Smartwatt’s Artificial Intelligence workflow, covering the whole AI Life Cycle from raw data to valuable insights, unlocks a world of possibilities on the path of better and faster results for the Renewable Energy industry.

Data, Data and more Data!

Once Albert Einstein said, “I like to think that the moon is there even if I am not looking at it”. Today, we live in a world of data, data, and more data. We can monitor and get data from all the components of our energy assets.

It has never been easier to collect and store structured and unstructured data in colossal data centres with zettabytes of data, especially with IoT systems. Today we can claim that is it possible to monitor almost everything.

But there are relevant questions to consider on this matter, such as:

If you can’t see the value of your data or you’ve never used it, does it mean it doesn’t exist to you?

Are you getting the right data for you?

Do you have the right talent to extract value from your data?

Collecting data just to create a vast knowledge base is not a solution for any problem. Data is one of the critical ingredients for revenue increase and performance, but we need to join other ingredients to blend until the perfect recipe.

As important as the data collected, there is the team experience. Years and years on the field, solving real problems and learning how the energy components work is a true superpower. If we join team experience with data, we will have a clear view of the scope of our problem. This experience helps us to understand how this data is connected.

It is possible to add value to your business with these two ingredients, but we have some limitations as humans. Our brain is the most advanced intelligent system there is, but we visualize the world from a 3D perspective.

When we have to create a multi-dimensional analysis, it is not easy for the human brain to deal with it. To tackle these problems, humans are using a (new) tool – Artificial Intelligence (AI).

Artificial Intelligence has evolved into a formidable tool in recent years, enhancing human intelligence in many ways. In the renewable energy sector, we have more data than ever. We have years of experience and AI tools capable of combining the information available to add value to the future of the renewable energy business.

So, there is data and experience, but how can AI tools add value to my business?

At Smartwatt, we have a dedicated team of Data Scientists and Data Engineers with a background in mathematics, statistics, computer science, and physics, ready to implement an AI workflow called the “Predictive Thinking approach” to develop customized algorithms using data to add value to your business.

Smartwatt’s Artificial Intelligence Workflow

Clear definition of scope and need: the problem

Understanding the problem is not as easy as it may seem since there are many factors to consider to ensure the correct problem is being tackled. The ability to clearly state the problem is more of an art, and to do that, our team uses the design thinking approach to be more effective.

Operationally, the problems in wind energy can vary from yaw and pitch misalignment to the underperformance of wind turbines. Whether that in solar photovoltaic energy can go from soiling to inverter underperformance. In the end, the possibilities are as many as the data available.

To define the problem, these are usually the topics discussed:

Define the problem in terms of environment, entities and data.

How is the problem currently being solved?

Who are the key people that are involved?

How is the problem currently being solved?

Who are the key people that are involved?

Data Collection and Understanding

Data is where most bottlenecks occur in renewable energy AI projects: all of the data we need is rarely readily available. If that data does exist, obtaining and storing it can be difficult.

Together with the partner IT team, our data engineering team designs a strategy using the most efficient techniques to obtain this data. Once we have the data, it is necessary to understand the variables involved and the time-frame synchronization.

Usually, these are the topics involved:

– All data sources as a Single Version of Truth (SVOT) – store all of an organization’s data in a consistent and non-redundant form.

Identify, record and review: data access, data ownership, data stewardship, metadata, data ethics, governance and regulations;

Compare industry benchmarks and algorithmic baselines for similar problems;

Checking and validating data quality;

External data sources that could impact the problem;

Data exploration;

Data exploration

Also known as exploratory data analysis (EDA), is a process where our team looks at it and understands the data with statistical and visualization methods.

This process helps identify patterns and problems in the dataset and decide which model to use in subsequent steps. This step is essential to assure the AI model will work.

Without data exploration, you may even spend most of your time checking your model without realizing the problem in the dataset.

Our workflow

Bellow is usually the workflow our team performs:

1. Data wrangling – cleaning and unifying messy and complex data sets for easy access and analysis;

2. Data augmentation – increase the amount of data by adding slightly modified copies of already existing data or newly created synthetic data from existing data;

3. Addressing class imbalance – oversampling or undersampling – when one or more classes have insufficient proportions in the training data as compared to the other classes;

4. Transfer learning from synthetic data – storing knowledge gained while solving one problem and applying it to a different but related problem;

5. Feature engineering – dynamic time warping: to select and transform the most relevant variables from raw data when creating a predictive model;

6. Feature representation – vector symbolic architectures: a process of using knowledge to extract features (characteristics, properties, attributes) from raw data.

Let the AI experts come in: AI modelling

Once the data is ready and the problem is clear, Smartwatt is prepared to apply a range of techniques and algorithms to the data. The process usually follows these steps:

1. Building involves learning and generalizing a machine-learning algorithm using training data.

2. Fitting involves measuring the machine learning model’s ability to generalize to never-seen-before examples similar to the data available.

3. Validation involves evaluating a trained model using testing data that comes from a different portion of the training data.

Usually, this is the workflow involved:

1. Build the kick-off AI model.

2. Develop a benchmark model.

3. Build multiple AI models.

4. Determine the correct model evaluation metrics to create an effective evaluation metric that should be accurate, robust, agnostic, scalable and interpretable.

Say what? Making the model cristal clear: Explainability (or XAI)

AI models for the end-user are like black boxes. Something closed that they don’t see the inside or understand. However, the primary assumption is that the end-users will be using the output of these black boxes to support their daily decisions.

To use the AI models’ information and make decisions considering them, it is vital to TRUST. This is key for Smartwatt and the overall success of the workflow.

We continually explain what is behind the model that we are developing during all the processes. Every model should balance accuracy and explainability, focus on the end-user and establish key performance indicators (KPIs) to assess AI risk.

The process usually follows these steps:

1. Apply intrinsic methods (endemic methods to the algorithm) for interpreting and explaining the model output.

2. Apply extrinsic methods for interpretation and explainability.

Extrinsic methods include Partial Dependence Plots (PDP), Individual Conditional Expectation (ICE), Local Interpretable Model Explanation (LIME) and Shapley Addictive Explanations (SHAP).

Ready to launch: AI model deployment

Smartwatt is focused on iteration and experimentation to achieve the perfect model that solves the customer’s problem. However, the organizations manage a considerable portfolio of assets, and the perfect model should be applied and performed in real-time for all the portfolios.

Taking the perfect model from the laptop to the real world is essential for success. Smartwatt MLOps and DevOps teams are responsible for creating a workflow that allows the application of the perfect algorithm to all the assets and keeps it running every second, minute and hour if needed.

The process usually follows these steps:

1. Review and identify the most time-efficient and adaptable deployment approach.

2. Consider model compression and deployment options on-prem: device vs. cloud.

3. Determine CPU usage and memory usage metrics.

4. Consider real-time vs. batch execution/prediction.

5. Consider the number of end-users and applications.

6. Determine the expected formats of output.

7. Define the expected turnaround time.

8. Define the frequency of use.

9. Determine the assurance of the data inputs.

Reality-check: From AI Algorithms to Real-time (and world) Application 

Our partner organizations have large teams with different knowledge that will use the outputs of the AI models. The final goal is to create a tool for enhancing human productivity and capabilities using the value behind the large data sets.

Understanding how the end-users make decisions and interact with software is essential to create a seamless experience to help them trust and use the outputs generated by the AI tools. Accuracy of the results is critical, but having the results whenever they are needed is also essential.

The process usually follows these steps:

1. Monitor and evaluate the performance.

2. Operationalize using AI pipelines (MLOps, AIOps).

3. Train end-users to explore the model and what is inside of it.

The main benefits of AI for the renewable energy industry

Artificial Intelligence is a tool that allows us to unlock the value of renewable energy data. Renewable energy operators shall look to this tool as a new opportunity to be more productive and more efficient.

More importantly, Artificial Intelligence is a strategic decision, and designing a strategy to implement AI tools to renewable energy data will state the evolution from Renewable Producers and Renewables Producers 2.0.

This AI workflow is the foundation of the service we are providing to our partners and has become our DNA. Explainability gives us a competitive advantage and is a source of value creation.

In the end, we are giving a new superpower to our partners – Knowledge to decide in a better and faster way.


Smartwatt’s core competencies are energy systems optimization, from project to installation to asset management in real-time, covering the value chain from top to bottom.

Our company has a culture of future and our team develops solutions that increase the efficiency of their customer’s business processes.

Smartwatt’s mission is empowering our partner’s energy systems with the most advanced tools and processes for optimizing energy consumption, renewable energy production and maintenance operations, making energy simple, safe, affordable and sustainable. For additional info, please contact us.