Research On The Method And Application Of Optimal Scheduling For Hydropower Station Group Based On Double-layer Parallel Algorithm

In China,with the rapid expansion of the scale of hydropower station group in large river basins,the problem of optimal scheduling is becoming more and more complicated.In order to give full play to the role of compensation and adjustment among the hydropower stations,and improve the efficiency of hydropower utilization,many experts and scholars have carried out research on the related theory and solution method of the optimal scheduling model according to the characteristics of the hydropower station group.However,the optimal solution accuracy of the scheduling model and the calculating time of the corresponding algorithm are two major mutual restraint contradictions.In recent years,with the rapid development of computing technology,this opportunity can be used to make full use of computing resources.According to the different characteristics of models and algorithms,corresponding parallel algorithms are proposed.On the basis of ensuring or improving the accuracy of optimal solutions,to a certain extent It can relatively reduce or keep the calculating time of the original model optimal solution.Therefore,this article takes the cascade hydropower station group composed of some adjacent series hydropower stations(with annual adjustment function)in some large watersheds in southwestern China as the research background.According to the theoretical method and model characteristics of optimal scheduling for cascade hydropower station group,in order to obtain ideal optimal results,The double-layer parallel algorithms are proposed under the two architecture modes of networked multi-core computers and the single multi-core computer.The research on how to improve the calculating efficiency is carried out,which can further improve the management level of the optimal scheduling of cascade hydropower station group.The main research results of this article are as follows.(1)Applying artificial fish swarm algorithm to the generation optimal scheduling model of hydropower station,in order to avoid falling into the local optimal solution,it can be combined with chaos optimal algorithm to expand the traversal search range of single artificial fish.On the other hand,it can increase the number of single artificial fish,which further improves the accuracy of the optimal solution of the optimal scheduling model.However,the calculating time will increase to a certain extent.In response to this problem and the characteristics of the model,an in-depth analysis of its parallelizable factors is carried out.Based on the hardware of interconnection system combined with multiple multi-core computers,a double-layer parallel algorithm for chaotic artificial fish school is proposed.The generation optimal scheduling model has been verified by the example of one single hydropower station located in a watershed,southwest China.The results show that the double-layer parallel algorithm not only improves the accuracy of the overall optimal solution,but also effectively controls the calculating time within an acceptable range.(2)Applying the traditional dynamic programming algorithm to the optimal scheduling model of the hydropower station,increasing the discrete number of initial and final storage state variables in one period can further make the overall optimal solution converge to the theoretical optimal value,but the calculating time will increase exponentially.To solve this problem,based on the parallelism of the cyclic calculation of the initial and final storage state variables in one period,the double-level parallel dynamic programming algorithm is proposed.The example verification of a power generation optimal scheduling model of one hydropower station in a watershed in southwest China shows that the algorithm has a good check and balance effect on the two contradictions between the overall optimal solution accuracy and the calculating time.(3)When the standard dynamic programming algorithm is applied to the power generation optimal scheduling of cascade hydropower station group,if improving the accuracy of the optimal solution,the phenomenon of "dimensionality disaster" is easy to occur.Parallel processing of the calculation can reduce the calculating time to a certain extent,but it is not effective to reduce dimensionality,and the calculating time is still more.The secondary hierarchical structure scheduling model based on the principle of large-scale system decomposition-coordination can effectively reduce dimension,and the standard dynamic programming algorithm with multi-threading technology parallelizing can be used to find the optimal solution of the decomposed hydropower station subsystem.With the analysis and research of parallelizable factors in standard dynamic programming algorithm,an optimal scheduling model of cascade hydropower station group is proposed,which is based on large-scale system decomposition-coordination combined with double-layer parallel dynamic programming of single hydropower station after its decomposition.The parallel factors are mainly from two kinds of fine-grained sub task division for initial and final discretized state variables of reservoir capacity in each period after each hydropower station decomposed,and the accuracy of optimal results are improved by expanding the degree of discretization.The effectiveness of this method is verified by taking the cascade hydropower station group in a basin in Southwest China as an example.(4)Based on the large-scale system decomposition-coordination method and the system architecture of the interconnection by some multi-core computers,a heterochrony starting double-layer parallel computing for optimal scheduling model of cascade hydropower station group is proposed.The basic idea of this method is to distribute multi-core computers to their corresponding numbered hydropower stations respectively at first,and to realize independent calculating of each hydropower station by heterochrony starting parallel algorithm,which is the first layer of parallel computing by coarse-grained task partitioning;the second layer of parallel computing is based on parallel processing of standard dynamic optimal calculation of each hydropower station after its decomposition,which is a fine-grained sub-task partition for initial and final discretized state variables of reservoir capacity in each period.Taking a cascade hydropower station group in a basin in Southwest China as an example,it is proved that the method can effectively control the calculating time and the overall optimal solution accuracy.