node feature

[cshjin]

Provide a detailed process to extract the node features.

Keys associated with nodes

• bus
• gen
• gmd_bus
• bus_gmd

[cshjin]

bus node features

selected feature	column	type	description
	bus_i	int	bus number
	type	int	bus type (1-4)
	Pd	float	real power demand
	Qd	float	reactive power demand
	Gs	float	shunt conductance
	Bs	float	shunt susceptance
x	area	int	area number
	Vm	float	voltage magnitude
	Va	float	voltage angle
	baseKV	float	base voltage
x	zone	int	loss zone
	Vmax	float	maximum voltage magnitude
	Vmin	float	minimum voltage magnitude

• bus_i is the bus number, but not bus id.
  • [x] DO NOT include the bus_i in the node feature
  • [ ] Match the bus_i with bus index, probably from the bus_sourceID in mpc file.
• type has four categories
  • [x] change to one-hot encoder
  • [x] remove the type from column, but include four different categories as new features.
• Overall, bus has 15 features

area, zone,

[cshjin]

gen node features

column	type	description
bus	int	bus number
Pg	float	real power output
Qg	float	reactive power output
Qmax	float	maximum reactive power output
Qmin	float	minimum reactive power output
Vg	float	voltage magnitude setpoint
mBase	float	total MVA base of machine
status	float	status, > 0, in-service, <= 0 out-of-service
Pmax	float	maximum real power output
Pmin	float	minimum real power output
Pc1	float	lower real power output of PQ capability curve
Pc2	float	upper real power output of PQ capability curve
Qc1min	float	minimum reactive power output at PC1
Qc1max	float	maximum reactive power output at PC1
Qc2min	float	minimum reactive power output at PC2
Qc2max	float	maximum reactive power output at PC2
ramp_agc	float	ramp rate for load following/AGC
ramp_10	float	ramp rate for 10 minute reserves
ramp_30	float	ramp rate for 30 minute reserves
ramp_q	float	ramp rate for reactive power (2 sec timescale)
apf	float	area participation factor

• bus_i is matched with bus_i in bus data
  • [x] remove it from features
  • [ ] (TODO) decide what if the node has no generator, fill the empty features with SOME values.
• status is in float (int)
  • [x] (optional) convert to one hot encoder
• Overall, gen has 20 feature

[cshjin]

gmd_bus node features

columns	dtype	description
parent_index	int	index of corresponding ac network bus
status	binary	binary value that defines the status of bus (1: enabled, 0: disabled)
g_gnd	float	admittance to ground (in unit of Siemens)
name	string	a descriptive name for the bus

• parent_index is matched with bus_i
• name is in string format, seems have two type of features
  • [x] dc_sub1 has g_gnd value, split into two columns -- ac bus
  • [ ] dc_bus1 has g_gnd value, split into two columns
  • [x] (TODO) Is this true?
  • [x] filling value for missing buses
  • No need to fill missing bus
• Overall, gmd_bus has 4 features

[cshjin]

bus_gmd pos attr

columns	dtype	description
lat	float	latitude coordinate of ac network bus and corresponding dc network bus
lon	float	longitude coordinate of ac network bus and corresponding dc network bus

• bus_gmd is geo information, should be in the pos data.
• NO normalized required.

[cshjin]

• [x] figure out the index with missing values in epri21.m. Buses are indexed from 1 to 21, but missing 3 nodes in the bus table
  • [x] try the index_col to specify the first column as indices and match it with the values in the branch table.

Solved in #10

[cshjin]

Close for now

(option) select relevant features for problem formulation