Due to the development of photovoltaic (PV) technology and the support from governments across the world, the conversion efficiency of solar energy has been improved. However, the PV power output is influenced by environment factors, resulting in features of randomness and intermittency. These features may have a negative influence on power systems. As a result, accurate and timely power prediction data is necessary for power grids to absorb solar energy. In this paper, we propose a new PV power prediction model based on the Gradient Boost Decision Tree (GBDT), which ensembles several binary trees by the gradient boosting ensemble method. The Gradient Boost method builds a strong learner by combining weak learners through iterative methods and the Decision Tree is a basic classification and regression method. As an ensemble machine learning algorithm, the Gradient Boost Decision Tree algorithm can offer higher forecast accuracy than one single learning algorithm. So GBDT is of value in both theoretical research and actual practice in the field of photovoltaic power prediction. The prediction model based on GBDT uses historical weather data and PV power output data to iteratively train the model, which is used to predict the future PV power output based on weather forecast data. Simulation results show that the proposed model based on GBDT has advantages of strong model interpretation, high accuracy, and stable error performance, and thus is of great significance for supporting the secure, stable and economic operation of power systems.

A combined delivery of economic dispatch and frequency regulation (FR) services from electric vehicles (EVs) is proposed to exploit the dynamic flexibility of EV charging load and its revenue opportunities in the electricity markets. The economic feasibility is assessed with the combined operation model of EVs for optimal bidding in the day-ahead energy market and FR market. The proposed optimization adopts the hierarchical control framework with EV aggregator (EVA) as the intermediate control unit to interact with the power system operator. As EVs are paid for FR based on their actual performance, a dynamic control strategy for charging EVs in response to regulation signals is developed to validate the technical feasibility of combined operation. The autonomous control of EVA modulates the charging of each EV to ensure the optimal charging schedules and real-time frequency control. A case study reveals that the profitability of EV charging control can be significantly increased by energy cost reduction and additional regulation revenue. The proposed dynamic control with a novel power-sharing strategy of EVA provides effective frequency support based on the optimal charging schedule and capacity bids without violating charging demand and availability constraints of each individual EV.

To realize accurate estimation of SOC (state of charge) and SOH (state of health), Li-ion battery’s operating characteristic is analyzed in this paper while fully considering temperature, degree of aging and other practical factors that could impact their operating status. On the basis of nonlinear autoregressive with exogenous inputs (NARX) architecture, an improved dynamic recurrent neural network (DRNN) with ability of dynamic mapping is established, which is more suitable than static network for estimating the batteries’ state with strongly nonlinear and dynamic behaviors. Meanwhile, a self-adaptive weight particle swarm optimization (SWPSO) algorithm is introduced for training the network. Compared with gradient descent algorithm, the SWPSO algorithm could improve error convergence speed and avoid falling into local optimum. Validation results highlights that the presented method is able to improve the estimation accuracy of the SOC and SOH under different conditions including temperature, current and degree of aging, and has strong robustness and ability of generalization.

Due to the continual fusion reaction, the sun generates tremendous energy. This solar energy is freely available and can be extracted by installing a large-scale solar power plant. Therefore, such PV solar plants are key contributors to cutting the energy deficit in remote areas. This study focused on predicting a 10-year performance analysis of a large-scale solar power plant by using 1 year of real-time data from the Quaid-e-Azam Solar Park (QASP) situated in Bahawalpur, Pakistan. For the purpose of prediction, the ARIMA model was developed using Python, which is one of the best tools in machine learning. Since ARIMA is a statistical technique for prediction, by using the developed model through Python, we predicted the values of the performance ratio (PR), production amount (MWh), and plan of array (POA) of the solar plant for the next 10 years using 1 year of real-time data. This machine learning prediction technique is very effective and efficient, compared with other traditional prediction and forecasting techniques, for estimating the performance of the solar power plant and the status of the solar power plant in the long-term future. The forecasting/prediction results acquired from the process show that power production during the next ten years increases to approximately 400 MW and that POA will grow from 6.8 to 7.8 W/m². However, a decline occurred in the performance ratio, which decreased from 76.7% to 73%. Based on these results, the ARIMA model for predicting solar power generation is effective and can be utilized for any solar power plant.

In order to effectively identify complex power quality disturbances, a power quality disturbance classification method based on empirical wavelet transform and a multi-layer perceptron extreme learning machine (ELM) is proposed. The model uses the discrete wavelet transform (DWT) multi-resolution method to extract classification features. Combined with hierarchical ELM (H-ELM) characteristics, the particle swarm optimization (PSO) single-object feature selection method is used to select the optimal feature set. The hidden layer of the H-ELM classifier in the model is trained by forward training. Once the previous layer is established, the weight of the current layer can be fixed without fine-tuning. Therefore, the training speed can be accelerated, the recognition accuracy is almost independent of the parameter adjustment, and the model has strong robustness. In order to solve the problem of data imbalance in the actual power system, a data enhancement method is proposed to reduce the impact of data imbalance and enhance the generalization performance of the network. The simulation results showed that the proposed method can identify 16 disturbances efficiently and accurately under different noise conditions, and the robustness of the proposed method is verified by the measured data.