Milestone & Budget
Milestone Number 1
Name: (Power Plant Digital Model) PDM Sourcing
Description: Gathering, research, and analysis of available PDMs, which may be partially outsourced.
Deliverables: A report discussing different available PDMs, pointing out their strengths and weaknesses, and explaining why we're choosing one or two for our next steps.
Budget and Timeline: $1000, 4 weeks
Milestone Number: 2
Name: Proprietary PDM Purchase
Description: Should the assessment from Milestone 1 determine that a proprietary PDM is essential for the successful execution of our POC, this milestone involves the acquisition of the identified proprietary PDM.
Deliverables: Price and source information for the purchased PDM.
Budget and Timeline: Up to $1000, 2 weeks (concurrent with the last two weeks of milestone 1)
Milestone Number: 3
Name: Analyzing Existing PO Software
Description: We will explore the landscape of existing Power Plant Optimization (PO) software available in the market. Additionally, we will engage with the respective developer companies to gain deeper insights into the capabilities and methodologies employed by their software. This milestone may be partially outsourced.
Deliverables: A report outlining the various competing software, and information about mathematical optimization methods they employ.
Budget and Timeline: $1000, 4 Weeks
Milestone Number: 4
Name: Creating a Test Case
Description: Utilizing insights from Milestones 1 and 3, here we'll design a test case that resembles how a real power plant operates.
Deliverables: A report that fully defines the test case.
Budget and Timeline: $500, 2 Weeks
Milestone Number: 5
Name: Conventional Optimization for Test Case Power Plant
Description: This milestone involves applying traditional optimization techniques, which are commonly used in existing software, to optimize the operation of the test case power plant defined in Milestone 4. The objective is to determine the optimal operational strategies and associated operational costs for each of the conventional optimization methods explored.
Deliverables: a report in the form of PowerPoint slides
Budget and Timeline: $500, 3 Weeks
Milestone Number: 6
Name: Generative Optimization for Test Case Power Plant
Description: This milestone involves utilizing our advanced generative optimization technology to optimize the operation of the test case power plant defined in Milestone 4. Here, our focus shifts from conventional methods to innovative generative techniques to evaluate optimal operational strategies and associated costs.
Deliverables: a report in the form of PowerPoint slides
Budget and Timeline: $500, 5 Weeks
Milestone Number: 7
Name: Comparison between Conventional and Generative Optimization for Test Case Power Plant
It will also involve iterating over milestones 4 and 6
Deliverables: final report of the POC
Budget and Timeline: $500, 4 Weeks
Long Description
Company Name
MLBoost
Summary
Our primary long-term goal is to create decision-making advisor software that utilizes generative technologies to make important industrial systems run better and use energy more wisely. This proposal serves as a crucial Proof of Concept (POC), demonstrating how our technology can enhance the efficiency of power plants. If successful, it paves the way for broader applications, including energy storage, supply chain management, agriculture technology, and financial optimization.
Funding Amount
$5,000
The Problem to be Solved
In the current electricity generation landscape, carbon emissions are a pressing concern, with conventional sources like coal and natural gas maintaining their significant role in the foreseeable future. In addressing this challenge, Power Plant Operation Optimization (PO) software becomes essential. Such software yields operational set points for the plant, guiding decisions such as load allocation among different generators. These set points allow power plant operators to adeptly manage plant operations, ensuring the fulfillment of committed electricity production within a specified timeframe, all while minimizing the associated costs.
For example, imagine you are the operator of a power plant that has two generators, and you have already committed to generating a certain amount of electricity at let's say 5 pm tomorrow. The big question is how to split that committed amount between the two generators in a way that not only meets your electricity target but also keeps your operational costs and carbon footprint as low as possible.
However, due to the slow nature of the core mathematical optimization methods underlying current PO software, operators often have to make simplifications when using software to determine how to run the plant. These simplifications lead to power plants operating in sub-optimal conditions, higher fuel consumption, and increased carbon emissions. Solving this issue holds substantial significance, as our analysis indicates that even a marginal enhancement of plant operational efficiency by a mere 0.1 percent across the USA could lead to a noteworthy reduction of tens of millions of dollars in annual fuel costs and Carbon credits over the forthcoming three decades, all achieved through intelligent decision-making without the need for additional hardware investments.
So, our objective is to address the limitations of current PO software.
Our Solution
In the pursuit of our goal to develop advanced PO software, this proposal specifically establishes a POC stage. The primary aim at this stage is to convincingly showcase the superiority of emerging generative mathematical optimization methods over conventional ones that are at the core of current PO software.
To accomplish this, the initial emphasis is on acquiring available Power Plant Digital Models (PDM). These models, which can be thought of as virtual power plants, use laws of physics to connect operational decisions such as the amount of fuel injected to different boilers, and the ambient temperature and humidity to how much power is produced. An example of such PDM can be found in a paper co-authored by the lead author of the current proposal and published in a top ML conference [1]. A Power Plant Operation (PO) software uses these models to suggest optimal or smart decisions for running the plant. The "smartness" of these decisions is key in judging how good the software is. For example, if we want a certain amount of power, superior software can suggest ways to get that power while using the least fuel possible, all within the safe limits set by the operator.
Once we have the digital models, the next step is to carefully compare how good the operational decisions suggested by our generative method are compared to those from existing PO software.
Upon the successful validation of the potential inherent in our generative decision-making technology, the subsequent step will involve proposing, as part of a future Deep Funding round, the development of software that will utilize the technology and run on the SingularityNet platform.
External services
Such software can utilize the power of SingularityNet's forecasting service SYBIL. Users can tap into that service to obtain valuable predictions, whether it's weather forecasts, energy demand projections, or market trends. These forecasts then act as essential inputs to our decision-making advisor, which operates as a sophisticated black-box inputting forecasts and outputting smart decisions. By integrating these forecasts, our software can suggest decisions that are informed by and adopt to latest forecasts. This advanced pipeline of SingularityNet services will empower power plant operators to make the smartest decisions that lead to more efficient plant operations. It acts as a trusted partner in the decision-making process within power plants, ultimately delivering benefits for both budgets and the environment.
Crucially, the decision-making capabilities of this software are not limited to power plants alone; they can be extended to various domains, including energy storage, agriculture, and supply chain optimization. For example, energy traders can use SingularityNet's SYBIL forecasting service alongside our advanced decision-maker software to boost their battery arbitrage strategies. SYBIL will offer precise electricity price trends, load patterns, and renewable energy forecasts. Then, our proposed decision-making software takes these forecasts and transforms them into decisions, optimizing charging and discharging schedules for maximum profit. As another example, farmers can use it to make informed decisions about water management, which will be especially valuable in regions facing water scarcity. They can utilize SYBIL's forecasts, such as soil moisture predictions, and then our software to determine the optimal irrigation strategy. This approach leads to more efficient resource utilization, increased profitability, and a reduced environmental impact, making it a valuable tool for modern agriculture.
Marketing Strategy
After successfully completing the POC, our primary marketing focus will be to clearly demonstrate the undeniable superiority of our solution compared to existing technologies. Our target audience, including power plant operators, utility companies, energy traders, and sustainability agencies, will witness the compelling evidence of how our generative decision-making advisor software excels in reducing both costs and emissions. We will develop comprehensive case studies, conduct in-depth comparisons, and provide quantifiable metrics that leave no room for doubt. These compelling demonstrations will be shared through various platforms, including our well-established YouTube channel, which already boasts hundreds of subscribers. Collaborations with renowned industry experts and active participation in industry events, as well as publication in top-tier conferences and journals, will further bolster our credibility.
Our Project Milestones and Cost Breakdown
Milestone Number 1
Name: (Power Plant Digital Model) PDM Sourcing
Description: Gathering, research, and analysis of available PDMs, which may be partially outsourced.
Deliverables: A report discussing different available PDMs, pointing out their strengths and weaknesses, and explaining why we're choosing one or two for our next steps.
Budget and Timeline: $1000, 4 weeks
Milestone Number: 2
Name: Proprietary PDM Purchase
Description: Should the assessment from Milestone 1 determine that a proprietary PDM is essential for the successful execution of our POC, this milestone involves the acquisition of the identified proprietary PDM.
Deliverables: Price and source information for the purchased PDM.
Budget and Timeline: Up to $1000, 2 weeks (concurrent with the last two weeks of milestone 1)
Milestone Number: 3
Name: Analyzing Existing PO Software
Description: We will explore the landscape of existing Power Plant Optimization (PO) software available in the market. Additionally, we will engage with the respective developer companies to gain deeper insights into the capabilities and methodologies employed by their software. This milestone may be partially outsourced.
Deliverables: A report outlining the various competing software, and information about mathematical optimization methods they employ.
Budget and Timeline: $1000, 4 Weeks
Milestone Number: 4
Name: Creating a Test Case
Description: Utilizing insights from Milestones 1 and 3, here we'll design a test case that resembles how a real power plant operates.
Deliverables: A report that fully defines the test case.
Budget and Timeline: $500, 2 Weeks
Milestone Number: 5
Name: Conventional Optimization for Test Case Power Plant
Description: This milestone involves applying traditional optimization techniques, which are commonly used in existing software, to optimize the operation of the test case power plant defined in Milestone 4. The objective is to determine the optimal operational strategies and associated operational costs for each of the conventional optimization methods explored.
Deliverables: a report in the form of PowerPoint slides
Budget and Timeline: $500, 3 Weeks
Milestone Number: 6
Name: Generative Optimization for Test Case Power Plant
Description: This milestone involves utilizing our advanced generative optimization technology to optimize the operation of the test case power plant defined in Milestone 4. Here, our focus shifts from conventional methods to innovative generative techniques to evaluate optimal operational strategies and associated costs.
Deliverables: a report in the form of PowerPoint slides
Budget and Timeline: $500, 5 Weeks
Milestone Number: 7
Name: Comparison between Conventional and Generative Optimization for Test Case Power Plant
It will also involve iterating over milestones 4 and 6
Deliverables: final report of the POC
Budget and Timeline: $500, 4 Weeks
Risk and Mitigation
When dealing with cutting-edge technology, especially in complex systems like power plants, inherent risks emerge. We want to reduce these risks through the POC phase before we fully develop our PO software. The POC acts as a feasibility study, aiming to demonstrate the superior performance of generative decision-making technologies compared to conventional methods used in power plant operations. While this superiority has been evident in simpler academic optimization scenarios, our goal is to affirm its adaptability and effectiveness within real-world power plant operations.
In the context of this POC, the first key concern revolves around the availability of proprietary PDMs, which could potentially be limited due to our budget constraints. To address this challenge, we are undertaking an analysis of both proprietary and open-source PDMs in Milestone 1. This dual approach also ensures a comprehensive assessment of all available options.
The second risk is not obtaining enough information about existing PO software beyond what's on company websites. To counter this, we've set aside enough time for Milestone 3. If companies can't provide details, we'll dive into research papers authored by their founders and researchers to gather insights into their methods.
The third long-term risk for this project involves the inherent cautiousness among industry players in adopting new technologies within the high-stakes power plant domain. Stakeholders may understandably exhibit reluctance or a measured approach in embracing unfamiliar methods. While this concern is not an immediate focus during the POC phase, it warrants early consideration. To mitigate this risk, we have implemented several measures:
A) We have allocated a budget to account for the potential purchase of proprietary Power Plant Digital Models (PDMs). This allows us to ensure access to the necessary data sources for accurate analysis. Proprietary PDMs are favored because, in general, they offer a closer representation of real-world plant dynamics compared to open-source alternatives.
B) To address the concern of practical relevance, we have allotted a generous timeline to meticulously define the benchmark test case in milestone 4. This time allocation is crucial to ensure that the test case closely mirrors the intricacies of real-world plant operations, thereby enhancing the applicability of the insights derived. Performing our study on such a realistic setting not only enhances the validity of our findings but also helps establish credibility and trust among stakeholders as we progress in the project.
C) We have designated a substantial timeline for milestone 7. This milestone involves a comprehensive comparison between the outcomes of conventional and generative technologies. This thorough evaluation necessitates iterative engagement with milestone 5 (Conventional Optimization Cost Analysis) and milestone 6 (Generative Optimization Cost Analysis), ensuring a robust assessment of their respective efficiencies and advantages.
D) Recognizing the importance of practical expertise, we plan to recruit an industry expert with experience in power plant operations during the later stages of the project, once the POC is successfully completed.
While the current proposal does not discuss our intellectual property (IP) strategy, we want to assure stakeholders that this critical aspect is not overlooked. Our current priority is to successfully complete the POC phase and validate the core generative decision-making technology. Once this milestone is achieved and the effectiveness of our methods is firmly established, we are fully committed to promptly addressing the IP concerns surrounding our innovations. We will engage with legal experts to devise a comprehensive IP protection strategy that safeguards our pioneering techniques from competitors, while still making them accessible to non-commercial uses.
Voluntary Revenue
During the POC phase, revenue generation is not the focus. Instead, our current objective is to establish the viability and effectiveness of our approach. However, should the future software development get funded by a subsequent Deep Funding round, a revenue-sharing model will come into play that will adhere to API Calls “user-friendly” template for revenue sharing: revenues exceeding $1,000 per month will be subject to a 10% fee.
Open Source
After the successful completion of POC, our power plant optimization project will offer select components and tools under permissive licensing to promote collaboration and community-driven development while safeguarding our core proprietary algorithms. The specific details of this open-source strategy, including licensing terms, will be determined through consultation with our legal team after the successful completion of POC. During POC, we'll actively engage with the SingularityNET (SNET) and Deep Funding communities, keeping them informed of our progress and seeking valuable feedback. Regular updates will be accessible mainly on our YouTube channel MLBoost and SNET discord channel to promote discussions and share insights. As part of our commitment to community involvement, we'll introduce opportunities for direct contribution and interaction through various channels.
Our Team
Currently, our team consists of Dr. M. Torabi Rad, who holds a Ph.D. in energy conversion and has co-authored papers on power plant and engineering system optimization, Kevin R.C. a DF2 2x Awardee, and, as our advisor, Dr. Matthew Ikle, Chief Science Officer at SingularityNET.
Mahdi T.R. - Ideation Lead and Senior Data Scientist
Kevin R.C. - Senior Data Scientist
- Senior Data Scientist / AI Researcher
- DF2 2x Awardee (
and
)
- Former Lead Core Developer and Maintainer of NeuralProphet
- 8+ years in the finance and crypto domains
- 3+ years of Python PyPI open-source experience (including
,
, and
)
- LinkedIn
-
GitHub
Advisor
Matt I. - AI and Climate Advisor
- Chief Science Officer (CSO) at SingularityNET
- LinkedIn
Related Links
1. https://www.climatechange.ai/papers/neurips2022/19
Sort by