基于CLARA算法的考虑时序特性分布式电源规划

DG planning based on CLARA algorithm with consideration of timing characteristics

针对配电网分布式电源(DG)的优化选址定容问题,以配电网年损耗电量最小为目标,考虑不同类型DG出力的随机性和时序特性,建立多场景多时段混合整数非线性随机优化模型。利用蒙特卡洛方法生成描述DG出力的序列场景,通过考虑各场景需要满足的约束条件近似对随机过程进行约束,将原随机优化问题转化为确定性优化问题。利用CLARA算法对各场景各时段模拟得到的样本进行聚类,以避免全场景下样本规模大、模型求解难等问题。IEEE 14和IEEE 33节点标准系统的测试表明,所提模型和算法能有效利用不同类型DG的时序互补作用,提高配电网对DG出力的消纳能力。与不采用聚类方法全场景代入及采用PAM聚类方法相比,所提算法在保证优化结果近似误差低于3%的同时,能显著降低模型的求解难度和节省计算时间。

Aiming at the optimal locating and sizing of DG（Distributed Generator） in distributed network,a multi-scene multi-period mixed-integer nonlinear stochastic optimization model is built,which takes the minimum annual power loss as its objective and considers the randomness and timing characteristics of different DGs. Monte Carlo method is applied to generate the time-series scenes and the constraints of each scene are considered to approximately constrain the stochastic process for converting a stochastic optimization problem into a deterministic optimization problem. CLARA（ Clustering LARge Application）algorithm is adopted to cluster the samples obtained by simulations for each scene and each period to avoid the large sample scale,model solving difficulty,etc. The tests of IEEE 14- and 33-bus standard systems show that,the proposed model and algorithm can effectively make use of the time-series complementation among different DGs to improve the DG output accommodation capability of distributed network. Compared with the complete scene method without clustering and PAM（ Partitioning Around Medoids） clustering method,the proposed algorithm,with the approximate error of optimized result lower than 3%,reduces the model solving difficulty and shortens the computation time significantly.