CHAPTER 6. CONCLUSIONS
6.2 Future Work
The dynamic VAR planning study methodology developed in this thesis work can be further enhanced by: 1) Comparing the effectiveness of different types of dynamic VAR support strate- gies and their cost/benefit assessment; 2) Utilization of reactive support from geographically distributed, power electronics based distributed generation (DG) resources; and 3) Improvi- sation in the dynamic optimization procedure through parallelization, multistart algorithms.
6.2.1 Integrated supply side and demand side solutions
If the system does not experience short-term voltage problems frequently, then load shedding can be considered as a viable and cheaper option to mitigate these problems. In planning studies, while evaluating the dynamic VAR requirements of the system, load shedding can also be considered as an option in addition to adding new VAR resources. To accomplish this, an integrated control scheme that effectively combines both supply-side (addition of dynamic VAR resources) and demand-side controls (load shedding) to meet the dynamic VAR requirements of the system. The optimization formulations discussed in this work can be extended to include demand side controls. A preliminary description of identifying DVCA’s demand side control locations is provided in Chapter4.6. Depending upon the severity of the problem and cost of
installing new devices the user can select appropriate choices to arrive at both economical and compliant solution for short-term voltage problems. Some preliminary results on the benefits offered by load shedding strategies are discussed in Chapter 5.6.1.
In this thesis, the majority of work in identifying DVCA was based on supply-side per- spective. The DVCA identification can be extended to accommodate both supply-side and demand-side controls in an integrated framework. Then the controls in DVCA will include addition of new dynamic VAR resources in the bulk system and rank ordered load shedding locations and amount of load shedding. Further, the reactive support from DG resources and wind farms can also be used as controls in each DVCA. The extension of the formulation pro- vided in this work for multiple resources require processing of large data, and proper modeling (eg: smart inverters) of reactive support capability from DG resources and wind farms.
6.2.2 Reactive support from DG resources
There has been a significantly increasing trend in the development and usage of distributed energy resources including distributed generation (DG) and demand response (DR) in power system operations. DG in the form of solar PV, small wind, biomass, gas-fired micro-turbines, combined heat and power (CHP) resources, and energy storage have been increasingly deployed in the distribution system and expected to continue its growth in the future. DG with inverter controls can be used to provide reactive power and voltage control. Solar PV or electric vehicles (EV) with an inverter, or wind generators with converters can accomplish the same function as STATCOMs or SVCs but at much less cost. This functionality is most beneficial when active power generation is low e.g. solar outside of peak production periods. Also, the DG resources are distributed geographically, therefore providing greater flexibility in providing reactive support. To utilize DG resources for providing reactive support requires significant modeling efforts. One approach is to develop aggregate DG models for transmission level studies. Recently, the modeling and validation work group (MVWG) of the western electricity coordinating council (WECC) expanded its scope of its Renewable Energy Modeling Task Force (REMTF) to address the modeling and representation of PV systems for transmission level studies. The REMTF recognizes the fact that representing distribution-connected PV
systems in transmission studies is more challenging. Another approach is to represent induction motor loads and DG resources at the distribution feeder level and evaluate the dynamic VAR requirements using an integrated transmission-distribution analysis. Since the geographically distributed DG resources are coordinated properly to provide reactive support, these resources can reduce the installments of new dynamic VAR devices.
Smart inverters in PV systems can manipulate and control real power and reactive power independently. These inverters sense local conditions, such as voltage and frequency, and respond with autonomous actions. Their functionalities include volt-var control, frequency- watt control, and provide dynamic grid support as a part of low voltage ride through.
6.2.3 Dynamic optimization solution enhancements
Parallel computations can be used to speed up some time-consuming procedures in the NLP routine to improve overall computational performance of the CVP algorithm. The NLP routine spends majority of its time in calculating the finite difference gradients. Parallel finite difference gradient computation will greatly reduce the overall solution time of the NLP opti- mization. Parallelism can be introduced at two levels, (a) Calculating the constraints Jacobian matrices for different contingencies concurrently (b) Calculating the sensitivities of different candidate locations in parallel to construct the constraints Jacobian matrix corresponding to a single contingency. Parallel computing can also be utilized to increase the computational efficiency of performing time domain simulations for different contingencies while evaluating NLP constraints, calculating KL sensitivity matrix for MILP optimization, and contingency clustering.
The convergence of the CVP algorithm can be improved by using multistart algorithms. The multistart procedure uses many initial points for the NLP optimization problem and can be run in parallel on shared memory multi-processor machines. The NLP optimization routine finds a local optimal solution from each of the initial points and returns the solution with the best objective function as the final solution. The use of many initial points will also help to find a pool of local optimal solutions.
Extension of clustering methods to other power system problems: In this work, clustering methods are used to identify similarly behaving contingencies under various operating condi- tions. The construction of similarity matrix forms the essential part of clustering methods. KL measure of monitored buses for different contingencies is used to construct the similarity ma- trix. By developing appropriate similarity matrix, the clustering methods described in this work can be utilized to study other power system problems like generator coherency identification, contingency classification for studies using steady-state analysis etc.