MQ-4 sensore metano

Quella che hai fatto è una mappatura lineare.
Dai datasheet visti in rete (non so quale sia il tuo) il sensore sembra logaritmico. Inoltre dovrai compensare le variazioni dovute a temperatura ed umidità che mi sembrano consistenti.
Lo schema della curva mostra che a 1000ppm le due resistenze (quella del sensore e quella imposta dal circuto) sono equivalenti, quindi, se non dico castronerie, il valore centrale è 512. (pari a 1000ppm)

EDIT:
Nel playground ho trovato una funzione per la mappatura logaritmica.
--> Arduino Playground - Fscale

/* fscale
 Floating Point Autoscale Function V0.1
 Paul Badger 2007
 Modified from code by Greg Shakar

 This function will scale one set of floating point numbers (range) to another set of floating point numbers (range)
 It has a "curve" parameter so that it can be made to favor either the end of the output. (Logarithmic mapping)

 It takes 6 parameters

 originalMin - the minimum value of the original range - this MUST be less than origninalMax
 originalMax - the maximum value of the original range - this MUST be greater than orginalMin

 newBegin - the end of the new range which maps to orginalMin - it can be smaller, or larger, than newEnd, to facilitate inverting the ranges
 newEnd - the end of the new range which maps to originalMax  - it can be larger, or smaller, than newBegin, to facilitate inverting the ranges

 inputValue - the variable for input that will mapped to the given ranges, this variable is constrained to originaMin <= inputValue <= originalMax
 curve - curve is the curve which can be made to favor either end of the output scale in the mapping. Parameters are from -10 to 10 with 0 being
          a linear mapping (which basically takes curve out of the equation)

 To understand the curve parameter do something like this: 

 void loop(){
  for ( j=0; j < 200; j++){
    scaledResult = fscale( 0, 200, 0, 200, j, -1.5);

    Serial.print(j, DEC);  
    Serial.print("    ");  
    Serial.println(scaledResult, DEC); 
  }  
}

And try some different values for the curve function - remember 0 is a neutral, linear mapping

To understand the inverting ranges, do something like this:

 void loop(){
  for ( j=0; j < 200; j++){
    scaledResult = fscale( 0, 200, 200, 0, j, 0);

    //  Serial.print lines as above

  }  
}

*/


#include <math.h>

int j;
float scaledResult; 

void setup() {
  Serial.begin(9600);
}

void loop(){
  for ( j=0; j < 200; j++){
    scaledResult = fscale( 0, 200, 0, 200, j, -1.5);

    Serial.print(j, DEC);  
    Serial.print("    ");  
    Serial.println(scaledResult , DEC); 
  }  
}

float fscale( float originalMin, float originalMax, float newBegin, float
newEnd, float inputValue, float curve){

  float OriginalRange = 0;
  float NewRange = 0;
  float zeroRefCurVal = 0;
  float normalizedCurVal = 0;
  float rangedValue = 0;
  boolean invFlag = 0;


  // condition curve parameter
  // limit range

  if (curve > 10) curve = 10;
  if (curve < -10) curve = -10;

  curve = (curve * -.1) ; // - invert and scale - this seems more intuitive - postive numbers give more weight to high end on output 
  curve = pow(10, curve); // convert linear scale into lograthimic exponent for other pow function

  /*
   Serial.println(curve * 100, DEC);   // multply by 100 to preserve resolution  
   Serial.println(); 
   */

  // Check for out of range inputValues
  if (inputValue < originalMin) {
    inputValue = originalMin;
  }
  if (inputValue > originalMax) {
    inputValue = originalMax;
  }

  // Zero Refference the values
  OriginalRange = originalMax - originalMin;

  if (newEnd > newBegin){ 
    NewRange = newEnd - newBegin;
  }
  else
  {
    NewRange = newBegin - newEnd; 
    invFlag = 1;
  }

  zeroRefCurVal = inputValue - originalMin;
  normalizedCurVal  =  zeroRefCurVal / OriginalRange;   // normalize to 0 - 1 float

  /*
  Serial.print(OriginalRange, DEC);  
   Serial.print("   ");  
   Serial.print(NewRange, DEC);  
   Serial.print("   ");  
   Serial.println(zeroRefCurVal, DEC);  
   Serial.println();  
   */

  // Check for originalMin > originalMax  - the math for all other cases i.e. negative numbers seems to work out fine 
  if (originalMin > originalMax ) {
    return 0;
  }

  if (invFlag == 0){
    rangedValue =  (pow(normalizedCurVal, curve) * NewRange) + newBegin;

  }
  else     // invert the ranges
  {   
    rangedValue =  newBegin - (pow(normalizedCurVal, curve) * NewRange); 
  }

  return rangedValue;
}

quindi, inserendo la funzione nel tuo codice e l'include della libreria math, dovresti avere:

valoreReale =fscale(0, 1023, 200, 10000, valoresensore, 10);

Devi controllare se il valore centrale è rispettato. 512 = 1000ppm