Data-driven Optimal Dynamic Dispatch for Hydro-PV-PHS Integrated Power Systems Using Deep Reinforcement Learning Approach

To utilize electricity in a clean and integrated manner,a zero-carbon hydro-photovoltaic(PV)-pumped hydro storage(PHS)integrated power system is studied,considering the uncertainties of PV and load demand.It is a challenge for operators to develop a dynamic dispatch mechanism for such a system,and traditional dispatch methods are difficult to adapt to random changes in the actual environment.Therefore,this study proposes a real-time dynamic dispatch strategy considering economic operation and complementary regulatory ability.First,the dynamic dispatch of a hydro-PV-PHS integrated power system is presented as a multi-objective optimization problem and the weight factor between different goals is effectively calculated using information entropy.Afterwards,the dispatch model is converted into the Markov decision process,where the dynamic dispatch decision is formulated as a reinforcement learning framework.Then,a deep deterministic policy gradient(DDPG)is deployed towards the online decision for dispatch in continuous action spaces.Finally,a case study is applied to evaluate the performance of the proposed method based on a real hydroPV-PHS integrated power system in China.Simulations show that the system agent reduces the power volatility of supply by 26.7%after hydropower regulating and further relieves power fluctuation at the point of common coupling(PCC)to the upperlevel grid by 3.28%after PHS participation.The comparison results verify the effectiveness of the proposed method.

Parallel Dispatch:A New Paradigm of Electrical Power System Dispatch

Modern power systems are evolving into sociotechnical systems with massive complexity, whose real-time operation and dispatch go beyond human capability. Thus,the need for developing and applying new intelligent power system dispatch tools are of great practical significance. In this paper, we introduce the overall business model of power system dispatch, the top level design approach of an intelligent dispatch system, and the parallel intelligent technology with its dispatch applications. We expect that a new dispatch paradigm,namely the parallel dispatch, can be established by incorporating various intelligent technologies, especially the parallel intelligent technology, to enable secure operation of complex power grids,extend system operators' capabilities, suggest optimal dispatch strategies, and to provide decision-making recommendations according to power system operational goals.

Reserve Constrained Dynamic Economic Dispatch with Valve-point Effect:A Two-stage Mixed Integer Linear Programming Approach

This paper proposes a deterministic two-stage mixed integer linear programming(TSMILP)approach to solve the reserve constrained dynamic economic dispatch(DED)problem considering valve-point effect(VPE).In stage one,the nonsmooth cost function and the transmission loss are piecewise linearized and consequently the DED problem is formulated as a mixed integer linear programming(MILP)problem,which can be solved by commercial solvers.In stage two,based on the solution obtained in stage one,a range compression technique is proposed to make a further exploitation in the subspace of the whole solution domain.Due to the linear approximation of the transmission loss,the solution obtained in stage two dose not strictly satisfies the power balance constraint.Hence,a forward procedure is employed to eliminate the error.The simulation results on four test systems show that TSMILP makes satisfactory performances,in comparison with the existing methods.

An Efficient Multi-objective Approach Based on Golden Jackal Search for Dynamic Economic Emission Dispatch

Dynamic Economic Emission Dispatch(DEED)aims to optimize control over fuel cost and pollution emission,two conflicting objectives,by scheduling the output power of various units at specific times.Although many methods well-performed on the DEED problem,most of them fail to achieve expected results in practice due to a lack of effective trade-off mechanisms between the convergence and diversity of non-dominated optimal dispatching solutions.To address this issue,a new multi-objective solver called Multi-Objective Golden Jackal Optimization(MOGJO)algorithm is proposed to cope with the DEED problem.The proposed algorithm first stores non-dominated optimal solutions found so far into an archive.Then,it chooses the best dispatching solution from the archive as the leader through a selection mechanism designed based on elite selection strategy and Euclidean distance index method.This mechanism can guide the algorithm to search for better dispatching solutions in the direction of reducing fuel costs and pollutant emissions.Moreover,the basic golden jackal optimization algorithm has the drawback of insufficient search,which hinders its ability to effectively discover more Pareto solutions.To this end,a non-linear control parameter based on the cosine function is introduced to enhance global exploration of the dispatching space,thus improving the efficiency of finding the optimal dispatching solutions.The proposed MOGJO is evaluated on the latest CEC benchmark test functions,and its superiority over the state-of-the-art multi-objective optimizers is highlighted by performance indicators.Also,empirical results on 5-unit,10-unit,IEEE 30-bus,and 30-unit systems show that the MOGJO can provide competitive compromise scheduling solutions compared to published DEED methods.Finally,in the analysis of the Pareto dominance relationship and the Euclidean distance index,the optimal dispatching solutions provided by MOGJO are the closest to the ideal solutions for minimizing fuel costs and pollution emissions simultaneously,compared to the latest published DEED solutions.

Advances of machine learning in multi-energy district communities‒ mechanisms, applications and perspectives

Energy paradigm transition towards the carbon neutrality requires combined and continuous efforts in cleaner power production, advanced energy storages, flexible district energy demands and energy management strategies. Applications of cutting-edge machine learning techniques can improve the system reliability with advanced fault detection and diagnosis (FDD), automation with agent-based reinforcement learning, flexibility with model predictive controls, and so on. In this study, a comprehensive review on artificial intelligence applications in carbon-neutral district community, has been conducted, from perspectives of energy supply, energy storage, district demands and energy management. Classifications and underlying mechanisms on ML techniques have been demonstrated, including supervised, unsupervised, reinforcement and deep learning. Afterwards, practical applications of ML have been reviewed, in respect to renewable energy supply, hybrid energy storages, district energy demand and advanced energy management. Results indicate that, supervised learning was mainly applied in classification and regression, and unsupervised learning was mainly applied in clustering. The reinforcement learning is mainly applied in on-line optimal scheduling for building energy management. With respect to clean energy supply, ML in solar and wind energy systems mainly include solar irradiance forecasting, wind resource forecasting, PV power prediction, maximum power point tracking (MPPT) for smart control, fault detection and diagnosis. ML in fuel cells mainly includes performance prediction, material selection, combination and so on. Furthermore, in respect to hybrid energy storages, ML in electrochemical battery includes dynamic thermal/ electrical behavior, battery sizing and optimization, state-of-charge prediction, battery lifetime estimation, fault detection and diagnosis analysis. ML in sensible energy storages mainly include load prediction and storage capacity sizing, dynamic scheduling for cost saving, thermal stratification analysis and dynamic performance prediction. Advances in energy management with ML mainly include dispatch on stochastic and intermittent renewable power, microgrid adaptive control, smart energy trading with controls and decision-marking. Research tendency over the recent past several years indicates that, critical areas for low-carbon energy systems transit from the only renewable systems (59.4% in 2016) towards both renewable energy supply and energy storages (35.1% and 34.1%, respectively), such as battery, capacitors/supercapacitors, sensible/latent heat storages, compressed air storage and hydrogen storage. This study can provide a holistic overview and in-depth thinking on artificial intelligence in the carbon-neutral district transition.