Mathficast Software Services Limited

Mathfi-Environment ©
Classifier/Predictive AI

A new AI-powered service helping environmental scientists and experts to make better informed decisions whenever a binary or multi-class prediction or classification is required. For example:

1. Predicting wind turbines’ output power levels under different conditions: Wind turbines as a source of renewable energy play a crucial role in battle against global warming and having a greener and cleaner environment. Butterfly AI can help designers and wind turbine technology firms accurately predict the output power of the wind turbines under different conditions to optimise the process. For example, you may divide the output power range of wind turbine (if it is between 0KW and 2000KW) into ten buckets/classes of (Class 1 between 0 and 200KW, …, Class 10 between 1800KW and 2000KW) and then predict under certain conditions into which bucket, a future unseen prediction sample will fall in.

Having measured the output power and established the data features of the selected wind turbines during thousands of past time snapshots and under different conditions including wind turbine ID, absolute wind direction, outdoor temperature, min outdoor temperature, max outdoor temperature, std of outdoor temperature, grid frequency, min grid frequency, max grid frequency, std of grid frequency, grid voltage, grid voltage min, grid voltage max, grid voltage std, rotor speed, rotor speed min , rotor speed max, rotor speed std, rotor bearing temperature, min of rotor bearing temperature, max of rotor bearing temperature, std of rotor bearing temperature, absolute wind direction, nacelle angle, pitch angle, min pitch angle, max pitch angle, std of the pitch angle, hub temperature, min hub temperature, max hub temperature, std of hub temperature, generator converter speed, min generator converter speed, max generator converter speed, std of hub temperature, generator speed, min generator speed, max generator speed, std of generator speed, generator bearings No. 1 to N temperature, min generator bearings No. 1 to N temperature, max generator bearings No. 1 to N temperature, std of generator bearings No. 1 to N temperature, generator stator temperature, min generator stator temperature, max generator stator temperature, std of generator stator temperature, gearbox bearings No. 1 to N temperature, min gearbox bearings No. 1 to N temperature, max gearbox bearings No. 1 to N temperature, std of gearbox bearings No. 1 to N temperature, gearbox inlet temperature, min gearbox inlet temperature, max gearbox inlet temperature, std of gearbox inlet temperature, gearbox oil sump temperature, min gearbox oil sump temperature, max gearbox oil sump temperature, std of gearbox oil sump temperature, nacelle angle, min nacelle angle, max nacelle angle, std of nacelle angle, nacelle temperature, min nacelle temperature, max nacelle temperature, std of nacelle temperature, and having established to which power bucket/class each one of those past measurement slots (i.e. training samples) belong and labelled them as Classes 1 to 10, predict and establish what level of power a wind turbine will generate under a future scenario (i.e. an unseen prediction sample).

If you wish to predict the output power as continuous numbers rather than categories or classes, you may predict the winning bucket as winning class of the prediction process and then divide that bucket into further M more granular sub-buckets and perform further rounds of training and prediction across the range of the winning bucket to get the winning sub-bucket and take the average value of that sub-bucket as the final continuous output value of the prediction. Please see the overview page for more details regarding the prediction with continuous numbers.

2. Site Energy Use Intensity (EUI) rating prediction: Energy efficient buildings are an essential part of an environmentally friendly and green ecosystem. EUI is the total amount of energy used in a building in a year divided by its floor area which is an indicator of the energy efficiency of a building's design and/or operations. Butterfly AI can help predict the EUI ranking of a building so that building designers have an idea about a site EUI before designing and building it. Knowing the past data features of hundreds of buildings including building site, floor area, orientation, location. altitude, window sizes, building envelope, ventilations, insulation, sociodemographic census data such as population and traffic patterns to building, number of workers or residents, building category (Residential, Non-residential), category of non-residential building (e.g., insurance company, bank, major corporate, etc.), built year, land use, surrounding buildings' types, surrounding buildings' patterns, surrounding buildings’ density, average number of hours in shadow per day in each month of year, surrounding vegetation type, surrounding vegetation density, surrounding vegetation area, max temperature in each month of year, min temperature in each month of year, number of sunny days per year, number of cloudy days, number of rainy days, number of days per year heating was on and its max temperature, number of days per year cooling was on and its min temperature, precipitation (inches), snowfall (inches), snow depth (inches), days below 30F, days below 20F, days below 10F, days below 0F, days above 80F, days above 90F, days above 100F, days above 110F, direction of wind with max speed, max wind speed, days with frost, days with fog, and average temperature and having verified and measured the EUI of those buildings in past and labelled those as (1: Highest EUI, 2, 3, 4, 5: Lowest EUI) which is the label and target of the prediction, predict the EUI rating of a future building provided that all the information (listed above) would be available for that building during the prediction process.

3. Predict/Classify air quality: Knowing the data features of past samples of air including the area the sample was taken, the date of samples, PM2.5, PM10, NO, NO2, NOx, NH3, CO, SO2, O3, Benzene, Toluene, Xylene, AQI, etc. and knowing each one of those air samples, belongs to one of the following classes (Good, Moderate, Poor, Satisfactory, Severe, Very Poor), classify and decide whether any new air sample belongs to one of the classes, (Good, Moderate, Poor, Satisfactory, Severe, Very Poor).

4. Forecasting solar energy efficiency: Solar energy is one of the most environmentally friendly sources of energy. It is possible to forecast the output power of the solar panels during different time of the day, depending on different meteorological parameters over a period. For example, DHI (Diffused Horizontal Irradiance) represents the solar radiation that does not arrive on a direct path from the sun but has been scattered by clouds and particles in the atmosphere and comes equally from all directions. Knowing the past values of thousands of the data features such as time stamp (date and time of measurement), temperature, dew point, surface albedo, pressure, wind direction, wind speed, Ozone (Dobson Units), cloud type, solar zenith angle, precipitable water, relative humidity and other potential relevant parameters during those past times and having measured DHI value in Watts/m2 as of three levels (0: No Power, 1: 1 Watts/m2 to 300 Watts/m2 and 2: above 300 Watts/m2) during those past time slots, for a future time slot forecast the level of DHI, provided that all those data features are available (and also show the probabilities).

5. Predict the battery failure in electric cars next 2 hours: Butterfly AI can help predict whether battery of an electric car will fail in next 2 hours or not. In battery electric cars, it is crucial to have a predictive view of the performance high-voltage batteries as they face significant load fluctuations due to environmental parameters such as humidity, temperature, driving conditions and driver behaviour. Having measured and recorded the key data features of thousands of past journeys of the electric cars as indicators of the performance of their batteries including time (s), velocity (km/h), elevation (m), battery voltage (V) at max speed, battery voltage at max acceleration, battery voltage at max outside car temperature, battery voltage at max air conditioning usage, battery voltage at max heating usage, battery current (A) at max speed, battery current at max acceleration, battery current at max outside car temperature, battery current at max air conditioning usage, battery current at max heating usage, battery temperature (°C) at max speed, battery temperature at max acceleration, battery temperature at max outside car temperature, battery temperature at max air conditioning usage, battery temperature at max heating usage, max battery temperature , throttle (%), motor torque (Nm), longitudinal acceleration (m/s^2), regenerative braking signal, SoC (%), displayed SoC (%), min. SoC (%), max SoC (%), heating power CAN (kW), heating power LIN (W), requested heating power (W), AirCon power (kW), heater signal, heater voltage , heater current , ambient temperature , ambient temperature sensor , coolant temperature heater core , requested coolant temperature , coolant temperature inlet , coolant volume flow +500 (l/h), heat exchanger temperature , cabin temperature sensor , temperature coolant heater inlet , temperature coolant heater outlet , temperature heat exchanger outlet , temperature defrost lateral left , temperature defrost lateral right , temperature defrost central , temperature defrost central left , temperature defrost central right , temperature foot wheel driver , temperature foot wheel co-driver , temperature feet vent co-driver , temperature feet vent Driver , temperature head co-driver , temperature head driver , temperature vent right , temperature vent central right , temperature vent central left , temperature vent right and having verified, recorded and labelled whether the battery has failed 2 hours after the point of measurements of the parameters (described above) during those past journeys, for current ongoing electric car journey, predict whether the battery will fail in next 2 hours (and present the probability and certainty of prediction) provided that the data features (presented above) are available during the time of prediction.

6. Decide whether a water sample is drinkable by human: Having the data features (characteristics) of hundreds of past water samples such as Nitrates and Nitrites , turbidity and total suspended solids (TSS), pH scale, water temperature, hardness of water, fecal indicator bacteria, flow of the water, total dissolved solids (TDS), total organic carbon (TOC), Polycyclic Aromatic, Hydrocarbons (PAH), Pesticides, Lead, Iron, Radionuclides, and some other potentially relevant data features, and having verified and labelled whether those past water samples were consumable by human or not, predict/decide whether a new water sample is also consumable by human or not (and with what probability and certainty).

Or any other binary or multi-class prediction or classification use cases required by environmental experts to make decisions.