. However, the performance of these models is not as good as expected due to the gradient vanishing problem in recurrent neural network. Transformer is a deep learning model initially designed for natural language processing, it calculates input–output representations and captures long dependencies entirely on attention mechanisms which has great performance for capturing the complex dynamic nonlinear sequence dependence on long sequence input. In this work, we proposed a model Time Augmented Transformer (TAT) for short-term electrical load forecasting. A temporal augmented module in TAT is designed to learn the temporal relationships representation between the input history series to adapt to the short-term power load forecasting task. We evaluate our approach on a real-word dataset for electrical load and extensively compared it to the performance of the existed electrical load forecasting model including statistical approach, traditional machine learning and deep learning methods, the experimental results show that the proposed TAT model results in higher precision and accuracy in short-term load forecasting.

Short-term load forecasting is viewed as one promising technology for demand prediction under the most critical inputs for the promising arrangement of power plant units. Thus, it is imperative to present new incentive methods to motivate such power system operations for electricity management. This paper proposes an approach for short-term electric load forecasting using long short-term memory networks and an improved sine cosine algorithm called MetaREC. First, using long short-term memory networks for a special kind of recurrent neural network, the dispatching commands have the characteristics of storing and transmitting both long-term and short-term memories. Next, four important parameters are determined using the sine cosine algorithm base on a logistic chaos operator and multilevel modulation factor to overcome the inaccuracy of long short-term memory networks prediction, in terms of the manual selection of parameter values. Moreover, the performance of the MetaREC method outperforms others with regard to convergence accuracy and convergence speed on a variety of test functions. Finally, our analysis is extended to the scenario of the MetaREC_long short-term memory with back propagation neural network, long short-term memory networks with default parameters, long short-term memory networks with the conventional sine-cosine algorithm, and long short-term memory networks with whale optimization for power load forecasting on a real electric load dataset. Simulation results demonstrate that the multiple forecasts with MetaREC_long short-term memory can effectively incentivize the high accuracy and stability for short-term power load forecasting.

With the large-scale development of the power industry, new requirements are put forward for the stable operation of the power system. Power system load forecasting refers to the use of historical load data to predict the future load value, which is an important part of energy management system. The short-term load forecasting process is often combined with the basic mechanism of power grid dispatching to achieve the balance of power grid supply and demand, reflecting the highly nonlinear computing ability. Power load forecasting is the premise of power grid real-time control, operation planning and development planning. At present, the grey model model used for power system load forecasting generally has the problems of large calculation amount, no mature theoretical basis for selecting structures and parameters, etc. This paper discusses the application of grey model in short-term power load forecasting, and puts forward a principal component analysis method suitable for ordinary daily power load forecasting data, which improves the accuracy of short-term power load forecasting.

Short-term load forecasting (STLF) of power systems is an important portion of the daily dispatch of the power industry. The preciseness of STLF straightforwardly disturbs the reliability, security, and economy of power system function. Thus, the research on STLF techniques is the main focus of researchers at abroad and home. Recently, artificial neural networks (ANN) were broadly studied as an intellectual method and implemented in the domain of short-term power load forecasting. Distinct methods like hybrid, conventional, and Artificial Intelligence (AI) methods were advanced to examine STLF. In this view, this study develops an Automated Short Term Load Prediction in Power Systems using Collision Bodies Optimization with MultiHead Deep Learning (AS TLP-CBMDL) model. The major intention of the AS TLP-CBMDL methodology is to predict the load in power systems which are adaptable to the time varying characteristics. To accomplish this, the ASTLP-CBMDL system model applies multihead attention based long short-term memory (MHALS TM) technique for performing load prediction. In addition, the colliding body's optimization (CBO) algorithm is utilized to optimally tune the hyperparameters related to the MHALSTM model to enhance the prediction efficacy. The experimental validation of the ASTLP-CBMDL model is tested using open access dataset and the outcomes are examined extensively. The comprehensive result analysis stated the enhanced performance of the ASTLP-CBMDL model over recent approaches.

For improving the accuracy and speed of short-term power load forecasting, a new on-line power load-forecasting method based on regularized fixed-memory extreme learning machine (RFM-ELM) is proposed. This method can choose the prediction model adaptively and adjust model parameters automatically. Considering uncertain factors, actual load data and real time meteorological data are used to train a forecasting model based on RFM-ELM, which improves the load forecasting accuracy effectively. RFM-ELM adopts the latest training sample and abandons the oldest training sample iteratively to achieve the online training of network weights. The structural risk is integrated into the model in order to enhance the generalization ability and robustness of the load-forecasting model. This paper verifies the method and the model by using the real data of a region. Experimental results show that the method significantly increases the precision of prediction. This approach provides superior accuracy and adaptability compared with the method of ELM when applied in short-term load forecasting.

Power load forecasting is important to ensure the stability and reliability of regional power systems. Researchers have put forward many combined forecasting models, but most of them cannot capture the global characteristics of data well. So as to improve the accuracy of short-term power load forecasting, this paper puts forward a combined forecasting model based on long-term and short-term memory networks (LSTM) and time convolution networks (TCN). In terms of the power load data, the LSTM and TCN forecasting models are established at first, and then the output results of LSTM and TCN are weighted and combined according to the reciprocal ratio of the square error, and the LSTM-TCN combined forecasting model is obtained. Finally, an example is analyzed by using the real data of the Australian Energy Administration. The LSTM-TCN model constructed in this paper has more advanced model performance, and its error is obviously lower than that of a single forecasting model and other classical network models, indicating that the LSTM-TCN model has higher accuracy in short-term load forecasting.

Short-term power load forecasting is an important link in power grid dispatching, which has an important impact on unit combination, economic dispatching, optimal power flow, etc. As meteorological factors have great influence on load, the influence of meteorological factors should be considered reasonably in short-term load forecasting. The accuracy of short-term electric charge prediction results can help senior personnel of the power system to make accurate and feasible power operation methods. In order to ensure the safe and stable operation of power grid in various special periods and to ensure the economic benefits of related power enterprises to the greatest extent, it is imperative to establish a highly accurate prediction model. The accuracy of power load forecasting will directly affect the position of each power enterprise in the market. Based on the function of time series embedding, this paper analyzes the rule of cumulative effect on load. The correlation between temperature and load is greatly improved after considering the cumulative effect to deal with temperature correction.

A reliable supply of power systems is critical for industry, commerce, and residential life. Improving the accuracy and reliability of short-term electricity load forecasting plays a crucial role in ensuring the satisfaction of electricity demand and the stable operation of the power system. Therefore, to realize accurate and efficient prediction of short-term power loads, a short-term power load prediction method based on multi-intelligence deep reinforcement learning is proposed to address the complex nonlinear characteristics of load data. In this paper, we analyze the multi-intelligence application architecture in power load forecasting, and analyze the function of each intelligent unit applied to short-term power load forecasting; based on clarifying the interaction relationship of each intelligent unit in short-term power load forecasting, we model short-term power load forecasting as a distributed and partially observable Markov decision-making process, which is suitable for multi-intelligence deep reinforcement learning; based on the MATD3 algorithm, a centralized training-distributed execution framework is used to train multiple intelligences within the model to achieve short-term power load forecasting. The experimental results show that in the August short-term electricity load forecasting using the design method, the obtained MAE value is 35.94 kW, MAPE value is 4.05%, and RMSE value is 32.71 kW. In the short-term power load forecasting evaluation conducted for December, the average absolute error (MAE) value obtained was 36.75 kilowatts, the average absolute percentage error (MAPE) value was 4.51%, and the root mean square error (RMSE) value was 34.82 kilowatts. These evaluation results fully demonstrate that the design method adopted has high prediction accuracy and forecast precision. This method has demonstrated good practical value and broad application prospects in practical applications due to its high-precision prediction performance and strong prediction stability.

Load Forecast (LF) is an important task in the planning, control and application of public power systems. Accurate Short Term Load Forecast (STLF) is the premise of safe and economical operation of a power system. In the research of short-term power load forecasting, machine learning and deep learning are the most popular methods at present, but there still exists a problem that the single and simple structure of power load forecasting model leads to low accuracy of load forecasting. In order to improve the accuracy of STLF, a Gated Cycle Unit (GRU)-Transformer combined neural network model is proposed. Transformer encoder structure is used as feature extractor to mine the complex mapping relationships between the input features and load. The advantage of self-attention mechanism is used to solve the problem of information loss of long sequences in short-term power load forecasting. At the same time, the multivariate time series model of GRU is used for model training. The experimental results on the power load data set of a certain region in southwest China and Panama City show that the proposed combined model prediction method has higher accuracy than those proposed in other literatures, which further proves its feasibility and superiority.

To handle the data imbalance and inaccurate prediction in power load forecasting, an integrated data denoising power load forecasting method is designed. This method divides data into administrative regions, industries, and load characteristics using a four-step method, extracts periodic features using Fourier transform, and uses Kmeans++ for clustering processing. On this basis, a Transformer model based on an adversarial adaptive mechanism is designed, which aligns the data distribution of the source domain and target domain through a domain discriminator and feature extractor, thereby reducing the impact of domain offset on prediction accuracy. The mean square error of the Fourier transform clustering method used in this study was 0.154, which was lower than other methods and had a better data denoising effect. In load forecasting, the mean square errors of the model in predicting long-term load, short-term load, and real-time load were 0.026, 0.107, and 0.107, respectively, all lower than the values of other comparative models. Therefore, the load forecasting model designed for research has accuracy and stability, and it can provide a foundation for the precise control of urban power systems. The contributions of this study include improving the accuracy and stability of the load forecasting model, which provides the basis for the precise control of urban power systems. The model tracks periodicity, short-term load stochasticity, and high-frequency fluctuations in long-term loads well, and possesses high accuracy in short-term, long-term, and real-time load forecasting.

Short-term power load forecasting for combined heat and power using CNN-LSTM enhanced by attention mechanism

A model integrating Particle Swarm Optimization (PSO) and support vector machines (SVM) is presented to forecast short-term load of electric power systems in this paper. PSO is a method for finding a solution of stochastic global optimizer based on swarm intelligence. Using the interaction of particles, PSO searches the solution space intelligently and finds out the best one. The PSO-SVM method proposed in this paper is based on the global optimization of PSO and local accurate searching of SVM. Practical example results indicate that the application of the PSO-SVM method to short term load forecasting of power systems is feasible and effective. And to prove the effectiveness of the model, other existing methods are used to compare with the result of SVM. The results show that the model is effective and highly accurate in the forecasting of short-term power load.

In the process of our daily use of electricity, power load forecasting is very important. Short-term power load forecasting can effectively manage electric energy. In order to improve the intelligence of power grid, this paper studies the analysis and prediction methods of power load based on deep learning. Short-term power load forecasting can accurately predict the load and electricity consumption of a certain area, and provide reference for the operation of the power system. In daily life, our power load data are interfered by many factors, so we need to fully find out the temporal characteristics of power load data, so as to improve the accuracy of our power load forecasting. In this paper, a convolutional neural network-long-term and short-term memory neural network (CNN-LSTM) neural network prediction model based on convolution neural network and long-term and short-term memory neural network is proposed. The data characteristics are extracted by convolutional neural network (CNN), and then the accurate prediction of power load is completed by using the unique memory and prediction ability of long-term and short-term memory neural network (LSTM) neural network. The experimental results show that compared with the single LSTM neural.

Artificial intelligence and machine learning methods have gradually matured and have been widely used in short-term power load forecasting. In order to make better use of the advantages of different artificial intelligence prediction models and traditional prediction models and improve prediction accuracy, this paper proposes a short-term load prediction model based on multi-model stacking. Different from the combined prediction method, the model first uses three machine learning models, support vector machine (SVM), back propagation neural network (BPNN), and extreme learning machine (ELM) as the base learners, and uses different training data sets. Training is performed on the model, and then the prediction results of the three basic learners are used as the input of Gaussian Process Regression (GPR), and multiple models are integrated to obtain the final prediction result. In order to verify the effectiveness of the Stacking prediction model, this paper applies the short-term load data of the PJM market to this model. Compared with the three base learners, the prediction results show that the model can make full use of the advantages of different prediction models and effectively reduce Forecasting errors have practical significance for solving short-term load forecasting problems.

When applying extreme learning machine (ELM) to short-term power load forecasting, its randomized weights and thresholds result in relatively low prediction accuracy and stability. Meanwhile, the pelican optimization algorithm (POA) suffers from the limitation of easily falling into local optima. To address these issues, this study proposes an improved pelican optimization algorithm (IPOA) to optimize ELM for short-term power load forecasting. The proposed method first incorporates an improved one-dimensional chaotic mapping (1-SCEC), Levy flight strategy, and adaptive weight strategy to enhance the optimization capability of POA, with the superiority of IPOA validated through two standard test functions. Subsequently, IPOA is employed to optimize ELM parameters, establishing an IPOA-ELM-based short-term power load forecasting model. The feasibility of the IPOA-ELM model is verified using actual power load forecasting data from an Australian region. Experimental results demonstrate that the proposed model achieves closer prediction results to actual loads for both weekends and weekdays, exhibiting superior prediction accuracy and stability compared to alternative methods.