Electric Generator Modeling and Automatic Generation Controller

Avista Utilities is building a microgrid within the downtown area of Spokane, Washington. In the event of a natural or manmade disaster, Avista wants to build a self-sustaining and self-contained small power system using the two hydroelectric generators downtown in Spokane, some renewable energy sources, large batteries, and a control system to make the microgrid work in a stable and reliable fashion.

Abstract
Downtown Spokane is in need of a reliable backup power in the case that the main power supply lines feeding into the city fail. There are many parts to this, one part is the two hydroelectric generators. These generators located in downtown Spokane must be modeled in such a way that will allow them to run without the assistance of the larger, outside grid. This means that we must model, find the parameters of and develop a control scheme for these two generators.

Outcomes

 * Monroe Street Generator Simulink and PowerWorld models
 * Upper Falls Generator Simulink and PowerWorld models
 * Transfer models into RTDS (Real Time Digital Simulation)

Introduction
There are many models that are available to use for simulation and modeling of hydroelectric generators. This is complicated further as there is more than just the AC machine involved in generation. In addition to the AC Synchronous machine there is a governor, exciter, and an over excitation limiter. As there are many models for each of these parts we must decide on one model for each part and successfully use it in simulation to get an accurate representation of the generators.

Synchronous Machine
A synchronous machine is a popular choice for generating electric power. They come in several types. First there are the round rotor and the salient variety. As the name implies the round rotor variety has a round rotor while the salient is not round as it has protrusions where the magnetic poles are. The salient type is often used with larger, slower machines. Here in the Northwest these are a very common choice for hydroelectric generators. While a round rotor machine is more often used in high speed generation applications such as steam turbines. As such the motors that we want to model are salient.

There are many models of synchronous machines, several that we looked into are the GENROU, GENSAL, GENTPF and GENTPJ. Each of these models has their own advantages and disadvantages. The GENROU is for modeling a round rotor machine; ours being a salient machine this isn't a good option. GENSAL is an older model for a salient machine, it has the advantage in requiring less computation to simulate. The GENTPF and GENTPJ are rather similar, both may be used to model round rotor or salient rotor synchronous machines. We had the parameters for a GENTPJ from testing Avista had completed on one of the generators so looked into this model in depth.



The GENTPJ differs from the GENTPF in it's handling of saturation, in this aspect it is more accurate. It also has the most parameters of any of the generators. Luckily these parameters should be the same between models and as such we can use any of the other models as well.

In the end we ended up going with a MatLab-Simulink model that is likely based on the GENSAL model.

Governor
The Governor is a mechanical and or electrical device that uses a feedback control to change the rate of rotation of the machine to get the desired output. This allows for the control of the output frequency of the machine. For the Synchronous Machine water turbine this is done by controlling the flow of water through the turbine and measuring the rate of rotation of the machine.



Exciter
Main functions of excitation system are to provide variable DC current with short time overload capability, controlling terminal voltage with suitable accuracy, ensure stable operation with network and/ or other machines, contribution to transient stability subsequent to a fault, communicate with the power plant control system and to keep machine within permissible operating range.



Introduction
The Monroe Street Dam was built in 1890 and at it's peak consisted of five generators. Eventually reconstruction of the Monroe Street dam started in 1974 and was finished in 1992. This reconstruction removed the 5 generators from the 1900-era and replaced them with 1 generator utilizing a Kaplan turbine. In the process the generation was moved underground, a beautification aspect for the riverfront park.



Monroe Street Simulink Model
The first step of the model process for this project is to model the generators in Matlab Simulink. Tests have been done for the Monroe Street generator giving us usable parameters. The values from the GENTPJ model were given and can be used with Simulink models from the Sim Power library. The values for the exciter were also in the testing report for the machine. With this information the Simulink model of the generator could be built using a generic exciter model that should be close to the actual model. Below is the final model for the Monroe Street generator in Matlab Simulink with per unit values.



This Simulink model uses a fault system on the line to trigger a transient response. With this response in the system we can verify that this model is correct for the Monroe street generator. The three main blocks of the diagram are the Govern the Exciter and the Synchronous Machine itself. With these models the Monroe Street generator model is complete.

The figures above show the current results of the forced fault in the system and the generator models transient recover from this fault. From the voltage it clearly can be seen that the fault is doing its job forcing the voltage to zero. This forces the rotor to slow and the current to spike.

Introduction
The Upper Falls Generator was installed in the early 1920s by Washington Water and Power that later became Avista Utilities. This generator utilizes a Francis Turbine and outputs up to 10MW. Today it utilizes a solid state exciter and governor controller. However much of it is original.

The Upper Falls Generator will be one of our greatest modeling challenges. This is due to the lack of information we have. We do not have any parameters for the generator, nor are we able to revive data on a transient response to calculate the parameters.



Document Archive

 * Group Minutes