In Chapter 2, you looked at the IoT and IIoT and their uses. You also saw via a practical example how an IoT system is designed using Raspberry Pi, Arduino, and Python with machine learning. By seeing a practical example, you now have an understanding of the implementation aspects of any IoT or IIoT applications such as creating monitoring systems for giving advanced warning signals to users by gathering data from IoT sensors and then measuring it against a benchmark number such as a temperature threshold and alerting the users through email or GUI alerts.
In this chapter, you are going to see the state-of-the-art implementation of machine learning and IoT in the telecom, energy, and agriculture domains. I will be walking you through some rapid implementations of machine learning and the IoT in each of these sectors, helping you understand the nuances and the complexity involved in building applications for them.
State-of-the-Art Implementation of Machine Learning and the IoT in the Telecom Domain
The telecom domain is going through a rapid technological wave right now all over the world. The heart of this change is the 5G spectrum introduction. In my blog post on two reasons why 5G IoT is safe (https://pmauthor.com/2019/07/02/3-reasons-why-5g-iot-is-safe/), I talk about what the spectrum of each telecom technology was able to achieve. Mobile technology has seen four generations of technological development with 1G, 2G, 3G, and 4G. Each generation is a set of telephone network standards for communication across mobile devices such as mobile phones and tablets. The 1G gave 2.4 kbps of connectivity, 2G gave 64 Kbps of connectivity and was based on GSM, 3G gave 144 kbps to 2 Mbps of connectivity, and 4G gave 100 Mbps to 1 Gbps of connectivity and is based on LTE technology. The 1G systems started in Australia and were analog in nature. Truly digital telecommunication started with the introduction of the 2G standard, which brought with it the CDMS and GSM protocols with SMS and MMS services, among others. 3G led to the introduction of smartphone technology such as web browsing, video streaming, and more. 4G has revolutionized the cost and usage of data as we knew it; with 3G, it was very costly. 4G gave high speed, high quality, and high capacity at affordable rates to users.
In my blog post at https://pmauthor.com/2019/07/02/3-reasons-why-5g-iot-is-safe/, I also pointed out the key benefits of 5G. The biggest benefit that 5G will give mankind, apart from high data connectivity speeds of up to 35.46 Gbps, is the promise of connecting up to 100 billion devices through its network. The communication protocol that is going to enable this is known as mmWave, or millimeter wave mobile communication. This is the biggest breakthrough for the 5G telecommunication standard. We’ll look in detail at how this works technically.
Now let’s look at the state-of-the-art implementation of machine learning and the IoT in the telecom domain.
HUAWEI: Yes, this is one company in the AI sector to watch out for in spite of the controversy of spying and data breaches that it has been embroiled in. This company was the first one to launch the full mobile 5G chipset. In my blog post on three top machine learning skills for 2019 (https://pmauthor.com/2019/07/05/3-top-machine-learning-skills-for-2019/) and another post covering two reasons why 5G IoT is safe (https://pmauthor.com/2019/07/02/3-reasons-why-5g-iot-is-safe/), I say that 5G IoT is going to be the next game changer in the world of AI. Huawei has launched a full capability 5G chipset. It has also launched the world’s first data center switch with an AI brain. Huawei has also launched full-stack, all-scenario AI technologies to enable an autonomous driving network and developed the SoftCOM AI solution to help mobile operators maximize energy efficiency, network performance, O&M efficiency, and user experiences. Huawei's solution focuses on building a digital village, as mentioned in the article at www.busiweek.com/huawei-presents-its-simplified-5g-and-softcom-ai-solutions/. IoT will also enable its applications to do 8K high-definition (HD) live streaming of audio and video between devices. Real-time virtual healthcare doctors will be able to stream both ways with a patient in critical condition from home, and employees can live stream a business meetings with much clarity and low feedback along with transfer of high speed data. With the launch of 5G services in every country, it’s now possible to provide IoT cloud services, personal mobile services, five-star premium home broadband, and cloud-network convergence. Although Huawei does not describe much about its OceanConnect IoT platform on its website (www.huawei.com/minisite/iot/en/overview.html), it is not difficult to see the key areas of IoT that it sees developing in society. See Figure 1 on the aforementioned webpage for Huawei's IoT solution.
5G and the IoT are destined to take off together since the technology that the 5G telecom spectrum provides not only compliments the capabilities of IoT but also enhances them by providing very fast and efficient data transfers, which are the backbone of IoT devices and applications in every country where it is implemented.
The OceanConnect IoT platform offers to integrate the IoT at all levels of human life. The base of the OceanConnect IoT platform is the 5G chipset or 5G devices which have their own cloud internetworking gateway on top of 2G for 3G and other IoT and 5G connectivity infrastructure networks, combined with a smart home gateway. On top of this layer is the OceanConnect IoT platform, which has three major components inside it. The first is IoT connection management wherein it sits on top of the 2G, 3G, 4G, or 5G infrastructure that has the smart home IoT gateway sitting in the individual houses of people. The connection management provides access to the OceanConnect IoT platform to these individual smart home gateways. Each of these smart home gateways will have enabled IoT devices and the IoT platform will provide device management services such as adding a device to the IoT platform or removing the device from the IoT platform period even if you have the best infrastructure of a smart gateway combined with connection management and device management services on the IoT platform; however, if you do not have an application enablement capability on the platform, it will not have any practical use. It is the application that actually gives the benefit to the user and provides a stable 5G connection for devices. So this IoT platform provides application enablement services as well. Huawei looks at the commercialization of its IoT platform; it sees applications in public utility services such as energy, water, and transportation. It also sees industrial applications where industry 4.0 makes IIoT applications to adapt to its IoT platform and utilizes it to provide smart industrial solutions. It also looks at creating a smart home that is connected to its IoT platform and gets various services such as healthcare, grocery management, transportation management, financial monitoring and management. The OceanConnect IoT platform talks about an Internet of Vehicles, which is a network of small vehicles with various IoT sensors in them that are capable of communicating with each other by sending data between them. Such a network of Internet of Vehicles will definitely help make traffic management easier and more real-time. However, in my opinion, the future will also lead to the creation of something known as the Internet of Devices, which will be the network of interconnected devices, not just limited to vehicles but to anything that uses IoT sensors, which are capable of getting onto any IoT platform with unique applications. It is this Internet of Devices, which is going to enable the smart life revolution period.
Another state-of-the-art implementation for IoT is the open connectivity foundation for all IoT connectivity. The problem that IoT is currently facing, which is going to become a major problem in the future, is that of common IoT standards in society. Without an IoT common standard, there will be various IoT devices and applications that will be isolated and not have the ability to talk to other IoT devices outside themselves. Nevertheless, there is a standard to do so now. There are three sets of people who are involved in the setting of standards for society. The first is the person developing the IoT application. The second is the set of consumers for whom the business is developing the IoT application. The third set is the developer or the creator of the IoT application who brings the IoT to life. This open connectivity foundation for IoT standards bridges the gap between all three. The key advantage is that there is a common definition of security interoperability and a common standard platform and an open source implementation and certification of IoT devices and applications. There are a lot of new models using this framework and they are getting released every month, such as smart pantry IoT devices, monitoring applications, and innovative applications such as smart doghouses, which the OCF currently supports. The complete ecosystem is being created by this Open IoT standard in order to make the IoT more secure and trustworthy.
Let’s move on to the next section regarding a state-of-the-art implementation of machine learning and the IoT in the energy domain. We’ll look at the key technological advancements that are enabling the energy sector to grow.
State-of-the-Art Implementation of Machine Learning and the IoT in the Energy Domain
The energy domain is going through a rapid technological wave right now all over the world. The heart of this change is the introduction of the IoT and machine learning for energy systems around the world. However, the adoption of the IoT and machine learning is not as straightforward as it is in the telecom sector. This is not just about creating new spectrum fields and using them. The energy sector has traditionally been a capital-intensive industry, be it oil, gas, hydro power, or even renewable energy sources such as solar or wind. This aspect of the energy domain makes it a slow adopter of technology such as the IoT and machine learning. The real reason is that most of the energy companies employ a lot of energy equipment and plants cost billions of dollars to commission and install.
The Current State of the IoT in the Energy Domain
The real challenge faced in this sector is about using the IoT and turning on old energy systems. It is easy to implement on energy equipment that has the capability to communicate using protocols such as Modbus interfaces. Very old equipment in the energy domain that does not have the capability to communicate using common industrial interfaces is definitely a big challenge for any energy company. It’s not easy to replace some parts of a working mechanical plant in production because these energy systems can’t have downtime. In such cases where the replacement of plant equipment cannot happen easily, the IoT and machine learning will have to use innovative ways in order to read data from such legacy equipment. After all, machine learning needs data in order to analyze and predict energy sector problems.
Solutions for Embracing the IoT in the Energy Domain
- 1.
Renewable energy plants are rising. This is happening in two ways. One is by old fossil fuel energy generation plants being replaced by solar or wind energy firms. The second is by the new energy plants being added to the grades which are renewable in nature.
Decentralizing of energy operations is happening and this means that the power plant units are becoming more and more independent in order to generate energy. The old way of making energy, like oil and gas, required a centralized plant facility that would help create them. However, renewable energy generation systems like solar grids and wind grades allow the decentralization of energy generation. Decentralization allows the plans for energy generation to be operated in a small facility and also to make independent decisions more conveniently for localized community energy needs. Centralization gives the energy companies the freedom to spread their operations across geographies and use technology with the IoT and machine learning in order to generate energy. Decentralization does bring challenges for making use of economies of scale; however, it is compensated by the ability to make fast and localized decisions on energy generation and the flexibility that it brings along with it. The localized small energy generation unit can make decisions very quickly on when to cut the power generation or up the power generation based on the consumption pattern it observes using techniques like machine learning or deep learning. The ability to adjust the energy needs becomes the key differentiating factor for establishing decentralized energy generation units for the future.
- 2.
We all know that the energy needs of humans, especially oil and gas, are increasing around the world and the power plants are being forced to perform outside of their original build specifications. This is a serious threat for the future as the capacity for the plans has already been breached and there is a stress on their operations for the future. There are various techniques that companies are adopting; some are using renewal energy plants beside the fossil fuel plants whereas others are increasing the capacity of existing fossil fuel plants.
- 3.
A problem that most countries face is that of stabilization of the energy power grid and for that they need to have a good system that is able to predict energy needs well in advance for any part of the grade so that they can increase or decrease the energy generation before a spike or a bust happens. Blackouts can create problems for industries and citizens who are dependent on energy for their operations. So most countries are planning operations that will help them be more productive regarding the energy generation requirements of their communities. Having a stable grid for energy generation reduces the cost by providing a stable power economy versus if there is an unstable grade; the latter worsens the cost of energy generation and has a deeper economic impact on the community.
Siemens, a company that deals with the power plant and energy generation technologies, has conducted a survey of its clients on futureproofing the power plant with its energy domain. See https://assets.new.siemens.com/siemens/assets/public.1552332517.656a83b0-646a-4286-9546-dc54bd8a2e35.future-proofing-power-plant-ebook-2019.pdf. Customers in the survey talk about the future utility challenges that power plants anticipate facing in the next three years. The second problem mentioned by 30% of the respondents concerns adopting new AI technologies. This shows that the energy generation companies see challenges around adopting artificial intelligence machine learning and the IoT in their current environment. The outlook given in the survey was that of next three years, so it’s a short-term outlook. In the question on success through planning and data analytics, a field that is closely associated with data science and machine learning, it asked its customers how successful plant modification and update projects have been in the past three years; the responses from the companies that are delivering plant operational objectives show that over 50% of them met this target. However, the major point that comes out of the survey is that the energy generators are investing heavily in planned upgrades and modifications including new digital technologies, but the confidence levels in deciding on modification and upgrades stands around 20 to 27% for prioritizing it and 203% for deciding which partners should be involved in such an upgrade project and between 18 to 33% for deciding when the investment should be made or at what point of time that investment should be made. The overall level of success reported by the upgrade projects in the past three years is around 25%, which means that although upgrade projects have been adopted by the energy generation companies, they have met with very little success and there is a high failure rate.
So we have seen that the energy sector is in a state of flux, with many companies being forced to upgrade their plans to more renewable energy sources like solar and wind; however, the success rate of such project upgrades is very low, leading to depletion in the return on investments. The adoption of technologies like machine learning and artificial intelligence is a challenge, as pointed out in the survey, and this means that the sector is less prepared to adapt and use these technologies to its advantage. But in any state-of-the-art implementation, a company like Siemens provides solutions by using IoT sensors and devices and connecting them to machine learning models in order to predict the energy demands in any grid. This innovative solution can be applied even to fossil fuel energy plants and to old data stores to analyze and synthesize data to create prediction models. From predicting the periodic maintenance cycles of plants and equipment to predicting energy spikes, the machine learning models are being used by companies in order to solve legacy problems (https://assets.new.siemens.com/siemens/assets/public.1534921431.05f93d5a3c096441998512706a42840c51fd3f68.2018-05-ew-article-iot-manfred-unterweger-en.pdf).
- 1.
Complex operations distributed over the border involving exploration, generation, and distribution of energy requires the use of machine learning so that the data collected from all of these operations is analysed carefully by creating productive business models to aid the decision-making process of the energy sector management executives.
- 2.
The high number and spread of customers pose challenges for energy companies in collecting data from the end user. The use of smart metering, which has the ability to communicate back data from individual customer devices such as smartgrid smartcloud storage applications and the use of machine learning to predict energy consumption based on this data can help the decision-making process in the energy sector.
- 3.
Investments in infrastructures require major equipment and facilities and fleet upgrades or new plant commissioning and installation. The use of machine learning and AI for creating models before making such investments can greatly help in making the right decisions by providing data on what failed in the past and how to avoid the problem areas while implementing new upgrade or installation projects.
To improve operational performance, energy generation plants are using the IoT for data by connecting their industrial assets to IoT sensor devices and monitoring the performance of machinery at the plant level and by measuring organizational KPI such as machinery productivity and equipment efficiency. Predictive maintenance of old plants is definitely a problem that IoT-based applications are solving for the energy sector. IoT data can be generated such that it makes them aware of the equipment performance payloads and locations of deteriorating energy assets. IoT-based pipeline energy meters have become very common; they use machine learning for measuring and predicting the flow of energy liquids like oil and gas. The quality of the flow can also be monitored and any deterioration can be predicted well in advance to avoid any huge losses during the production process. The energy sector is also using machine learning to preidct energy consumption in the fossil fuel plant so that it is in ready state to accommodate any spike in consumer energy requirement by having a standby renewable energy plant power up to take up the load. This balancing based on a predictive demand forecasting of energy requirements is an excellent use case for the energy sector.
You have looked at the state-of-the-art implementation of machine learning in the energy sector. The next section in this chapter covers the state-of-the-art implementation of machine learning and the IoT in the agriculture domain.
State-of-the-Art Implementation of Machine Learning and the IoT in the Agriculture Domain
The agriculture domain is seeing the emergence of an AI revolution right now. This technological wave is happening all over the world. The key transformation that is happening worldwide in farming is that major food conglomerates are trying to take over farming for their own consumption in producing products for the end consumer. The bigger food conglomerates want the farmers to produce specific patented varieties of crops in order to maintain good food quality and productivity. So in some countries they are tying up with farming communities and buying raw vegetables and beef from them. In other countries, they are hiring farmers on contract to produce specific varieties of food grains and other products such as potatoes, lettuce, carrots, dairy, and rapeseed, which are used as inputs for end consumer products such as burgers by McDonalds through the Flagship Farmers program (www.flagshipfarmers.com/en/about-the-program/).
The effort by these food giants is to create a bank of patented varieties of their core materials such as potatoes and lettuce. For example, McDonalds uses its famous patented Russet Burbank potato for its french fries (www.earthandtablelawreporter.com/2015/09/11/patenting-the-potato-not-all-taters-are-created-equal/). Another example is that of Monsanto, a food conglomerate that owns many patents for its vegetables and seeds, such as beans, broccoli, carrots, cucumber, and melons, among others. Many of these vegetables are commercially used by other companies. Monsanto’s potatoes are used by PepsiCo for its flagship potato chip brand Lays (https://feast.media/food-brands-owned-by-monsanto). Monsanto holds 14 varieties of patents and the most popular is Roundup Ready® Corn 2 (www.monsantotechnology.com/content/genuity-traits-corn.aspx).
This is just an example of how food companies are buying a patent and utilizing it commercially. There is currently a controversy surrounding the use of GMO products as raw material for commercial foods. Genetically modified foods can cause various diseases, so they are not being used by major food companies; however, there is pressure on them to become commercially viable and to adapt to GMO-based practices to increase profits, which GMO food varieties offer.
The bold new alternative to genetically modified crops is computationally bred crop technologies combined with AI and machine learning to give excellent yields to farmers. The first one uses the IoT to gather data about the crop and its environment, such as the components of air using air filter sensors and oil sensors to gather data about the soil mix. The applications are already available through something known as statistical breeding patterns, which allow the farmer to get the best crop for the environment based on the atmosphere comprising of the air and soil mix. In this system, the farmer uses intelligent machine learning and IoT systems to select the best seed breed for the given air and soil mix based on past data in other regions of the world that have similar climates in order to produce maximum yields for their farms. This is being done by a company known as HFG; it is using farming data sets and predictive data science to build a predictive breeding platform for crops. (www.zdnet.com/article/computational-breeding-can-ai-offer-an-alternative-to-genetically-modified-crops/).
Drone robot for detecting disease onset in crops
A drone named 91108 rich's was developed by a startup company. Its main function is to disperse pesticides and fertilizers on crops. It’s able to do in 15 minutes work that would take a farmer 60 minutes. Control of such a drone through an iPad merged with a machine learning application to identify areas that require more fertilizer or pesticides can be developed on top of such an application. The diagnostic drones can detect onset of disease by continuous play Selma the crops and protecting any microorganisms near The Cross set on it pacific Best Buy all those areas are such an infestation occur. This is the where the use of IoT sensors on pest control along with machine learning applications running from nearby Raspberry Pi application station can help to automate the entire farming cycle (www.techrepublic.com/article/how-drones-are-changing-farming-in-rural-japan/?ftag=CMG-01-10aaa1b).
Summary
All of the applications you saw in this chapter are experimental in nature and have been newly introduced with the state-of-the-art implementation in the IoT of 5G by the Chinese company Huawei. You also saw how the challenges faced by energy sector are being resolved by the use of the IoT in areas of power generation both for fossil fuels and renewable energy sources. You then looked at the farming sector where the intelligent use of the IoT and machine learning promises to solve the human drain problem of farmers selling their land and migrating to the cities for a better living. In Chapter 5, you will prepare your setup for implementing case studies in these three domains.