Developing of method for primary frequency control droop and deadband actual values estimation

Operation of thermal power plant generation equipment, which participates in standardized primary frequency control (SPFC), must meet specific requirements. These requirements are formalized as nine algorithmic criteria, which are used for automatic monitoring of power plant participation in SPFC. One of these criteria - primary frequency control droop and deadband actual values estimation is considered in detail in this report. Experience shows that existing estimation method sometimes doesn’t work properly. Author offers alternative method, which allows estimating droop and deadband actual values more accurately. This method was implemented as a software application.

Operation of thermal power plant generation equipment, which participates in standardized primary frequency control (SPFC), must meet specific requirements. These requirements are formalized as nine algorithmic criteria, which are used for automatic monitoring of power plant participation in SPFC [1]:  Failure to provide information;  Time step of data transmission doesn't meet the requirements;  Primary frequency control range doesn't meet the requirements;  Precision of data logging doesn't meet the requirements;  Automatic load frequency control system is in non-automatic mode;  Load control precision doesn't meet the requirements;  Primary frequency control droop and deadband actual values don't meet the requirements;  Lack of response to frequency change;  Oscillation. These criteria are presented in the document 'The procedure for determining the volume of SPCF services (Appendix 2. Criteria for monitoring of power plant participation in SPFC)' of Joint-stock Company 'System Operator of the United Power System'.
Primary frequency control droop and deadband actual values estimation is considered in detail in this report. Author offers a method, which allows estimating droop and deadband actual values accurately enough. Primary frequency control deadband is a specified value of the frequency deviation from the nominal value, which does not require primary control. Primary frequency control deadband minimal value is equal to actual deadband of automatic load frequency control system. Primary frequency control droop is a coefficient that determines changing of active power under the action of frequency control system at frequency changing.
At present droop and deadband actual values are estimated in monitoring using statistical methods of data processing. The droop and the deadband are determined as parameters of the frequency deviations Δ and active power deviations Δ regression function.
The use of the criterion for the operation of Karmanovskaya thermal power plant (TPP) unit 2 is shown in the figure 1. The droop and the deadband values are estimated ( = 6 %, = 0.024 Hz). But, as can be seen from the figure 1, the deadband estimate is not correct enough. Probably, this is due to the regression function smoothing and association to origin.
In this regard, consider an alternative method. Below are the steps in the alternative method algorithm.
From source variables , , , , where -frequency measurements array, -actual active power measurements array, -planned active power, move to vectors of relative variables , : • 100 Here -electrical power unit nominal power. A 'cloud' of points , is drawn up. The range from x min to x max is divided into N intervals. The width of the interval should be no more than 1.5 mHz and the number of points in the interval must ensure the accuracy of the calculation. From these conditions, the number of intervals is taken.
At each interval, the average value of у is calculated: where m -the number of points in the interval 1..N, j = 1…m. As a result we obtain two vectors х av , у av with N elements in an each vector. Vector у av contains of each interval, vector х av contains х values in the middle of each interval.
graph is constructed in conjunction with a 'cloud' of points and the so-called 'theoretical' graph based on the specified values of deadband and droop (in most cases for equipment, which participates in SPFC, = 5 %, = 0.02 Hz). Here: x 1 and y 1 are values from vectors , for intervals located at distances of 3%, 6%, 10% (for each droop value) of the range width from the range boundaries, x 2 and y 2 are values from vectors , for intervals located at a distance of 5% of the range width from the deadband boundaries. The values of 3%, 5%, 6%, 10% were obtained empirically.
Based on the estimated values of droop and deadband, a piecewise linear graph of the "estimated" static characteristic can be constructed. An example is shown in the figure 2. In the figure 2: green is the graph; red is the piecewise linear graph; s11, s21, s31, s12, s22, s32 are points located at a distances of 3%, 6%, 10% (for each droop value) of the range width from the range boundaries; s01, s02 are points located at a distance of 5% of the range width from the deadband boundaries; max1, max2 are maximum distances between the auxiliary line and the graph. Thus, the algorithm allows to estimate the deadband and the droop separately for negative and positive frequency deviations. Often these deviations are asymmetric with respect to zero.
The algorithm flow chart is shown in the figure 3. The partition of range from x min to x max into N intervals The calculation of the mean value M y at each interval.
The formation of variables х av , у av .
The My graph construction.
The auxiliary lines construction y aux =a 1 x+b 1 and y aux =a 2 x+b 2 Finding the maximum distances between the M y graph and the auxiliary lines for estimating the boundaries of the deadband The droop estimation End Figure 3. The algorithm flow chart.
The described algorithm was implemented as a software application.The arrays x and y calculated from the source data must be inserted into the two fields of the application window. After that, you must click the 'Get Arrays' button and specify the values of x min , x max and N. After clicking on the "Get Average" and "Inclined Line Method" buttons, the program calculates the values of the droop and the deadband and displays them on the screen. Using the "Graph" button, graphs are displayed.  The values of deadband and droop obtained by different methods are given in tables 1 and 2.  Thus, we can see that the alternative method gives more adequate and reliable estimation.

Conclusion
The existing method for primary frequency control droop and deadband actual values estimation sometimes doesn't work properly. The alternative method, which allows estimating droop and deadband actual values more accurately, has been developed.