Research On Multi-energy Cooperative Control Strategy Based On Source Grid Load Situation

In recent years,the provincial power grids encounter the new problems such as the slowing power load growth,rapid increase of new energy power generation capacity,the difficulty of valley frequency modulation and peak shaving during heating period,large fluctuation net of inter-grid exchange power.There are some solutions by coordinating source,grid and load,but few have focused on the power control mechanism of power source side coordination.In order to solve the problem of unclear coordination mechanism on the power source side,this dissertation proposes the multiple energy generation control strategy(MEGC).According to the characteristics of each power source and the grid operation modes,the real-time cooperative control method of multiple types of power source is proposed.A wind-PV-hydropower-thermal coordinated automatic power generation control mechanism was built.And the MEGC platform was developed for Liaoning power grid.The demonstration results proved that the proposed strategy promoted the safe and stable operation of power source side and clean energy consumption effectively.The scientific problem of the dissertation lays that the short time-scale operation state is predicted by the long time scale data of whole year.The running state point is got by the analysis of the power source side operation characteristics.Based on the combination of point prediction and running state,the future power system operation states are matched.And,the optimization results are got by the power source side point and scenario prediction and the future state optimizing.The multi-energies are matched precisely by prediction point,scenic point and operation optimization point.The important innovations in this dissertation include three parts: i)short-term point prediction of future state of typical power source based on deep adaptive filtering;ii)future state scenario generation of typical power source based on MD-K2-Bayes network and improved K-means;iii)the coordinated and optimized scheduling of multi-energy systems for wind,hydropower,thermal and storage;iv)the platform demonstration of MEGC collaborative power generation control system.Firstly,the multi-power source operation law of 8760 hours around one year is analyzed to generate the power source side operation state scenario.Secondly,combined with the 15-minute day-ahead prediction results of typical power points,the future states of the power system are matched.Finally,the multi-time scale optimization operation points of future state of wind-PV-hydropower-thermal power are solved.The main contents are as follows:1)Different prediction models are selected adaptively based on different predicted objects.A short-term point prediction method is proposed for the future state of typical power source based on deep adaptive filtering.Firstly,Kullback Leibler(KL)divergence was used for automatic selection of variational mode decomposition(VMD)core parameters to improve model adaptability and reduce human experience error.Secondly,in order to prevent mode aliasing and beneficial information loss of high-frequency components,the data sequence is decomposed by KL-VMD.And,the effective components and noise dominant components are determined based on sample entropy.Aiming at the noise dominant components,the non-local means denoising(NLM)method is used to further get the feature information.Then the effective component and the filtered dominant component are reconstructed to obtain the denoising sequence.Finally,an object-based alternative library prediction method is proposed to select the optimal model to predict the future operation points of typical power source.2)The wind-PV-hydropower-thermal power operation law is analyzed based on long time scale data.MD-K2-Bayes network is firstly applied in the generation of wind-PV-hydropower-thermal power scene.The k-means algorithm is improved.The center curve of each power source is determined.The difficult problems of initial clustering center and the number of algorithm clusters selecting are solved.Then the operation state scene of each power source is formed.Secondly,by combining point prediction and operational scenario,the trained MD-K2-Bayes network is transformed into a multi-objective constrained optimization problem,so as to predict the scene and generate the wind-PV-hydropowerthermal power.The future operation state of wind-PV-hydropower-thermal power is matched with the operating state scene and future state scene.And the future state curve of each power source at short time scale is obtained to achieve future state scene prediction.Finally,the practical correctness and practicability of the proposed improved MD-K2-Bayes network are verified by the examples of the operating state and the future state.3)Firstly,the day-ahead/intra-day optimal scheduling model is built based on the output characteristics of wind-PV-hydropower-thermal multiple power sources.The scheduling strategy is obtained,which makes full use of renewable energy by the analytic hierarchy process(AHP)to promote the consumption of renewable energy.Secondly,the wind-PV-hydropower-thermal power automatic generation control is built based on the intra-day rolling scheduling strategy.For the case that conventional power depth peak shaving and spare are insufficient,the turn on and off of heat storage and electricity storage,and emergency regulation of wind power and photovoltaic are controlled automatically.It promotes storage adjustable capacity of hydropower and thermal and makes full use of the new energy.The schedulable plenty of AGC is improved.The effective coordination,full consumption and efficient control are improved.