What Is Curve Fitting? Fitting Models to Data Made Easy with MATLAB

1. Audio
   Observation

   The speaker says, “First, we select the data we want, wind speed on the x-axis and power output on the y.”

2. Diagram
   Observation

   In the Curve Fitter dialog, X data is set to windSpeed.

x

Independent variable in the fitted curve, here wind speed.

wind speed values shown on the plot axis

1. Audio
   Observation

   The speaker says, “wind speed on the x-axis and power output on the y.”

2. Diagram
   Observation

   In the Curve Fitter dialog, Y data is set to powerOutput.

y

Dependent variable in the fitted curve, here power output.

power output values shown on the plot axis

1. Formula
   Observation

   Code overlay shows parameters b1-b7 and predictor variable x.

2. Audio
   Observation

   The speaker says a nonlinear model can be created from the Statistics and Machine Learning Toolbox.

1. The exact physical meaning of each coefficient beyond b1-b3 for wind speed and b4-b7 for other factors is not stated in the clip.

b1-b7

Parameters of the nonlinear model; b1-b3 represent coefficients for wind speed and b4-b7 represent linear relation for other factors.

model coefficients

1. Formula
   Observation

   Code overlay includes “& predictor variable 'x'” and uses x(:,1), x(:,2), x(:,3), x(:,4), x(:,5).

1. The clip does not explicitly map each column x(:,1) through x(:,5) to a named factor in the visible code.

x

Predictor variables passed to the nonlinear model function.

matrix or table of predictors

1. Formula
   Observation

   Code overlay defines modelFun = @(b,x) ... with $b(1)$./($1+\exp$(-b(2)*(x(:,1)-b(3)))) + ...

$b(x, x)$

Nonlinear model function returning predicted power output from parameters and predictors.

model evaluation function

1. Formula
   Observation

   Code overlay shows fitnlm(X, y, modelFun, initialParams).

1. The exact contents of X, y, and initialParams are not shown in this clip.

fitnlm(X, y, modelFun, initialParams)

MATLAB call used to fit the nonlinear model to predictor matrix X and response y.

nonlinear model fitting command

1. Formula
   Observation

   The table displays the header 'p-value' and the value '0.32' for the 'Wind Shear' predictor.

p-value

A statistical measure used to determine the significance of a predictor variable in a model.

(0, 1]

1. Formula
   Observation

   The comparison table displays the column header 'R_squared' with values 0.98962 for both approaches.

R_squared

Coefficient of determination, representing the proportion of the variance in the dependent variable that is predictable from the independent variables.

[0, 1]

1. Formula
   Observation

   The comparison table displays the column header 'RMSE' with values 3.358 and 3.3579.

RMSE

Root Mean Square Error, a standard way to measure the error of a model in predicting quantitative data.

[0, $\infty$)

1. Formula
   Observation

   The comparison table displays the column header 'SSE' with values $7.9075e+05$ and $7.9073e+05$.

SSE

Sum of Squared Errors, the sum of the squares of the errors (deviations between the model predictions and the corresponding observations).

[0, $\infty$)

1. Diagram
   Observation

   The Simulink model shows an input port labeled windSpeed feeding two 1-D Lookup Table blocks.

2. Audio
   Observation

   The narrator says the second lookup table represents the relationship between wind speed and power output.

windSpeed

Input variable representing wind speed in the Simulink model.

Scalar input to the lookup tables; units not stated in this clip.

1. Diagram
   Observation

   The upper lookup block is labeled mpptOmega and outputs to wReference.

2. Audio
   Observation

   The narrator identifies the first lookup table as showing the MPPT omega relationship.

mpptOmega

Lookup-table output associated with the MPPT omega relationship.

Output of the first 1-D lookup table; exact physical units not shown in this clip.

1. Audio
   Observation

   The speaker defines curve fitting as processing data by applying a trend to raw data through literally fitting a curve to it, by fitting a mathematical expression that best represents the data.

2. Animation
   Observation

   A scatter plot of colored dots appears, and a blue curve is drawn through them to illustrate fitting a mathematical expression to data points.

Curve fitting is the process of applying a trend to raw data by fitting a mathematical expression that best represents the data.

1. Audio
   Observation

   The speaker states that data in its raw form is largely unhelpful for an engineer trying to make computations or conclusions, and must be processed to become truly usable.

Raw data must be processed to become usable for making computations or drawing conclusions.

1. Audio
   Observation

   The speaker mentions that applying techniques like pre-processing and cleaning up data makes the curve fit more robust against pesky outliers and disruptive noisy data points.

Pre-processing and cleaning up data are techniques used to make curve fitting more robust against outliers and noise.

1. Definition of Curve Fitting

1. Audio
   Observation

   The speaker explains that curve fitting allows us to see the story the data is trying to tell by cutting through noise and individual data points to get to overall trends and understand the effects of different factors and parameters.

Curve fitting helps identify overall trends in data, allowing engineers to understand the effects of different factors and parameters despite individual noise.

1. Definition of Curve Fitting

1. Audio
   Observation

   The speaker outlines the video's goals: quickly and interactively fit a curve, extend to non-linear regression, optimize by constraining bounds, and integrate into Simulink.

2. Caption evidence
   Observation

   On-screen text lists: 'Fitting Curves to Data', 'Non-Linear Regression', 'Optimization with Constraints', 'Integration into Simulink'.

The video will cover interactive curve fitting, extending to non-linear regression for statistical significance, optimizing with coefficient bounds, and integrating into Simulink.

1. Audio
   Observation

   The speaker describes wind turbines rotating when wind blows, turning a generator which creates power in the form of electricity.

2. Animation
   Observation

   An animation shows a wind turbine spinning, then transitions to show the internal generator mechanism connected to power lines.

Wind causes turbine blades to rotate, which turns a generator to produce electrical power.

1. Audio
   Observation

   The speaker states that based on the chain of events, we can plot out the sigmoidal relationship between wind speed and the power output of the turbine, assuming this relationship is sigmoidal since the wind needs to be at a certain speed to begin pushing the blades.

2. Animation
   Observation

   A white S-shaped curve is drawn across the screen next to a wind turbine icon.

The relationship between wind speed and power output in a wind turbine is modeled as a sigmoidal curve because a minimum wind speed is required to start generating power.

1. Wind Turbine Power Generation Mechanism

1. Audio
   Observation

   The speaker asks how to get from data with outliers to the sigmoidal curve and answers, “Well, curve fitting, of course.”

The video presents curve fitting as the practical way to move from raw wind-speed/power-output observations, including outliers, to a smooth functional relationship represented by a sigmoidal curve.

1. Raw data contain multiple observations and some outliers.

2. The goal is to estimate a smooth relationship between variables.

1. Audio
   Observation

   The speaker says there is a dataset of various factors affecting the power output of a wind turbine and that they can look at the important relationship between wind speed and power output.

2. Diagram
   Observation

   Spreadsheet columns include turbine_id, time_stamp, wind_speed, wind_direction, air_density, turbulence_intensity, wind_shear, power, and outlier.

The example uses a tabular dataset containing several environmental and operational factors for wind turbines, with power as the response variable and wind speed singled out as the main predictor for the first curve fit.

1. Data are organized in rows of observations and columns of variables.

2. Power output is treated as the dependent variable.

1. Audio
   Observation

   The speaker says, “Before we do that, we can apply some basic preprocessing steps to clean our data and prepare it for curve fitting.”

2. Audio
   Observation

   The speaker adds, “We've already identified anomalous events from the performance logs, so we'll just remove those.”

3. Diagram
   Observation

   A spreadsheet view highlights rows associated with outlier handling, and a MATLAB documentation page for Data Cleaner appears.

1. The exact rule used to mark anomalous events is not stated in the clip.

Before fitting a curve, the workflow removes anomalous events already identified in the logs so the model is estimated from cleaned data rather than from records flagged as outliers.

1. Anomalous events have been identified in the performance logs.

2. The goal is to prepare the data for curve fitting.

1. Dataset structure for wind turbine power modeling

1. Audio
   Observation

   The speaker says, “Let's now use the curve fitter app in MATLAB to not only plot the data points showing this relationship, but also fit a curve to it.”

2. Audio
   Observation

   The speaker describes selecting wind speed on the x-axis and power output on the y-axis, then choosing a sigmoidal curve.

3. Diagram
   Observation

   The MATLAB Curve Fitter window shows Select Data, Fit Options, a scatter plot, and a fitted curve.

The video demonstrates an interactive fitting workflow: open Curve Fitter, choose the predictor and response variables, inspect the scatter plot, select a model family, and obtain a fitted curve together with goodness-of-fit information.

1. The data are available in MATLAB.

2. The user wants both visualization and model fitting in one interface.

1. Curve fitting as the method for turning noisy data into a smooth model
2. Dataset structure for wind turbine power modeling

1. Audio
   Observation

   The speaker says, “As we can see by the shape of the data, we want to select a sigmoidal curve.”

2. Diagram
   Observation

   The fitted plot shows an S-shaped curve through the scatter of wind speed versus power output.

The model family is chosen visually: because the scatter plot rises from low values, steepens in the middle, and flattens at higher wind speeds, the video identifies the relationship as sigmoidal and fits a sigmoid curve.

1. The plotted relationship has an S-shaped trend.

2. The response saturates at higher predictor values.

1. Using MATLAB Curve Fitter to plot data and fit a model

1. Audio
   Observation

   The speaker says, “As shown here, wind speed clearly has a big impact, and all the others have very small impacts, with all their curves looking close to horizontal.”

2. Diagram
   Observation

   In the Plot Slice window, the Wind Speed panel has a steep S-shaped curve, while the other panels are nearly flat.

1. The claim is presented for this dataset and fitted model, not as a universal law for all turbines.

For the fitted wind-turbine model shown in the clip, wind speed has a much larger effect on predicted power output than wind direction, air density, turbulence, or wind shear.

1. A nonlinear model has been fitted using wind speed plus four other factors.

2. The Plot Slice display is based on that fitted model.

For the dataset and model shown in this video segment.

1. Audio
   Observation

   The speaker says, “for the other factors, we can assume a linear relationship for simplicity.”

2. Formula
   Observation

   Code overlay uses $b(4)$.*x(:,2) + $b(5)$.*x(:,3) + $b(6)$.*x(:,4) + $b(7)$.*x(:,5).

3. Diagram
   Observation

   The Plot Slice panels for Wind Direction, Air Density, Turbulence, and Wind Shear appear close to horizontal.

1. The clip states the linearity assumption for modeling convenience; it does not prove that the true relations are exactly linear.

In the multi-factor model shown, the non-wind-speed factors are assumed to enter linearly, and their fitted effects appear small enough that their plot-slice curves are close to horizontal.

1. The model uses linear terms for the other factors.

2. The displayed Plot Slice corresponds to the fitted model.

For the fitted model and plotted example in this segment.

1. Audio
   Observation

   Speaker concludes wind shear lacks statistical significance based on its p-value.

2. Formula
   Observation

   p-value for Wind Shear is 0.32.

Because the p-value for the 'Wind Shear' predictor is 0.32, which is greater than 0.05, it does not have a statistical significance on the power output and can be removed from the model.

1. The significance level is 0.05.

No quantifiers are stated in the source.

1. Audio
   Observation

   Speaker states: 'As we can see, the two models are quite similar, and either functionality is usable depending on the requirements and the problem to solve.'

2. Formula
   Observation

   Table shows nearly identical values for R_squared, R_squared_adj, RMSE, and SSE across both approaches.

The standard curve fitting approach (Fitnlm) and the constrained optimization approach produce statistically similar models for this dataset.

1. The evaluation is based on R-squared, Adjusted R-squared, RMSE, and SSE.

No quantifiers are stated in the source.

1. Audio
   Observation

   The narrator states that the second lookup table represents the relationship between wind speed and power output.

2. Diagram
   Observation

   The lower branch is labeled mpptActivePowerDemand and feeds through MW to W to activePowerDemand.

In the shown subsystem, the lower 1-D lookup table represents the relationship between wind speed and power output.

1. The model is the one displayed in the Simulink window at the start of the clip.

For the specific subsystem shown in the video.

1. Audio
   Observation

   The narrator says the subsystems and lookup tables are calibrated based on the latest experimental data.

2. Diagram
   Observation

   The updated model contains curveblock derived from the earlier fitted data plot.

After replacing the lower lookup with the exported curve-fit block, the narrator characterizes the subsystem and lookup tables as calibrated based on the latest experimental data.

1. The exported block comes from the Curve Fitter model shown immediately before replacement.

For the demonstrated workflow in this clip.

1. Audio
   Observation

   The speaker moves from describing the dataset, to preprocessing, to selecting variables in Curve Fitter, to choosing a sigmoidal curve.

2. Diagram
   Observation

   Spreadsheet data, then Curve Fitter scatter plot, then fitted S-shaped curve.

1. Expression

   $$$$
   Explanation

   Start with a tabular dataset containing wind-related factors and power output.

   Justification

   The speaker introduces the dataset and the spreadsheet view shows columns such as wind_speed and power.

   Shown in the video
2. Expression

   $$$$
   Explanation

   Remove anomalous records identified from performance logs before modeling.

   Justification

   The narration explicitly says basic preprocessing is applied and anomalous events are removed.

   Shown in the video
3. Expression

   $$$$
   Explanation

   Select wind speed as the x-variable and power output as the y-variable in Curve Fitter.

   Justification

   The speaker states this selection and the dialog shows X data = windSpeed, Y data = powerOutput.

   Shown in the video
4. Expression

   $$$$
   Explanation

   Inspect the scatter plot shape and choose a sigmoidal model family.

   Justification

   The narration says the shape of the data indicates a sigmoidal curve, and the interface changes to a sigmoid fit.

   Shown in the video
5. Expression

   $$$$
   Explanation

   Obtain the fitted curve and goodness-of-fit summary.

   Justification

   The resulting plot shows an S-shaped curve over the data, and the side panel displays fit statistics.

   Shown in the video

The example demonstrates a practical sequence: clean the data, choose predictor and response, inspect the scatter shape, and fit a sigmoidal curve in MATLAB.

1. Audio
   Observation

   The speaker transitions from the fitted sigmoidal wind-speed relation to considering other wind factors and creating a nonlinear model.

2. Formula
   Observation

   Code overlay shows modelFun = @(b,x) with a sigmoid term for x(:,1) and linear terms for x(:,2) through x(:,5).

3. Diagram
   Observation

   Plot Slice window displays separate effect curves for each factor.

1. The clip does not show the numerical estimation procedure for b1-b7 in detail.

1. Expression

   $$P \sim \frac{b_1}{1+\exp(-b_2(x_1-b_3))} + b_4 x_2 + b_5 x_3 + b_6 x_4 + b_7 x_5$$
   Explanation

   Use a sigmoid term for wind speed and add linear terms for the remaining factors.

   Justification

   The narration says wind speed is sigmoidal and the other factors can be assumed linear for simplicity; the code overlay matches this structure.

   Shown in the video
2. Expression

   $$\texttt{fitnlm(X, y, modelFun, initialParams)}$$
   Explanation

   Fit the nonlinear model to the predictor matrix X and response y using initial parameter guesses.

   Justification

   The code overlay explicitly shows this MATLAB fitting call.

   Shown in the video
3. Expression

   $$$$
   Explanation

   Inspect each predictor’s partial effect in the Plot Slice window.

   Justification

   The speaker says the impact of each factor can be seen in the plot slice window, and the interface shows one panel per factor.

   Shown in the video
4. Expression

   $$$$
   Explanation

   Compare curve steepness across panels to judge relative impact.

   Justification

   The narration interprets the steep wind-speed curve versus nearly flat curves for the other factors as evidence of differing impact sizes.

   Shown in the video

The video extends the single-variable curve fit into a nonlinear regression model and uses partial-effect plots to compare the influence of each factor on predicted power output.

1. Audio
   Observation

   Speaker says: 'We can define the curve fitting problem as minimizing the objective function, and we will rewrite the fitting equation we used earlier as the objective function which minimizes the sum of squared errors.'

2. Diagram
   Observation

   The Objective field in the Optimize task shows: sum((clean_turbine_data.power - (x1 ./ ($1 + \exp$(-x2 * clean_turbine_da...)))^2).

1. The full formula is truncated in the UI, but the structure is clear.

1. Expression

   $$\text{Objective} = \sum (y_{data} - f(x; \theta))^2$$
   Explanation

   The goal is to minimize the sum of the squared differences between the observed data (power) and the model's predicted values.

   Justification

   Definition of least squares optimization.

   Shown in the video

The curve fitting problem is successfully translated into a standard nonlinear least squares optimization problem.

1. Diagram
   Observation

   The clip first shows the fitted scatter plot and logistic equation in Curve Fitter.

2. Animation
   Observation

   The user opens Export and selects Create Simulink Lookup Table.

3. Diagram
   Observation

   The Simulink model then shows the lower lookup replaced by curveblock.

4. Audio
   Observation

   The narrator explains that the earlier model can be exported as a Simulink lookup table and replaced in the subsystem.

1. Expression

   $$$$
   Explanation

   Start from the previously fitted wind-speed/power-output model displayed in Curve Fitter.

   Justification

   The video explicitly returns to the earlier model created with the Curve Fitter app.

   Shown in the video
2. Expression

   $$f(x) = \frac{a}{1 + \exp\left(-\frac{x-b}{c}\right)}$$
   Explanation

   Use the displayed logistic fit as the mathematical representation of the data relation.

   Justification

   The equation and fitted parameters are visible in the results panel.

   Shown in the video
3. Expression

   $$$$
   Explanation

   Export the selected fit as a Simulink lookup table.

   Justification

   The Export menu is opened and Create Simulink Lookup Table is chosen on screen.

   Shown in the video
4. Expression

   $$$$
   Explanation

   Insert the generated lookup block into the lower branch of the Simulink subsystem, replacing the prior mpptActivePowerDemand lookup.

   Justification

   The updated model shows the new block named curveblock in the lower path.

   Shown in the video

The clip demonstrates a direct workflow from a fitted logistic curve to a Simulink lookup-table implementation for the wind-speed/power-output branch.

1. Diagram
   Observation

   A slide lists Fitting Curves to Data, Non-Linear Regression, Optimization with Constraints, and Integration into Simulink.

2. Audio
   Observation

   The narrator summarizes going from data to fitted curves, regression with multiple predictors, optimization with constraints, and finally creating a Simulink lookup table from the original curve fit.

1. Expression

   $$$$
   Explanation

   Begin with data.

   Justification

   The first bullet on the summary slide is Fitting Curves to Data.

   Shown in the video
2. Expression

   $$$$
   Explanation

   Move to fitted curves and non-linear regression with multiple predictors.

   Justification

   The slide lists Non-Linear Regression and the narration mentions regression with multiple predictors.

   Shown in the video
3. Expression

   $$$$
   Explanation

   Then apply optimization with constraints.

   Justification

   Optimization with Constraints appears as the third bullet on the slide.

   Shown in the video
4. Expression

   $$$$
   Explanation

   Finally integrate the result into Simulink as a lookup table.

   Justification

   The fourth bullet is Integration into Simulink, and the narration ends with creating a Simulink lookup table from the original curve fit.

   Shown in the video

The closing recap frames the tutorial as a sequence from data fitting to regression, constrained optimization, and Simulink integration.

1. Audio
   Observation

   The speaker introduces data from wind turbines as an example, explaining the physical mechanism and the resulting sigmoidal relationship between wind speed and power output.

2. Animation
   Observation

   Animations of wind turbines, generators, and a sigmoid curve are shown to illustrate the example.

Model the relationship between wind speed and power output for a wind turbine.

1. Wind turbines rotate when wind blows.

2. Rotation turns a generator to create electricity.

3. Wind needs to be at a certain speed to begin pushing the blades.

Plot the relationship between wind speed and power output.

1. Expression

   $$$$
   Explanation

   Identify the physical chain of events: wind blows -> blades rotate -> generator turns -> electricity is created.

   Justification

   Stated by the speaker.

   Shown in the video
2. Expression

   $$$$
   Explanation

   Determine the shape of the relationship based on the physical constraints (minimum wind speed required).

   Justification

   Stated by the speaker.

   Shown in the video
3. Expression

   $$$$
   Explanation

   Conclude that the relationship is sigmoidal.

   Justification

   Stated by the speaker and illustrated with an S-curve animation.

   Shown in the video

The relationship between wind speed and power output is sigmoidal.

No verification step is shown in the source.

1. Audio
   Observation

   The speaker walks through loading the dataset, cleaning it, selecting wind speed and power output, and fitting a sigmoidal curve in MATLAB.

2. Diagram
   Observation

   Spreadsheet columns, Curve Fitter dialog, scatter plot, and fitted S-shaped curve are all shown.

1. The final fitted equation coefficients for the one-variable sigmoid are visible only briefly and are not fully discussed in the narration.

Given a wind-turbine dataset with multiple factors, estimate the relationship between wind speed and power output using curve fitting.

1. A spreadsheet/table with columns including wind_speed, power, and an outlier flag.

2. Anomalous events have already been identified from performance logs.

3. MATLAB Curve Fitter is available.

Produce a fitted curve describing how power output varies with wind speed.

1. Expression

   $$$$
   Explanation

   Load the dataset and identify the relevant variables for the first model.

   Justification

   The narration says the data are pulled from the spreadsheet and prepared for plotting in MATLAB.

   Shown in the video
2. Expression

   $$$$
   Explanation

   Apply basic preprocessing by removing anomalous records.

   Justification

   The speaker explicitly says anomalous events are removed before curve fitting.

   Shown in the video
3. Expression

   $$x = \text{windSpeed},\quad y = \text{powerOutput}$$
   Explanation

   Choose wind speed as the predictor and power output as the response.

   Justification

   The Curve Fitter selection dialog shows X data = windSpeed and Y data = powerOutput.

   Shown in the video
4. Expression

   $$$$
   Explanation

   Select a sigmoidal fit type because the scatter plot has an S-shaped trend.

   Justification

   The narration states that the shape of the data indicates a sigmoidal curve should be selected.

   Shown in the video
5. Expression

   $$$$
   Explanation

   Review the fitted curve and goodness-of-fit panel, then decide the fit is acceptable without further refinement.

   Justification

   The speaker says there is a pretty good fit and mentions optional refinement by excluding outliers or changing advanced options.

   Shown in the video

A sigmoidal curve is fitted to the wind-speed/power-output data in MATLAB, giving a smooth model of the relationship.

The fitted curve is visually overlaid on the scatter plot, and the interface reports goodness-of-fit measures such as SSE, R-square, and adjusted R-square.

1. Audio
   Observation

   The speaker lists additional factors and says a nonlinear model can be created from the Statistics and Machine Learning Toolbox.

2. Formula
   Observation

   Code overlay defines modelFun with a sigmoid term for wind speed and linear terms for the other factors.

3. Diagram
   Observation

   Plot Slice window shows separate effect curves and p-values for Wind Speed, Wind Direction, Air Density, Turbulence, and Wind Shear.

1. The exact numeric coefficient estimates are not narrated in this clip.

2. The precise mapping from each x(:,j) column to each named factor is implied by context but not verbally enumerated one by one.

Extend the single-variable fit to understand the combined effects of wind speed, wind direction, air density, turbulence, and wind shear on power output.

1. The wind-speed/power relation is known to be sigmoidal.

2. Additional factors are wind direction, air density, turbulence, and wind shear.

3. A nonlinear fitting function from MATLAB is available.

Create and interpret a nonlinear model including all listed factors.

1. Expression

   $$$$
   Explanation

   List the additional factors beyond wind speed.

   Justification

   The slide and narration name Wind Direction, Air Density, Turbulence, and Wind Shear.

   Shown in the video
2. Expression

   $$P = \frac{b_1}{1+\exp(-b_2(x_1-b_3))} + b_4 x_2 + b_5 x_3 + b_6 x_4 + b_7 x_5$$
   Explanation

   Define a model with a sigmoid term for wind speed and linear terms for the other factors.

   Justification

   The narration states this modeling choice and the code overlay shows the same structure.

   Shown in the video
3. Expression

   $$\texttt{fitnlm(X, y, modelFun, initialParams)}$$
   Explanation

   Fit the nonlinear model in MATLAB.

   Justification

   The code overlay explicitly shows the fitnlm call.

   Shown in the video
4. Expression

   $$$$
   Explanation

   Open the Plot Slice window to inspect each factor’s effect.

   Justification

   The speaker says the impact of each factor can be seen in the plot slice window.

   Shown in the video
5. Expression

   $$$$
   Explanation

   Compare the plotted curves: wind speed is steep and S-shaped, while the others are nearly horizontal.

   Justification

   The narration directly interprets the visual output as wind speed having a big impact and the others having very small impacts.

   Shown in the video

The multi-factor nonlinear model shows wind speed as the dominant predictor, with the other factors contributing much smaller effects in this example.

The conclusion is checked against the Plot Slice display and the accompanying p-values, where Wind Speed has p-value 0.00 and the other factors have much larger p-values, especially Wind Shear at 0.32.

1. Audio
   Observation

   Speaker analyzes the Wind Shear predictor and decides to remove it.

2. Formula
   Observation

   Table shows Wind Shear has a flat curve and p-value = 0.32.

Determine if the 'Wind Shear' variable should be included in the power output model.

1. The p-value for Wind Shear is 0.32.

2. The significance threshold is 0.05.

Decide whether to keep or remove the variable.

1. Expression

   $$0.32 > 0.05$$
   Explanation

   Compare the p-value of the Wind Shear predictor to the standard significance level.

   Justification

   Statistical hypothesis testing rule.

   Shown in the video
2. Expression

   $$\text{Remove Wind Shear}$$
   Explanation

   Since the p-value is high, the variable is not statistically significant and is axed from the model.

   Justification

   Feature selection based on p-values.

   Shown in the video

Wind shear does not have a statistical significance on the power output and is removed from the model.

The speaker explicitly states the conclusion based on the visual evidence in the table.

1. Audio
   Observation

   Speaker compares the two models using statistical metrics.

2. Formula
   Observation

   Table displays R_squared, R_squared_adj, RMSE, and SSE for both 'Fitnlm approach' and 'Optimization approach'.

Evaluate the performance of a standard curve fit versus a constrained optimization fit.

1. Fitnlm approach: R_squared = 0.98962, RMSE = 3.358, SSE = $7.9075e+05$.

2. Optimization approach: R_squared = 0.98962, RMSE = 3.3579, SSE = $7.9073e+05$.

Determine which method is better or if they are equivalent.

1. Expression

   $$\text{Compare metrics}$$
   Explanation

   Look at the values for R-squared, RMSE, and SSE across both rows.

   Justification

   Model evaluation methodology.

   Shown in the video
2. Expression

   $$\text{Values are nearly identical}$$
   Explanation

   The metrics show negligible differences between the two approaches.

   Justification

   Direct observation of the table.

   Shown in the video

The two models are quite similar, and either functionality is usable depending on the specific requirements.

The speaker's verbal conclusion matches the numerical data presented in the table.

1. Diagram
   Observation

   The Simulink model shows windSpeed entering two lookup branches, with the lower branch later replaced by curveblock.

2. Formula
   Observation

   The Curve Fitter panel shows the logistic fit and its coefficients.

3. Animation
   Observation

   The user exports the fit as a Simulink lookup table and returns to the model.

4. Audio
   Observation

   The narrator explains that the second lookup table represents wind speed versus power output and can be replaced by the exported model.

1. The exact numeric breakpoints inside the generated lookup table are not shown.

2. The units of windSpeed and powerOutput are not explicitly stated in this clip.

Given a Simulink subsystem with a lower lookup table for wind speed versus power output, update that branch using a previously fitted Curve Fitter model.

1. Initial model input is windSpeed.

2. Lower branch initially uses a 1-D lookup labeled mpptActivePowerDemand.

3. The fitted model is a sigmoidal curve $f(x)=a/(1+\exp (-(x-b)/c))$ for powerOutput vs. windSpeed.

4. The Export menu provides Create Simulink Lookup Table.

Replace the lower manual lookup with a Simulink lookup generated from the fitted curve.

1. Expression

   $$$$
   Explanation

   Identify the lower branch as the wind-speed/power-output relationship in the subsystem.

   Justification

   The narrator explicitly says the second lookup table represents that relationship.

   Shown in the video
2. Expression

   $$f(x) = \frac{a}{1 + \exp\left(-\frac{x-b}{c}\right)}$$
   Explanation

   Reopen the earlier fitted logistic model in Curve Fitter.

   Justification

   The video returns to the plot and equation from the earlier modeling step.

   Shown in the video
3. Expression

   $$$$
   Explanation

   Use Export → Create Simulink Lookup Table to generate a lookup representation of the fit.

   Justification

   The menu selection is shown on screen.

   Shown in the video
4. Expression

   $$$$
   Explanation

   Place the generated block into the lower branch, where it appears as curveblock with an S-shaped icon.

   Justification

   The updated Simulink diagram shows the replacement in the lower path.

   Shown in the video

The lower branch is updated to use curveblock, a Simulink lookup table derived from the fitted logistic model.

The narrator states that the subsystems and lookup tables are now calibrated based on the latest experimental data, and the visual model shows the replacement completed.

1. Animation
   Observation

   A grid appears with scattered colored dots. A blue curve is drawn through the dots, illustrating the concept of fitting a mathematical expression to data.

1. Grid

2. Colored dots

3. Blue curve

1. Dots appear randomly on the grid.

2. A blue curve is drawn through the dots.

1. The grid remains static.

Visualizes the process of curve fitting by showing a continuous function approximating discrete data points.

1. Animation
   Observation

   A 3D model of a wind turbine spins, then the camera zooms into the nacelle to show the generator mechanism, followed by a schematic of the generator connected to power lines.

1. Wind turbine

2. Generator

3. Power lines

1. Turbine blades spin.

2. Camera zooms into the generator.

3. Schematic shows connection to power grid.

Illustrates the physical mechanism converting wind energy into electrical energy.

1. Animation
   Observation

   A white S-shaped curve is drawn horizontally across the screen next to a wind turbine icon.

1. White S-curve

2. Wind turbine icon

1. The S-curve is drawn progressively from left to right.

1. The wind turbine icon remains static.

Represents the non-linear relationship between wind speed and power output.

1. Animation
   Observation

   A presenter sits at a desk while a stylized wind turbine, clouds, leaves, and a cow graphic appear and move around him.

2. Audio
   Observation

   The narration mentions that at a certain higher speed the generator maxes out how much electricity it can output.

1. presenter

2. stylized wind turbine

3. clouds

4. leaves

5. cow graphic

6. desk sign reading REAL MODELS HAVE CURVES

1. Clouds and leaves drift across the frame.

2. The turbine graphic rotates.

3. A cow graphic appears near the turbine.

1. The presenter remains seated at the desk.

2. The background sign stays visible on the shelf.

The animation illustrates the spoken point that turbine output does not keep increasing indefinitely; at higher speeds the generator reaches a maximum output, motivating a nonlinear curve shape.

1. Animation
   Observation

   The scene cuts from the presenter to a computer monitor showing MATLAB/Simulink-style wind-turbine models and then a wider workstation view with code on the left.

1. Some small interface text in the brief software montage is not fully legible.

1. computer monitor

2. keyboard

3. mouse

4. tower

5. software windows with turbine model diagrams

6. code window

1. The view zooms out from the monitor to the full workstation.

2. Software windows change from model diagrams to code and plotting views.

1. The workstation setup remains centered on computational modeling.

This visual transition signals that the discussion is moving from qualitative explanation to data-based modeling in software.

1. Diagram
   Observation

   A spreadsheet/table shows many rows and columns including turbine_id, time_stamp, wind_speed, wind_direction, air_density, turbulence_intensity, wind_shear, power, and outlier.

2. Animation
   Observation

   An orange arrow points to rows in the outlier column during the preprocessing discussion.

1. Individual cell values are too small to transcribe reliably except for column names.

1. spreadsheet grid

2. column headers

3. orange pointer arrow

4. outlier column

1. The view scrolls or shifts across the table.

2. The arrow highlights specific rows associated with outlier handling.

1. The table remains organized as rows of observations and columns of variables.

The visual establishes the raw input data structure and supports the narration that anomalous rows are removed before fitting.

1. Diagram
   Observation

   The MATLAB Curve Fitter window shows a scatter plot of powerOutput versus windSpeed, a Select Fitting Data dialog, and a right-side results panel.

2. Animation
   Observation

   The cursor selects data, opens fit-type choices, and the plot updates from scatter-only to scatter plus fitted curve.

1. Some small numeric entries in the results panel are only partially legible.

1. Curve Fitter window

2. scatter plot

3. Select Fitting Data dialog

4. Fit Options panel

5. fitted curve

6. goodness-of-fit table

1. X and Y variables are selected as windSpeed and powerOutput.

2. The fit type changes to Sigmoidal.

3. A smooth S-shaped curve is overlaid on the scatter plot.

4. Fit statistics populate the side panel.

1. The horizontal axis remains wind speed and the vertical axis remains power output.

The animation shows the full modeling loop: choose variables, inspect the data shape, select a model family, and obtain a fitted curve with quantitative diagnostics.

1. Diagram
   Observation

   An orange slide lists Factors Affecting Power Output: Wind Speed, Wind Direction, Air Density, Turbulence, Wind Shear.

1. title text

2. bulleted factor list

1. Each factor is presented in sequence on the slide.

1. All five factors remain listed together on the slide.

The slide marks the conceptual shift from a single-predictor sigmoid fit to a multi-factor modeling problem.

1. Formula
   Observation

   A code overlay appears beside the presenter showing comments about parameters b1-b7 and predictor variable x, a modelFun definition, and a fitnlm call.

1. The overlay is brief, so not every line of surrounding code is readable.

1. presenter

2. code overlay window

3. modelFun definition

4. fitnlm call

1. The code overlay appears while the presenter explains the model structure.

1. The overlay consistently shows a sigmoid term for x(:,1) and linear terms for x(:,2) through x(:,5).

The visual gives the mathematical form of the extended model: nonlinear in wind speed, linear in the other factors, then fitted with MATLAB's nonlinear model routine.

1. Diagram
   Observation

   A Plot Slice window shows five panels labeled Wind Speed, Wind Direction, Air Density, Turbulence, and Wind Shear, with p-values 0.00, 0.00, 0.00, 0.00, and 0.32 respectively.

2. Animation
   Observation

   Different panels are highlighted in sequence while the presenter discusses impact size.

1. The exact statistical test behind the p-values is not explained in the clip.

1. Plot Slice window

2. five predictor panels

3. predicted-y curve in each panel

4. p-value labels

1. Highlight moves from one predictor panel to another.

2. The Wind Speed panel shows a steep S-shaped curve, while the others are nearly flat.

1. All panels share the same overall purpose of showing one factor’s effect at a time.

The visual comparison supports the claim that wind speed dominates the predicted power output in this fitted model, whereas the other factors have much weaker marginal effects.

1. Diagram
   Observation

   The screen shows the MATLAB Live Editor with an 'Optimize' task block. It includes sections for 'Create optimization variables' (listing x1 to x6 with bounds), 'Define problem' (Goal: Minimize, Objective: sum(...)), and 'Specify problem-dependent solver options' (Solver: lsqnonlin (auto)).

1. Optimize task block

2. Variable table

3. Objective function input

4. Solver dropdown

1. The user scrolls through the configuration options for the optimization problem.

1. The overall structure of the Live Editor task remains constant.

This visual demonstrates how to configure a constrained nonlinear optimization problem using a graphical interface in MATLAB, mapping mathematical concepts (variables, bounds, objective) to UI elements.

1. Diagram
   Observation

   The screen displays a Simulink model titled 'Nacelle' and later zooms into a 'Power Controller' subsystem containing an 'MPPT Control' block.

1. Simulink canvas

2. Nacelle subsystem

3. Power Controller subsystem

4. MPPT Control block

1. The view transitions from a high-level system diagram to a detailed subsystem.

1. The underlying Simulink environment remains the same.

This shows the practical application of the fitted mathematical model within a larger dynamic system simulation, specifically for controlling a wind turbine's power output.

1. Audio
   Observation

   The speaker notes that data in its raw form is largely unhelpful for an engineer trying to make computations or conclusions.

Raw data is immediately useful for making computations or conclusions.

Raw data must be processed (e.g., via curve fitting) to become truly usable for engineering analysis.

1. Audio
   Observation

   The speaker emphasizes applying preprocessing and removing anomalous events before curve fitting.

2. Diagram
   Observation

   The outlier column and highlighted rows are shown in the dataset view.

One might think the raw data can be fitted immediately without addressing flagged anomalies.

The video explicitly removes anomalous events from the performance logs before curve fitting, indicating that preprocessing is part of the modeling workflow.

1. Audio
   Observation

   The speaker says, “As we can see by the shape of the data, we want to select a sigmoidal curve.”

A common mistake is to choose a curve type without looking at the scatter plot.

In this example, the S-shaped trend in the plotted data motivates selecting a sigmoidal model rather than an arbitrary fit type.

1. Audio
   Observation

   The speaker says, “for the other factors, we can assume a linear relationship for simplicity.”

Viewers may infer that the other factors are definitely linear in reality.

The video states the linear treatment is adopted for simplicity in the model, not demonstrated as an exact underlying law.

1. Audio
   Observation

   The speaker interprets the Plot Slice window by comparing steepness of the curves across factors.

2. Diagram
   Observation

   One panel is steep while the others are nearly horizontal.

One might expect all listed factors to matter about equally because they all appear in the model.

The Plot Slice display shows that, in this fitted example, wind speed has a large effect while the other factors have very small effects.

1. Audio
   Observation

   Speaker notes the models are similar based on high R-squared values (~0.99).

Believing that an R-squared value close to 1 guarantees the model is perfectly accurate or physically meaningful in all contexts.

While a high R-squared indicates a good fit to the training data, it does not account for overfitting, omitted variable bias, or whether the model respects physical constraints (which is why the optimization approach with bounds was introduced).

1. Animation
   Observation

   The video shows a fitted Curve Fitter model being exported directly as a Simulink lookup table.

2. Diagram
   Observation

   The lower branch changes from a generic 1-D lookup to curveblock with an S-shaped icon.

One might assume lookup tables in Simulink must always be built manually from breakpoints typed into the block.

This clip demonstrates that a model already fitted in Curve Fitter can be exported as a Simulink lookup table and inserted into the model automatically.

1. Audio
   Observation

   The speaker states 'Curve fitting does this processing...' referring back to the need to process raw data.

Curve fitting is a specific technique applied to achieve the goal of processing raw data.

1. Audio
   Observation

   The speaker links the physical mechanism of the wind turbine to the mathematical model of a sigmoidal relationship.

The physical operation of a wind turbine dictates the mathematical form (sigmoid) of its power output model.

1. Audio
   Observation

   The speaker asks how to get from data to the sigmoidal curve and answers curve fitting, then introduces the wind-turbine dataset.

Curve fitting is applied to the wind-turbine dataset to estimate the relationship between variables.

1. Audio
   Observation

   The speaker says preprocessing is applied before curve fitting and then moves to the Curve Fitter app.

Cleaning anomalous records is presented as a preparation step before using Curve Fitter.

1. Audio
   Observation

   The speaker says they will use the curve fitter app in MATLAB to plot the data and fit a curve.

The abstract idea of curve fitting is implemented through the MATLAB Curve Fitter workflow.

1. Audio
   Observation

   The speaker chooses a sigmoidal curve based on the shape of the data.

Within the Curve Fitter workflow, selecting a sigmoidal model is a specific case of choosing a fit type.

1. Audio
   Observation

   The speaker says the fit could be further refined by excluding outliers or modifying advanced options.

After obtaining the initial sigmoid fit, the workflow contains optional refinement steps.

1. Audio
   Observation

   The speaker transitions from the fitted sigmoidal wind-speed relation to understanding all factors with a nonlinear model.

2. Formula
   Observation

   The code overlay keeps the sigmoid term for wind speed and adds linear terms for other factors.

The one-variable sigmoid fit is generalized into a multi-factor nonlinear model that retains the sigmoid dependence on wind speed.

1. Audio
   Observation

   The speaker says the impact of each factor can be seen in the plot slice window.

2. Diagram
   Observation

   Plot Slice displays one panel per predictor from the fitted model.

The Plot Slice window is used to interpret the fitted multi-factor nonlinear model.

1. Audio
   Observation

   The speaker interprets the Plot Slice curves to conclude wind speed has a big impact and the others are small.

2. Diagram
   Observation

   The visual contrast between steep and flat panels supports the statement.

The claim about relative factor importance depends on the visual and statistical evidence shown in the Plot Slice window.

1. Audio
   Observation

   Speaker transitions from removing insignificant factors to adding bounds and constraints for a more precise model.

Identifying significant predictors via p-values is a preliminary step before formulating the final constrained optimization problem for curve fitting.

1. Audio
   Observation

   After setting up the optimization, the speaker immediately moves to comparing the resulting models using statistical metrics.

Statistical metrics like R-squared and RMSE are applied to evaluate the success of the curve fitting achieved through the optimization process.

1. Audio
   Observation

   Speaker defines curve fitting.

1. Definition of Curve Fitting

1. Audio
   Observation

   Speaker explains raw data is unhelpful without processing.

1. Purpose of Processing Raw Data

1. Audio
   Observation

   Speaker describes the sigmoidal relationship in wind turbines.

1. Sigmoidal Relationship in Wind Turbines
2. Wind Turbine Power Generation Mechanism

1. Audio
   Observation

   The speaker asks how to get from data with outliers to the sigmoidal curve and answers curve fitting.

1. Curve fitting as the method for turning noisy data into a smooth model

1. Audio
   Observation

   The speaker says basic preprocessing is applied and anomalous events are removed before curve fitting.

1. Preprocessing step: remove anomalous records before fitting

1. Audio
   Observation

   The speaker says wind speed is placed on the x-axis and power output on the y-axis.

2. Diagram
   Observation

   The Select Fitting Data dialog shows X data = windSpeed and Y data = powerOutput.

1. Using MATLAB Curve Fitter to plot data and fit a model
2. x
3. y

1. Audio
   Observation

   The speaker says the shape of the data indicates a sigmoidal curve should be selected.

1. Choosing a sigmoidal curve from the observed data shape

1. Audio
   Observation

   The speaker mentions excluding specific outlier points or modifying advanced fit options to refine the fit.

1. Possible refinements after an initial fit

1. Diagram
   Observation

   The slide lists Wind Speed, Wind Direction, Air Density, Turbulence, and Wind Shear.

2. Audio
   Observation

   The speaker names the additional factors affecting power output.

1. Extending from one sigmoidal fit to a multi-factor nonlinear model

1. Formula
   Observation

   The code overlay shows a sigmoid term for x(:,1) plus linear terms for x(:,2) through x(:,5).

1. Extending from one sigmoidal fit to a multi-factor nonlinear model
2. $b(x, x)$

1. Audio
   Observation

   The speaker says the impact of each factor can be seen in the plot slice window.

2. Diagram
   Observation

   Separate panels show each factor’s effect curve.

1. Using Plot Slice to compare factor impacts

1. Audio
   Observation

   The speaker says wind speed clearly has a big impact and the others have very small impacts.

2. Diagram
   Observation

   Wind Speed panel is steep; other panels are nearly horizontal.

1. Wind speed has the largest impact among the listed factors in this example
2. Using Plot Slice to compare factor impacts