Takaaki Ohishi

Optimal active power dispatch by network flow approach

An optimal active power dispatch problem is formulated as a nonlinear capacitated network flow problem with additional linear constraints. Transmission flow limits and both Kirchhoff laws are taken into account. The problem is solved by a generalized upper bounding technique that takes advantage of the network flow structure of the problem. The proposed approach has potential applications to power system problems such as economic dispatch, load supplying capability, minimum load shedding and generation-transmission reliability. The authors also review the use of transportation models for power system analysis. A detailed illustrative example is presented. >

A short term hydrothermal scheduling approach for dominantly hydro systems

A short-term hydrothermal scheduling approach is presented for predominantly hydroelectric systems. The model takes into account both the operating hydroelectric system and the electric transmission network constraints. The model consists of a simulation of the hydraulic system with the discharge decisions given by an optimal DC power flow algorithm. The release targets of the reservoirs, established by long-term operational planning, are enforced by a dual Lagrangean approach that fixes a penalty for the use of water in the reservoirs. Two illustrative examples have been solved in order to evaluate the efficiency of the approach. >

Dynamic dispatch of hydro generating units

This work is concerned with the dynamic dispatch of hydro generating units. A performance criterion that takes into account variations in tailrace elevation, penstock head losses and turbine-generator efficiencies is considered. A heuristic procedure based on Lagrangian relaxation is applied to solve the dynamic dispatch problem of scheduling generation on an hourly basis during a day. The approach has been tested on a hydro system composed of nine hydro plants of the Brazilian power system. The generation scheduling verified on a typical day was considered for comparison with the solution provided by the approach and the results show a significant improvement in terms of hydro generation efficiency.

A short-term bus load forecasting system

This paper proposes a methodology for a short-term bus load forecasting. This approach calculates the short-term bus load demand forecast using few aggregated models. The idea is to cluster the buses in groups with similar daily load profile and for each cluster one bus load forecasting model is adjusted. For each cluster, aggregated forecasting model is built based on the analysis of individual bus load data. The solution obtained through aggregated approach is similar to the solution obtained by individual bus load forecasting model, but requiring much less computational time. This proposed methodology was implemented in a friendly computational forecasting support system described in this paper.

Short-term hydropower scheduling via an optimization-simulation decomposition approach

This paper presents a decomposition approach based on an optimization-simulation approach to short-term hydropower scheduling. The problem is formulated as a mixed-integer nonlinear programming problem where the decision variables are the power output and the number of units committed at each hydro plant and hour of the day or week ahead. The goal consists of maximizing hydropower efficiency while reducing startup/shutdown costs and attaining system load, operational constraints, as well generation targets established by mid-term operation scheduling models. The approach proposed in this paper solves a relaxed version of the original problem in which hydraulic constraints are ignored. Eventual hydraulic infeasibilities are computed by a simulation step in order to either validate the solution or add violated constraints back into the problem. The approach is implemented and tested over the Brazilian power system for a study case comprised of 95 hydro plants, 447 generating units, and an average load of 41 GW for a week long horizon. Results confirm the approach to be very efficient in terms of computational costs, and both unit commitment and generation schedules.

Clustering bus load curves

This paper proposes the clustering of a set of busses through a fuzzy c-means clustering approach. The utilization of fuzzy techniques in a clustering problem aims the attainment of a partition fuzzy in the data set, allowing degrees of relationship between different elements of the set, this way, an element can belong to more than one group with different membership value. In this paper the clustering algorithm aims at exploring data characteristics and determining groups composed by busses with similar bus daily load. Results show the efficiency of the clustering method, where the data was classified into distinct groups such as: commercial, residential and industrial consumption profiles.

An hybrid aggregate model applied to the short-term bus load forecasting problem

In this paper we present a hybrid methodology built on a combination of clustering and forecasting techniques used to solve the short-term bus load forecasting problem. The proposed method was made in two phases: In the first phase a clustering algorithm is used to identify buses clusters with similar daily load profile and in the second phase is proposed an aggregate structure for to foresee each bus using a conventional prediction model. The methodology was applied on bus load data from the Brazilian North/Northeast system and the results showed that the model was efficient with 2% to 3.6% of the mean percentage error level on the buses.

Comparison of two heuristic approaches to hydro unit commitment

This paper is concerned with the unit commitment of hydro generating units on an hourly basis throughout a single day. The performance criterion to be optimized includes the efficiency of hydro conversion, which depends on variations in tailrace elevation, penstock head losses and turbine-generator efficiency, as well as the cost of startup/shutdown of the hydro generating units. The paper presents a comparison of two heuristic approaches for solving the optimization problem. One heuristic is based on decomposing the problem into Generation Scheduling (GS) and Unit Scheduling (US) sub-problems, and solving the sub-problems by Lagrangian Relaxation and Dynamic Programming, respectively. The other heuristic makes use of a Genetic Algorithm combined with Lagrangian Relaxation to solve the original problem. The two heuristics were tested on a system composed of sixteen hydro plants, one hundred generating units, and an installed capacity of 21,933 MW in the Brazilian power system. The actual scheduling of generation actual for a typical day was used for comparison with the solutions proposed by the two heuristics. The results of both heuristics show significant savings in terms of hydro conversion efficiency and startup/shutdown costs.

An aggregate model applied to the short-term bus load forecasting problem

In this paper we present a methodology based on a combination of clustering and forecasting techniques. The proposed method is built in two phases: In the first phase, a clustering algorithm is used to identify buses clusters with similar daily load profile. In the second phase we introduce an aggregate structure for to foresee each bus. The methodology was applied on bus load data from the Brazilian North/Northeast system and the results showed that the model was efficient with 2% to 4% average percentage error level on the buses. The obtained forecasting was compatible with the load safe operating levels of the Brazilian power system.

Ensembles of Selected and Evolved Predictors using Genetic Algorithms for Time Series Prediction

This work proposes the use of Neural Networks Ensembles to predict future values of an electrical load time series. At first, to generate these ensembles it is necessary to make several predictions of the same time series using various different networks in which every single one alone is sufficiently competent to predict the above mentioned time series. Therefore, we applied Genetic Algorithms to evolve the parameters of four types of networks: MLPs Neural Networks, Recurrent Neural Networks, Radial Basis Neural Networks and Neuro-fuzzy Networks. As a result, we came up with a set of genetically evolved networks as possible candidates to compose the final ensemble. Finally, in order to achieve a better model, selections (using Genetic Algorithms) of the most suitable networks were made to compose the final ensembles.