Prediction of solar irradiance is essential for minimizing energy costs and providing high power quality in electrical power grids with distributed solar photovoltaic generations. However, for ...residential and small commercial users deploying on-site photovoltaic generations, the historical irradiance data can not be obtained directly because of expensive solar irradiance meters. Thanks to increasingly improved weather forecasting service provided by local meteorological organizations, weather forecasting data such as temperature, dew point, humidity, visibility, wind speed and descriptive weather summary, are becoming readily available through the Internet, while the irradiance forecasting data are often unavailable. This paper proposes a novel solar prediction scheme for hourly day-ahead solar irradiance prediction by using the weather forecasting data. This study formulates the prediction problem as a structured output prediction problem jointly predicting multiple outputs simultaneously. The proposed prediction model is trained by using long short-term memory (LSTM) networks taking into account the dependence between consecutive hours of the same day. We compare persistence algorithm, linear least square regression and multilayered feedforward neural networks using backpropagation algorithm (BPNN) for solar irradiance prediction. The experimental results on a dataset collected in island of Santiago, Cape Verde, demonstrate that the proposed algorithm outperforms these competitive algorithms for single output prediction. The proposed algorithm is %18.34 more accurate than BPNN in terms of root mean square error (RMSE) by using about 2 years training data to predict half-year testing data. Moreover, compared with BPNN, the proposed algorithm also shows less overfitting and better generalization capability. For a case using 10 years of historical data to predict 1 year of irradiance data, the prediction RMSE using the proposed LSTM algorithm decreases by 42.9% against BPNN.
•Weather forecasting data are used as input variables for irradiance prediction.•The prediction problem is formulated as a structured output prediction problem.•LSTM is applied to prediction of solar irradiance.
Model reduction of a high‐dimensional distributed parameter system (DPS) reduces the complexity of the system for various applications, from monitoring to model predictive control, while retaining ...its intrinsic properties. Unfortunately, the assumption of time–space separability usually fails to hold for popular time–space separation model reduction methods because the space and time of the DPS are inherently coupled. In this study, a time–space coupled learning method for a data‐driven model reduction of the DPS is presented. The proposed method has the advantage of preserving the time–space coupling characteristics and increasing the number of degrees of freedom during the model reduction learning process. A novel deep‐learning architecture is presented by combining encoder‐decoder networks with recurrent neural networks. Given a high‐dimensional system without an exact partial differential equation description, the dimension‐reduced model and its temporal dynamics are jointly learned using the collected input and output data. The learned model is then applied to predict the low‐dimensional representations and reconstruct the high‐dimensional outputs. The proposed method was demonstrated on the catalytic rod in a tubular reactor with recycle, the results of which indicate a better modeling accuracy and lower intrinsic dimensionality compared with classical time–space separation model reduction methods.
Nonlinear high‐dimensional distributed parameter systems (DPSs) described by sets of parabolic partial different equations (PDEs) exhibit a dominant, low‐dimensional slow behavior that can be ...captured using model reduction. A time–space‐coupled model reduction architecture combining encoder–decoder networks with recurrent neural networks (RNNs) was presented in our previous work, for modeling the spatiotemporal dynamics of DPSs without recourse to the governing equations. In this work, we further understand the stability of the training dynamics of the deep architecture by using the Lyapunov exponents (LEs). Subsequently, we construct nonlinear model predictive control (MPC) formulations for the DPS based on the learned, dimensional‐reduced model. We use a path‐integral optimal control algorithm for MPC implementation to avoid any analytic derivatives of the dynamics. The effectiveness of integration of a deep neural network‐based model with MPC is demonstrated in a tubular reactor with recycle cases. The results of the simulation also show that the LE can serve as a readout of training stability for the learned dynamical model.
Objective: Daytime short nap involves individual physiological states including alertness and drowsiness. In order to have a better understanding of the periodical rhymes of physiological states and ...then promote a good interpretability of alertness, the aim of this study is to detect drowsiness during daytime short nap. Methods: A method of Bayesian-copula discriminant classifier (BCDC) was introduced to detect individual drowsiness based on the physiological features extracted from electroencephalogram (EEG) signals. As an extension of traditional Bayesian decision theory, the BCDC method tries to construct the class-conditional probability density functions by exploiting the theory of copula and kernel density estimation. Results: The proposed BCDC method was validated with experimental dataset and compared with other traditional methods for drowsiness detection. The obtained results showed that our method outperformed other methods in terms of three evaluation criteria. Conclusion: Our proposed method is effective to detect drowsiness with superior performance. Additionally, the BCDC method is relatively robust to different parameter settings on the group-level dataset. Significance: The proposed method is likely to be a useful tool to improve the correctness of the estimated class-conditional probability density functions. Since features are extracted from spontaneous EEG recordings, the results of this study can be further generalized to other experimental environment to detect vigilance level or driver drowsiness.
As a volcanic archipelago, the Republic of Cape Verde relies dominantly on diesel to power its electricity supply. Recognizing the financial and environmental burden of diesel generation and risk of ...energy security, the government of Cape Verde has launched an ambitious goal of 50% electricity from renewables by 2020, since the country is endowed with high potential of renewable energy resources such as wind and solar. Although the annual average penetration rate of wind power has reached 24% of total electricity production generated in Cape Verde, raising the wind energy penetration level in future will pose numerous challenges for the operation and control of the power system because of wind's inherent intermittency and unpredictability. In this study, a statistical analysis of the wind characteristics in Santiago island, is presented by using historical wind speed and power data of the Santiago wind farm in 2014. A two-parameter Weibull distribution is first applied to model the wind speeds on various timescales and to determine wind energy potential in Santiago island, Cape Verde. The annual average wind speed was 8.57 m/s with a standard deviation close to 3.29 m/s. The monthly Weibull scale parameters varied from 5.64 m/s to 13.7 m/s, while the monthly Weibull shape parameters varied from 1.97 to 9.13. Although the monthly mean power density of the rainy season from August to September was low, the annual mean power density shows that Santiago has good wind potential. Then, an approach to modeling the equivalent power curve based on available wind speed and power output data from the wind farm is proposed. By utilizing the estimated power curve, the uncertainty set of wind power generation, resulted from the uncertainty of wind speed forecast, can be obtained to quantify the power system reserve requirements. A statistical analysis of wind power ramp is also given for estimating the power capacity requirement of the energy storage system that can be considered as a reasonable way to mitigate the wind intermittency and minimize curtailment of wind. Results of this study contribute to assess the wind energy potential of Cape Verde for investors, and can be used to quantify the uncertainties of wind power generation for the power system operator.
•It presents a method to obtain the uncertainty set of wind power generation from the uncertainty of wind speed forecast. The uncertainty set can quantify the power system reserve requirements.•It determines the appropriate power size of the energy storage system by using the results of statistical analysis.
This paper presents a simple and theoretically sound submodule-based model to simulate the characteristics of a photovoltaic (PV) array with a series-parallel configuration. The proposed model can ...describe the behavior of bypass diodes as well as the full PV array characteristics under varying irradiance and temperature conditions. Rather than using the nonlinear system of equations solved with a Jacobian matrix, separate equations are employed to model the submodule-based PV array and solved by an easy-to-implement bisection search method. Consequently, the output current of the PV array can be readily determined when its output voltage, the irradiance levels, and temperature values of all submodules are given. The robustness and calculation efficiency of the proposed computational method are analyzed. Some test examples allow us to exhibit the acceptable accuracy of the proposed model. Special attention of this work is paid to the simulation approach to evaluate the electrical mismatch losses in large-scale PV arrays with nonuniform aging after several years of field operation and exposure.
► The extended set-membership filter (ESMF) is employed for power system dynamic state estimation. ► The ESMF provides 100% confidence to the estimated states for the safety and reliability of power ...systems. ► The proposed method is tested on two different single machine infinite bus systems and a multi-machine system. ► Simulation results demonstrate the effectiveness of the proposed method.
A new method for power system dynamic state estimation is presented. It is based on the application of the extended set-membership filter (ESMF). The ESMF provides an on-line nonlinear guaranteed estimation that assumes that noise sources are unknown but bounded, rather than stochastic as in probabilistic estimation algorithms such as the extended Kalman filter (EKF) and the unscented Kalman filter (UKF). Thus, the ESMF provides 100% confidence to the estimated states for the safety and reliability of power systems. In this paper, the ESMF is derived and demonstrated by using two different single machine infinite bus (SMIB) systems. The performance of the ESMF is compared with the classical UKF by using one SMIB system modeled as a second-order nonlinear dynamic equation. The feasibility of the ESMF method with unknown inputs is also presented on the other SMIB system, in which the exciter output voltage may be unavailable but bounded. The ESMF method is also tested on a multi-machine system model to demonstrate its effectiveness.
In recent years, with the increasing proportion of photovoltaic (PV) power generation in grid-connected microgrids, suppressing power fluctuations at the point of common coupling (PCC) has become a ...challenge. This paper proposes a collaborative power dispatch algorithm for battery energy storage systems (BESSs) based on multi-agent reinforcement learning (MARL), aiming to suppress the PCC power fluctuations caused by the uncertainty of PV power generation. First, a distributed multi-agent communication framework is developed, which defines the neighboring areas of agents based on the physical distances between BESSs to reduce the communication and computational cost of agents. Subsequently, a distributed multi-agent dueling double deep Q-network power dispatch algorithm based on the communication framework is proposed. In the proposed algorithm, a distributed Markov decision process is designed, enabling agents to share actions and rewards with neighboring agents locally to collaboratively learn optimal charging and discharging actions for suppress PCC power fluctuations. Finally, the scalability and effectiveness of the proposed algorithm in suppressing PCC power and voltage fluctuations and reducing operational cost are validated through simulation experiments based on the IEEE-33 bus and IEEE-141 bus systems. The simulation results demonstrate significant advantages of the proposed algorithm compared with other baseline MARL and traditional optimization methods.
Display omitted
•A study on PCC power suppression in AC grid-connected microgrids based on BESSs.•A distributed multi-agent communication framework is developed to reduce cost.•A distributed MAD3QN algorithm is proposed to collaboratively learn optimal actions.•Results show proposed algorithm outperforms baseline MARL and optimization methods.
•A decentralized unscented Kalman filter for multi-area dynamic state estimation in power systems is proposed.•The detailed consensus algorithm is presented.•Comparison of the results reveals the ...impressiveness of the algorithm.
A decentralized unscented Kalman filter (UKF) method based on a consensus algorithm for multi-area power system dynamic state estimation is presented in this paper. The overall system is split into a certain number of non-overlapping areas. Firstly, each area executes its own dynamic state estimation based on local measurements by using the UKF. Next, the consensus algorithm is required to perform only local communications between neighboring areas to diffuse local state information. Finally, according to the global state information obtained by the consensus algorithm, the UKF is run again for each area. Its performance is compared with the distributed UKF without consensus algorithm on the IEEE 14-bus and 118-bus systems. The low communication requirements and high estimation accuracy of the decentralized UKF make it an alternative solution to the multi-area power system dynamic state estimation.
As a data-driven, equation-free decomposition method, the DMD can characterise dynamic behaviour of a non-linear system by using the DMD modes and eigenvalues. However, all current provable ...algorithms suffer from a separate procedure for obtaining the DMD modes and determining the number of modes. In this study, the authors propose a nuclear norm regularised DMD (NNR-DMD) algorithm that produces low-dimensional spatio-temporal modes. A nuclear norm regularisation term is added to the optimisation problem of the standard DMD algorithm for prompting the sparsity of the projected DMD modes. Split Bregman method is applied to solve the regularised convex, but non-smooth optimisation problem. Several numerical examples demonstrate the potential of the proposed NNR-DMD algorithm: (i) it can identify the low-dimensional spatio-temporal DMD modes in which each of them possesses a single temporal frequency; (ii) the reconstruction errors based on the sparse DMD modes can be reduced when it compares with the sparsity-promoting DMD algorithm penalising the l1-norm of the vector of DMD amplitudes; and (iii) it can obtain low-dimensional coherent structures when the NNR-DMD algorithm is applied to coherency identification of generators in an interconnected power system.
