What Affects ORP?

Jump to latest Follow Reply
amattas

amattas

Gold Supporter
Bronze Supporter
Jun 18, 2017
110
Cincinnati, OH
I know in general ORP + SWG is not a great combination, but I have the Hayward sense and dispense system on my pool so I want to make it more meaningful.

I've been told TC, FC, CYA, sunlight, etc all impact the readings, are there other things?

Ultimately I'm planning to measure these factor at common times through the day over the period of a month, and build a machine learning/AI model that will take some of them into account, to "predict" what FC is based on the ORP reading and other observations.

I have no idea if this will work or not - it's an experiment - but I'd love to hear everyone's thoughts.
 
amattas said:
Unfortunately it's not a LLM, so that doesn't work - it's a cluster wise regression model.
Click to expand...
Then you will need to read the threads and assimilate it the old-fashioned way.
 
amattas said:
I see we're back to not wanting to discuss new ideas or help people over here.
Click to expand...
@amattas I don't think anyone is not wanting to discuss new ideas. AI and ORP are waaaay beyond the average TFP member. As I read the thread, not knowing much, @ajw22 suggested how to use your AI (assuming it was an LLM), but since you are not using that, you must use a different mechanism (read and feed) to satisfy your cluster wise regression model. Don't know if that is right or wrong, but I read it as suggestions to help you...

Not sure there are many here that can help you with your endeavor. Happy to discuss, but without the expertise to help, might be few that can actually discuss (or help).

Love to hear the results of your experiment!
 
PoolStored said:
@amattas I don't think anyone is not wanting to discuss new ideas. AI and ORP are waaaay beyond the average TFP member. As I read the thread, not knowing much, @ajw22 suggested how to use your AI (assuming it was an LLM), but since you are not using that, you must use a different mechanism (read and feed) to satisfy your cluster wise regression model. Don't know if that is right or wrong, but I read it as suggestions to help you...

Not sure there are many here that can help you with your endeavor. Happy to discuss, but without the expertise to help, might be few that can actually discuss (or help).

Love to hear the results of your experiment!
Click to expand...
That's fair - I don't expect anyone to be able to help with the computer science parts and I have read a lot of the ORP threads, but many of them either have a shortage of information and sometimes misinformation so I was hoping someone could give me a concise list based on testing they had done on what actually affects the readings and what doesn't - to avoid having to dive in to a bunch of chemistry texts, because while the CS part is easy for me - I am terrible at actual chemistry, and was hoping someone here had the experience.
 
Gemini:

You're right, relying solely on ORP with a Saltwater Chlorine Generator (SWG) can be tricky. The relationship between ORP and Free Chlorine (FC) in a swimming pool is complex and influenced by many factors. Your plan to build a machine learning model to predict FC based on ORP and other parameters is a smart approach to make your Hayward Sense and Dispense system more useful.
Beyond Total Chlorine (TC), Free Chlorine (FC), Cyanuric Acid (CYA), and sunlight, here are other factors that can significantly impact ORP readings in your pool, especially with an SWG:
Water Chemistry Parameters:
  • pH: This is arguably the most significant factor affecting ORP. ORP readings are highly sensitive to pH changes. A higher pH will generally lead to lower ORP readings for the same level of FC. Conversely, a lower pH will result in higher ORP readings. Your model will absolutely need to include pH measurements.
  • Total Alkalinity (TA): While not as direct as pH, TA influences pH stability. Fluctuations in TA can indirectly affect ORP by causing pH swings.
  • Calcium Hardness (CH): Extremely low or high calcium hardness can affect the overall water balance and potentially influence the efficiency of chlorine generation by the SWG, indirectly impacting ORP.
  • Other Oxidizers/Reducers:Any other chemicals present in the water that can either oxidize or reduce substances will directly affect the ORP reading. This includes:
    • Shock Treatments (Non-Chlorine): Potassium monopersulfate (MPS) shock will significantly increase ORP readings, even though it's not FC.
    • Algaecides: Some algaecides can have oxidizing or reducing properties.
    • Metal Ions: Dissolved metals like iron, copper, or manganese can affect ORP readings.
  • Total Dissolved Solids (TDS): High TDS levels, especially from the salt in your SWG, can influence the conductivity of the water and potentially affect the ORP sensor's performance and readings.
  • Temperature: Temperature can affect the kinetics of chemical reactions in the water, including the oxidation-reduction reactions that ORP measures. While the effect might be less dramatic than pH, it's worth considering, especially with seasonal temperature changes.
SWG Specific Factors:
  • Salt Level: The efficiency of your SWG in generating chlorine is directly dependent on the salt concentration. Insufficient salt levels will lead to lower chlorine production and thus lower ORP readings.
  • SWG Output Setting: The percentage or level at which your SWG is set to run directly dictates the amount of chlorine being generated, which will have a significant impact on ORP.
  • SWG Cell Condition: Over time, the SWG cell can develop scale buildup or wear out, reducing its efficiency in producing chlorine. This will lead to lower FC and consequently lower ORP readings for the same output setting.
  • Flow Rate: The rate at which water passes through the SWG cell can influence chlorine production efficiency to some extent.
Environmental and Usage Factors:
  • Bather Load: A higher number of swimmers introduces more organic matter (sweat, oils, etc.), which consumes chlorine and can affect ORP.
  • Rainfall: Significant rainfall can dilute the pool water, potentially affecting salt levels and other chemical parameters, indirectly influencing ORP. It can also introduce contaminants that consume chlorine.
  • Debris: Leaves, dirt, and other organic debris in the pool will consume chlorine and affect ORP.
  • Filter Condition and Circulation: Proper filtration and circulation are crucial for evenly distributing chlorine throughout the pool. Poor circulation can lead to localized variations in chlorine levels and ORP.
Sensor Specific Factors:
  • Sensor Calibration: An improperly calibrated ORP sensor will provide inaccurate readings. Regular calibration is essential.
  • Sensor Condition: Over time, the ORP sensor can become fouled or degraded, affecting its accuracy. Regular cleaning and eventual replacement are necessary.
  • Electrode Type: Different ORP sensors may have slightly different response characteristics.
Considerations for Your Machine Learning Model:
  • Data Collection Frequency and Timing: Consistent data collection times throughout the day (e.g., morning, midday, evening) will help capture diurnal variations.
  • Data Logging Accuracy: Ensure your measurement devices for TC, FC, CYA, pH, etc., are accurate and well-maintained.
  • Feature Engineering: You might need to create new features from your raw data. For example, instead of just sunlight intensity, you could categorize it (low, medium, high). You could also consider the change in certain parameters over time.
  • Model Selection: Experiment with different machine learning algorithms to find the one that best captures the complex relationships in your data (e.g., regression models, neural networks, etc.).
  • Validation: Thoroughly validate your model with unseen data to ensure its predictive accuracy.
  • Maintenance: Remember that the underlying relationships might change over time (e.g., as your SWG cell ages), so you may need to retrain your model periodically.
By diligently collecting data on these various factors and incorporating them into your machine learning model, you have a good chance of developing a more meaningful way to interpret the ORP readings from your Hayward Sense and Dispense system and get a better estimate of your Free Chlorine levels. Good luck with your project!
 
  • Like
Reactions: PoolStored
amattas said:
That's fair - I don't expect anyone to be able to help with the computer science parts and I have read a lot of the ORP threads, but many of them either have a shortage of information and sometimes misinformation so I was hoping someone could give me a concise list based on testing they had done on what actually affects the readings and what doesn't - to avoid having to dive in to a bunch of chemistry texts, because while the CS part is easy for me - I am terrible at actual chemistry, and was hoping someone here had the experience.
Click to expand...
Using the Search bar and reading threads would let you identify TFP Experts like @mgtfp who have posted about ORP.

You can review his postings here - Search results for query: orp

Much information is available here for those who take the initiative to find it.

The mission of this site, aptly named Trouble Free Pool, is to help average pool owners maintain their pools in the most trouble free least effort ways.

There is the Deep End section for pondering the deep mysteries of pool chemistry. Plunge into the more technical aspects of pool care. Discussions there may not be applicable to everyday pool care.
 
Last edited:
  • Like
Reactions: Newdude and PoolStored
Claude:

You're tackling an interesting project with your Hayward Sense and Dispense system. ORP readings can indeed be affected by multiple factors, making direct correlation with FC challenging.

Here are the key factors that impact ORP readings:
  1. Free Chlorine (FC) - Primary relationship you're investigating
  2. Cyanuric Acid (CYA) - Higher CYA levels reduce the effectiveness of chlorine and lower ORP
  3. pH - Higher pH reduces chlorine effectiveness and lowers ORP
  4. Temperature - Warmer water typically has lower ORP
  5. Total Dissolved Solids (TDS) - Higher TDS can lower ORP readings
  6. Combined Chlorine (CC) - Chloramines can affect readings
  7. Sunlight exposure - UV degradation of chlorine
  8. Bather load - Organic contamination
  9. Calibration status of probes - Affects accuracy
  10. Water flow rate past the probe - Can affect readings
  11. Salt levels (if using SWG) - Impacts conductivity
For your ML model, I'd recommend collecting:
  • Time of day
  • ORP reading
  • FC level (measured with a reliable test kit)
  • pH
  • Temperature
  • CYA level
  • Weather conditions (sunny, cloudy, etc.)
  • Recent bather load
  • Time since last filter backwash/cleaning
  • SWG output percentage
This is an excellent application for machine learning since the relationships between these variables are complex and often non-linear. By collecting data over a month, you should get enough variation to build a reasonably accurate model.

I'd recommend a few ML approaches for your pool ORP prediction model:

  1. Random Forest Regression - Great for this problem because:
    • Handles non-linear relationships well
    • Manages both numerical and categorical variables (like weather conditions)
    • Provides feature importance metrics to see which factors most affect ORP
    • Resistant to overfitting with smaller datasets
  2. Gradient Boosting Models (like XGBoost or LightGBM):
    • Often achieve higher accuracy than random forests
    • Excellent for time-series data patterns
    • Handle complex interactions between variables
    • Built-in regularization to prevent overfitting
  3. Neural Networks (simple architecture):
    • If you collect enough data (~100+ observations)
    • Can capture complex non-linear relationships
    • Consider a basic 2-3 layer network given your dataset size
  4. Multiple Linear Regression with Interactions:
    • Start here for baseline performance
    • Include interaction terms (like FC×CYA, pH×Temp)
    • Simple to interpret
For implementation, I suggest:
  • Use Python with scikit-learn, XGBoost, or TensorFlow
  • Split data 80/20 for training/testing
  • Use cross-validation to avoid overfitting
  • Try feature engineering (e.g., creating variables like "hours since sunrise")
  • Normalize your numerical variables

For evaluation, focus on RMSE (root mean squared error) and R² to assess how well your model predicts FC from ORP readings.
 

Enjoying this content?

Support TFP with a donation.

Give Support
PoolStored said:
Happy to discuss, but without the expertise to help, might be few that can actually discuss (or help).
Click to expand...
+1. ORP usually brings people to TFP. At some point they give up, or go back to ignoring its limitations.

Nobody to date has cracked the code, but that doesn't mean it can't be done. If it was (1st) repeatable and (2nd) Trouble Free, we would happily add it to the rotation of reliable options.

Maybe you can be the one @amattas :)

:epds:
 
CHATGPT:

That's a super interesting and ambitious project—especially with the Sense & Dispense system already giving you ORP data to work from.

You're right that ORP alone isn't super reliable in saltwater pools with a SWG, but modeling it with contextual data could give you a serious edge.

You’ve already nailed a bunch of the big influencers—TC, FC, CYA, sunlight (UV exposure). Here are some additional factors that can impact ORP readings and FC behavior that you might want to include in your data collection:

Additional Influential Factors:​

  1. pH – ORP is very sensitive to pH; even small pH changes can significantly shift ORP without any actual change in FC.
  2. Water Temperature – Affects reaction rates and the strength of oxidation; ORP sensors themselves can be affected by temperature.
  3. Salt Level (ppm) – SWG performance depends on proper salt levels; low salt can cause weaker chlorine production.
  4. Phosphates/Nitrates – Not directly affecting ORP, but they consume chlorine, which changes FC/TC balance and can alter ORP indirectly.
  5. Bather Load – Organic contaminants and ammonia from swimmers impact chlorine demand and ORP readings.
  6. Alkalinity – Can influence pH stability, which in turn affects ORP.
  7. Time of Day / Sun Angle – Related to UV but useful as a separate timestamp or feature for your model.
  8. Circulation Timing – ORP readings may vary depending on whether you're mid-circulation or at rest periods.
  9. Sensor Drift / Calibration Age – ORP sensors degrade over time, so tracking calibration intervals or age could help normalize long-term trends.
If you log all that over time and feed it into a model like a random forest or even something simple like linear regression at first, you might end up with a surprisingly accurate FC predictor from ORP and environmental data.

Are you planning to gather the data manually, or are you going to automate any of it with a Raspberry Pi or similar setup?
 
It just seems like a lot of trouble for an indirect control method. And it’s counter productive to loose FC to UV due to low CyA just to run a control method that is not as simple, precise or cost effective as testing and adjusting FC and CyA with a decent test kit.

You can try using the ORP control as the SWG’s set point control and adjust the set point in response to regular testing the same as adjusting a set point and runtime but automated pH control would be essential and your back to beginning every time an adjustment is made. Rain can have a significant impact while topping up with tap water will cause slow creep of the set point.

ORP is one or those systems that falls into the category of “seemed like a good idea at the time but”, and they are sold as a premium product. But once you start to understand what is required to effectively manage them they no longer look the same as they did in the glossy brochure. And to be fair they are not well matched to the TFP method where a high CyA level is typically recommended.

I can’t sum it up better so I’ve just added the quote and the link. There is a broken link to the “ORP” thread which contained loads of data but could end find it.

If you looked at the ORP post and the links to manufacturer's own tables of ORP vs. FC in various conditions, it's all over the map. That means using ORP as a reliable measure of disinfection isn't very useful except on a broad scale. The ORP data is all over the place. Also, as for the Oakton meter, it was used in a study of 194 pools taking 620 samples as shown in this post. Note how the ORP is much more highly correlated to the calculated hypochorous acid concentration (which is roughly half the FC/CYA ratio) than it is to FC alone. This is, of course, expected, yet the 2009 paper you referenced didn't even bother doing the calculations for HOCl. Also note that in the pool study using the Oakton meter, many pools also had built-in ORP systems yet in 23% of those measuring the same water they showed 100 mV or more differences, some higher some lower. This wasn't just an issue of calibration either -- the Oakton was regularly calibrated for each day's measurements.

ORP measures a thermodynamic quantity, not reaction rates which is really what is relevant, and even then it doesn't actually follow the Nernst equation for chlorine and no one has explained why. The closest fit implies a chemical reaction with 0.6 to 0.8 electrons instead of the expected 2-electron reaction. Also, many chemicals including dissolved hydrogen gas from saltwater chlorine generators or the use of non-chlorine shock (potassium monopersulfate) can all affect ORP readings but do not result in the corresponding implied changes in disinfection rates. At best, ORP can be used for process control, as a setpoint, when accurately measuring FC, CYA and other water chemistry parameters.

Why spend hundreds of dollars on an ORP sensor when given the FC and CYA at normal pH near 7.5 you will know the active chlorine level? All you have to do is choose your FC/CYA ratio for your purposes. For those on this and other pool forums, the goal is to have enough chlorine to kill green and black algae faster than they can grow. In manually dosed pools, this is roughly an FC that is 7.5% of the CYA level (in saltwater chlorine generator pools it's roughly an FC that is 5% of the CYA level). If you were wondering what the kill times are like, this post lists them for an FC that is around 10% of the CYA level -- the equivalent of 0.1 ppm FC with no CYA. For commercial/public pools, a decent minimum would be no lower than this and might be an FC that is 20% of the CYA level so equivalent to 0.2 ppm FC with no CYA. This balances disinfection and oxidation times against minimizing disinfection by-products and oxidation of swimsuits, skin and hair and of corrosion rates.
Click to expand...

Pool Water Chemistry

Nearly everything is still true at CYA levels over 100, but some of the secondary effects start getting to be major problems when CYA is that high, making it difficult to take care of the pool properly. For example, the FC levels required start to cause the PH test to not work. The amounts of...
www.troublefreepool.com www.troublefreepool.com
 
AUSpool said:
It just seems like a lot of trouble for an indirect control method. And it’s counter productive to loose FC to UV due to low CyA just to run a control method that is not as simple, precise or cost effective as testing and adjusting FC and CyA with a decent test kit.

You can try using the ORP control as the SWG’s set point control and adjust the set point in response to regular testing the same as adjusting a set point and runtime but automated pH control would be essential and your back to beginning every time an adjustment is made. Rain can have a significant impact while topping up with tap water will cause slow creep of the set point.
Click to expand...
Typically I would agree, but we sometimes aren't in a position to test chemicals for a couple weeks at a time in the summer. The onboard ORP control is terrible - it always overshoots and is late to respond. Every attempt at using it either resulted in too high of chlorine levels, or too low (pool would get cloudy).

That's why I want to approach it from this angle, if I can ballpark what FC actually is based on environmental data (temperature, UV, bather load from my cameras, PH from the PH sensor, CYA/Metals/Calcium/etc from the last test I actually did from the PoolMath API, output based on salt levels, etc), I suspect I can get closer by taking the derivation of that ballpark and adjusting the output earlier instead of either having to set a lower ORP value or having it adjust too late.

Someone made a local control plugin for Home Assistant for Hayward, so if I can find something that mostly works I can implement it in that plugin so anyone using it can take advantage of it (making it trouble free at that point for those users).
 
amattas said:
That's why I want to approach it from this angle, if I can ballpark what FC actually is based on environmental data (temperature, UV, bather load from my cameras, PH from the PH sensor, CYA/Metals/Calcium/etc from the last test I actually did from the PoolMath API, output based on salt levels, etc), I suspect I can get closer by taking the derivation of that ballpark and adjusting the output earlier instead of either having to set a lower ORP value or having it adjust too late.
Click to expand...

Give me half of that data, and I can adjust my SWG % using an app from anywhere in the world. No machine learning/AI mode is necessary.

My pool and others do fine running with the SWG unattended for 30 days or longer with some remote monitoring.
 
Coming from someone who manages commercial pools successfully with ORP automation, I’ll tell you PH suppression to maintain as stable a PH as possible is most important. Keeping CYA around 30 to prevent interfence is secondary. Temperature stability is helpful. Regular maintenance and calibration of the probes to a known good test kit is paramount.

ORP shines in commercial settings where bather loads can swing from 0 at opening, to 200 bathers at noon, and back down to 30 at close where testing and manual dosing multiple times a day isn’t practical. For homeowners who want less variables to deal with due to time constraints….adding more mechanical devices that need maintenance, calibration, cleaning, etc seems counterproductive and is usually the root cause of the failure of system. These systems add a higher threshold of hands on commitment to pool maintenance to keep them operating properly. They are a useful tool in the correct application, but will never be a substitute for regular testing with a reliable test kit. In fact these systems require that for their calibration and “seamless” operation.

Good luck with your project. Following with interest.
 
  • Like
Reactions: AUSpool, JamesW and PoolStored
Katodude said:
Or, you could just buy a FC sensor probe.

sensorex.com

FCL Amperometric Free Chlorine Sensor | Sensorex

Our dependable free chlorine sensor features amperometric measurement technology. Designed for use in disinfection applications with direct 4-20mA output.
sensorex.com sensorex.com
Click to expand...
lol at a meager $1,700 per probe everyone should have one. Of course they are considered consumable items, so add that to the annual replacement list. I wonder if it’s a direct fit for the sense and dispense or if a rig will need to be built to interpolate the data. Add another 2 grand. I’m in….🤣
 
  • Like
Reactions: JamesW

Enjoying this content?

Support TFP with a donation.

Give Support
You must log in or register to reply here.
Top Bottom
Back
Jump to latest Follow Reply