If you don’t mind combining awk with column, then it’s really straight forward without needing to store any amount of data in arrays

pmset -g custom | 

awk '/^((AC|Battery) Power[:]$|[ t]+displaysleep[ t])/ && 
     (ORS = ORS < FS ? FS : RS)' | 

column -t

Output:

Battery  Power:  displaysleep  2
AC       Power:  displaysleep  5

From here, filtering AC Power vs. Battery Power is trivial.

Using awk:

function get_displaysleep_from_power () {
    pmset -g custom | awk -v heading="$1" '
    BEGIN {heading_found=0}
    {
        if(!heading_found){
            heading_found=match($1,"^"heading)
        }
        if (heading_found){
            if ($1 == "displaysleep") {
                print $2;
                exit
            }
        }
    }
    ' - 
}

get_displaysleep_from_power "AC"
get_displaysleep_from_power "Battery"

outputs:

5
2

I would harness GNU AWK for this task following way, let command output be

Battery Power:
 lidwake              1
 lowpowermode         1
 standbydelayhigh     86400
 standbydelaylow      10800
 standby              1
 proximitywake        0
 ttyskeepawake        1
 hibernatemode        3
 powernap             0
 gpuswitch            2
 hibernatefile        /var/vm/sleepimage
 highstandbythreshold 50
 displaysleep         2
 womp                 0
 networkoversleep     0
 sleep                1
 lessbright           1
 halfdim              1
 tcpkeepalive         1
 acwake               0
 disksleep            10
AC Power:
 lidwake              1
 lowpowermode         0
 standbydelayhigh     86400
 standbydelaylow      0
 standby              1
 proximitywake        1
 ttyskeepawake        1
 hibernatemode        3
 powernap             0
 gpuswitch            2
 hibernatefile        /var/vm/sleepimage
 highstandbythreshold 50
 displaysleep         5
 womp                 1
 networkoversleep     0
 sleep                1
 tcpkeepalive         1
 halfdim              1
 acwake               0
 disksleep            10

then

command | awk '/^[A-Z]/{head=$1;next}{arr[head][$1]=$2}END{print arr["AC"]["displaysleep"]}'

gives output

5

Explanation: for lines starting with uppercase letter I set head value to 1st field (Battery and AC for given data) and instruct GNU AWK to go next line that is not doing anything more. For all others lines I set value in 2D array under head-1st field key. In END I print value under desired key.

(tested in GNU Awk 5.1.0)

If you are looking for a robust way to create a hierarchical data structure from what you see output from pmset there are several ways:

1. Use awk in paragraph mode to easily catch the header and the time of interest;
2. Use a regex to parse all the output and build a hash from that;
3. Use the Apple tools to parse the binary plist file.

Here is a simple awk to get the value of displaysleep for all sections:

x=$(pmset -g custom | sed -E 's/([^[:blank:]][^:]*:)/nn1/') 
echo "$x" | awk 'BEGIN{RS=""; FS="n"} {
      for(i=1;i<=NF;i++) if ($i~"^ displaysleep ") 
          print $1, $i
}'  

Prints:

Battery Power:  displaysleep         2
AC Power:  displaysleep         10

The problem with parsing the output from pmset -g custom is that the spaces may be ambiguous. Examples:

• Both keys and values can have spaces and one or more spaces are used to delimit between key and value.

• The value for the key "hibernatefile" is potentially a file with a space in it.

• The key "sleep" not only reports status but may report the reason the actual state is different than assigned. If I enter pmset -g | grep ' sleep ' the return is sleep 0 (sleep prevented by bluetoothd, coreaudiod, powerd, sharingd)

With those provisos, you can parse the output of pmset with the knowledge that some values for certain keys cannot be reliably parsed without further knowledge of that specific key.

Here is a Ruby that will parse the output of pmset -g custom into a Ruby hash. That hash can then be queried as a two level hash (["Battery Power"]["powermode"]) or turned into a structure like JSON to query in another way:

pmset -g custom | ruby -r json -e '
    dat=$<.read.scan(/^(S+s+Power:[sS]+?(?=(?:^S+s+Power:$)|z))/).
    map{|a| sa=a[0].split(/:s*R[ t]+/,2); [sa[0], sa[1].split(/R[  t]+/)]}.
    map{|sa| 
        [sa[0], sa[1].flatten.map{|s| s.scan(/^([^n]+?)s+(S+$)/)}.
        flatten(1).to_h] }.
    to_h

#output that hash
puts "pmset -g custom as hash:n#{dat}nn"

puts ""Battery Power"."powermode": #{dat["Battery Power"]["powermode"]}nn"

puts "pmset -g custom as JSON:n#{JSON.pretty_generate(dat)}"'

Prints:

pmset -g custom as hash:
{"Battery Power"=>{"Sleep On Power Button"=>"1", "powermode"=>"0", "standby"=>"1", "ttyskeepawake"=>"1", "hibernatemode"=>"3", "powernap"=>"1", "hibernatefile"=>"/var/vm/sleepimage", "displaysleep"=>"2", "womp"=>"0", "networkoversleep"=>"0", "sleep"=>"1", "lessbright"=>"1", "tcpkeepalive"=>"1", "disksleep"=>"0"}, "AC Power"=>{"Sleep On Power Button"=>"1", "powermode"=>"0", "standby"=>"1", "ttyskeepawake"=>"1", "hibernatemode"=>"3", "powernap"=>"1", "hibernatefile"=>"/var/vm/sleepimage", "displaysleep"=>"10", "womp"=>"1", "networkoversleep"=>"0", "sleep"=>"0", "tcpkeepalive"=>"1", "disksleep"=>"0"}}

"Battery Power"."powermode": 0

pmset -g custom as JSON:
{
  "Battery Power": {
    "Sleep On Power Button": "1",
    "powermode": "0",
    "standby": "1",
    "ttyskeepawake": "1",
    "hibernatemode": "3",
    "powernap": "1",
    "hibernatefile": "/var/vm/sleepimage",
    "displaysleep": "2",
    "womp": "0",
    "networkoversleep": "0",
    "sleep": "1",
    "lessbright": "1",
    "tcpkeepalive": "1",
    "disksleep": "0"
  },
  "AC Power": {
    "Sleep On Power Button": "1",
    "powermode": "0",
    "standby": "1",
    "ttyskeepawake": "1",
    "hibernatemode": "3",
    "powernap": "1",
    "hibernatefile": "/var/vm/sleepimage",
    "displaysleep": "10",
    "womp": "1",
    "networkoversleep": "0",
    "sleep": "0",
    "tcpkeepalive": "1",
    "disksleep": "0"
  }
}

Alternatively, you can do it the proper Apple Way which is to reference the actual plist file.

Here is a Ruby that parses the plist file into XML:

ruby -e '

# Depending on the version of OS X, the plist can be at 
# /Library/Preferences/SystemConfiguration/com.apple.PowerManagement.plist
info_plist=File.read("/Library/Preferences/com.apple.PowerManagement.plist")

puts IO.popen("plutil -convert xml1 -r -o - -- -", "r+") {|f|
  f.write(info_plist)
  f.close_write
  info_plist = f.read # xml plist
}'

Prints:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <key>AC Power</key>
    <dict>
        <key>DarkWakeBackgroundTasks</key>
        <integer>1</integer>
        <key>Disk Sleep Timer</key>
        <integer>0</integer>
        <key>Display Sleep Timer</key>
        <integer>10</integer>
        <key>System Sleep Timer</key>
        <integer>0</integer>
        <key>Wake On LAN</key>
        <integer>1</integer>
    </dict>
    <key>Battery Power</key>
    <dict>
        <key>DarkWakeBackgroundTasks</key>
        <integer>1</integer>
        <key>Disk Sleep Timer</key>
        <integer>0</integer>
        <key>Display Sleep Timer</key>
        <integer>2</integer>
        <key>ReduceBrightness</key>
        <integer>1</integer>
        <key>System Sleep Timer</key>
        <integer>1</integer>
        <key>Wake On LAN</key>
        <integer>0</integer>
    </dict>
    <key>SystemPowerSettings</key>
    <dict>
        <key>Update DarkWakeBG Setting</key>
        <true/>
    </dict>
</dict>
</plist>

You can then query that XML file.

This might work for you (GNU sed and bash):

 f(){ p=${1:-Battery};
      pmset -g custom | sed -En '/:$/h;G;s/ (displaysleep.*)n('$p'.*)/2 1/p'; }

Create a shell function that sends the output of the pmset command through sed and outputs by default Battery Power or AC Power if supplied with the parameter AC.

I wouldn’t normally answer a question that doesn’t contain an attempt by the OP to solve their own problem but since there are multiple answers already:

Using any awk:

$ pmset -g custom | awk -v pwr='Battery' -v tag='displaysleep' 'index($0,pwr" ")==1{p=1} p && $1==tag{print $2; exit}'
2

Assuming the headings don’t appear anywhere else and aren’t repeated, your question amounts to:

• skip over stuff until
• either of the headings is matched
• skip over more stuff until
• displaysleep is matched

And doing that twice