AI Applied to the Circular Economy: An Approach in the Wastewater Sector
In recent years, technological advances have opened up several opportunities. One of the most significant advances has been artificial intelligence (AI), which has made it possible to develop systems capable of performing functions associated with learning, reasoning, and problem-solving. Technological development, together with the ability to analyze and process information, has had positive impacts on the economic, social, and environmental spheres.
This paper aims to answer the following question: How can artificial intelligence be used to optimize business management and boost demand for the resources generated in the water reuse sector?
The use of AI, oriented to the commercialization of the resources that the water reuse sector is capable of generating, allows for the adjustment of supply to demand in order to incentivize their use and achieve a more sustainable society. This study evaluates the variables to be taken into account and their interactions in the generation of the price of the resources generated, in order to incentivize their demand. This last aspect has hardly been developed in the water reuse sector. Taking advantage of the potential of AI to direct the waste generated to other industrial sectors might make it possible to boost reuse and extend the useful life of resources in general. The goal is to ensure the sustainability of these infrastructures within the framework of the economic development provided by the circular economy.
2. Potential for the Use of AI/ML Targeted at Resources Generated in the Wastewater Treatment Sector
To this end, it is important to have forecasting techniques to strengthen the distribution channels that link the supply of these resources with their demand, to ensure that they are used efficiently and sustainably. Distribution and marketing channels are the links between producers and consumers; for the supply of resources, they are defined by the WWTPs, and for the demand for resources, they are defined by the potential users.
In the distribution and marketing channels, the interaction between suppliers and demanders requires a certain balance that guarantees the economic relationship between the parties. The main variables that intervene are the quantities offered of each resource, its availability, the price of the resource offered, and the prices of current substitute goods, in addition to other aspects. Price plays an incentive role in trade, so achieving an equilibrium price implies maximizing utility for the agents involved in the market. Some of the possible applications aimed at optimizing the supply and demand of the resources generated in the urban water cycle are presented below.
2.1. Water Distribution and Commercialization
2.2. Distribution and Commercialization of Nitrogen and Phosphorus
If we look at the current fertilizer market, international reference prices increased throughout 2022, and several quotations reached all-time highs. Fertilizer prices are determined by the interaction of supply and demand. Different causes have generated this relentless price increase—on the supply side, these include (i) high and increasing energy prices and (ii) trade disruptions and high transportation costs, while on the demand side, they include (iii) high crop prices (and therefore, high affordability).
The use of ML allows multiple variables to be included in order to establish a dynamic price and synchronize aspects such as daily quotations of industrial fertilizers, consequently allowing us to adjust the price of natural fertilizers produced in WWTPs in order to optimize their distribution at the farm level. This information can be defined at earlier stages of the wastewater treatment process. Knowing the characteristics and volume of water entering the plant makes it possible to project the quantities of fertilizers that the WWTPs can offer to the market, making it possible to plan and supply the demand to them while minimizing the impacts generated by the extraction and subsequent transformation of industrial fertilizers.
2.3. Sludge Distribution and Commercialization
In this case, the use of ML makes it possible to optimize the sludge distribution channel. There are different variables that can be monitored to automate shipments, adjusting quantities and qualities to the agricultural terrain. In addition, geographic localization can be of great help to optimize transport routes, minimizing total transport times and, consequently, economic costs and associated emissions. Finally, the economic prices associated with sludge EUR/tn.) can vary and adapt to market conditions, so a higher demand for sludge can generate slight price increases and vice versa, among other aspects.
2.4. Energy Distribution and Commercialization
In conclusion, AI/ML allows us to optimize decisions in the distribution and commercialization channels. It manages to automate decisions by considering a multitude of dynamic parameters that can be processed and evaluated almost instantaneously, in order to adapt the decisions or prices of the resources generated to the market. For this, in addition to the variables that define the market, it is necessary to have data relating to certain technical aspects, as well as the history of the plant in terms of flows treated and resources generated. This last aspect allows us to reduce uncertainty because we are able to make forecasts of both production and demand for the resources generated.
3. Machine Learning Applied to the Wastewater Treatment Sector
The wastewater treatment sector has great potential in terms of resource generation. It is considered a strategic sector due to the large amount of resources it produces, and it guarantees economic development within the framework of sustainability and the circular economy. The use of methodologies that monitor the production of the resources generated in the WWTP allows, beyond the management of processes, for the optimization of marketing and distribution channels, adjusting the demand and supply of these resources. In this last aspect, the design of appropriate tariffs will make it possible to incentivize the demand for these resources. In terms of their design, there are numerous aspects involved, which can be divided into internal factors—costs intrinsic to their production—and external factors—unique characteristics of the market that affect their demand. A constantly changing environment implies a continuous adaptation of the prices offered for the different resources that these infrastructures are capable of generating.
In a competitive market, consumers can choose between different goods/products to satisfy their needs. The concept of substitute goods refers to goods that compete with each other because they can perform the same function or service. For example, in the case of the water resource, it is common for both industrial and agricultural areas to resort to conventional fresh water supply, either through installations carried out by the drinking water supply company or through the direct extraction of water from aquifers, rivers, etc. However, WWTPs are capable of providing water in sufficient quantity and quality, adapting to the requirements of the industrial and agricultural sectors. In this example, there are two assets that can be used in the same way; one of them is conventional (limited natural resource), and the other is non-conventional (through regeneration). With regard to the consumption of water from the WWTP (non-conventional resource), it may depend on characteristics related to geographical areas. For example, areas with high water stress will be encouraged to use non-conventional water due to the possible limitations posed by the conventional resource. Between the two most polarized scenarios (situations of water stress and water surplus), there are multiple scenarios in which factors such as the price and availability of the resource will play a determining role in the final choice of the resource.
Similarly, in the agricultural sector, sludge and fertilizers are used with the aim of increasing crop productivity and improving soil characteristics. With regard to the use of these resources, WWTPs are capable of generating both manure and fertilizers (phosphorus and nitrogen). However, in the latter case, it is common to use industrial fertilizers, the availability of which is becoming increasingly limited. In a market with scarce resources, prices can suffer upward variations and, as in the case of fertilizers, costs have been rising in recent years, putting the food industry at serious risk. In this scenario, fertilizers obtained through sustainable processes, such as their recovery from wastewater, guarantee the long-term sustainability of the sector. However, the price of this resource may be affected by the prices of industrial fertilizers (substitute goods), so a shortage in the production of the latter might increase the consumption of fertilizers produced in WWTPs.
Finally, energy represents another strategic resource and, likewise, there is a range of alternatives with which to guarantee energy demand, whether domestic or industrial. In recent years, the prices associated with energy have experienced continuous growth, so industries that depend heavily on energy to carry out their production processes have seen an alarming increase in production costs. Furthermore, it is important to point out that the source of the energy can generate high environmental impacts and pose certain risks, mainly due to the process of obtaining it. In this scenario, having a clean, alternative source of energy, such as the WWTP, might guarantee the sustainability of nearby industries and agricultural fields (pumping water for irrigation) or the supply of this infrastructure, avoiding the consumption of other external energy sources.
3.1. Empirical Case: Applying ML to the Wastewater Treatment Sector
An important knowledge structure that can result from data mining activities is decision trees. They can be used for the classification of future events, and these events can be defined by both qualitative and quantitative criteria. The present work applies the decision tree methodology with the objective of integrating a large number of variables that can influence price.
The set of statistical characteristics can come from history, e.g., reclaimed water prices in recent years; these can be defined by infrastructure characteristics and specific market conditions in which the company operates.
The decision tree has leaf nodes, which represent class labels, and other nodes, which are associated with the classes (magnitude level in this case) being analyzed.
The branches of the tree represent each possible value of the parameter node from which they originate.
The decision tree can be used to express the structural information present in the data by starting at the root of the tree (the highest node) and moving through a branch to a leaf node.
The level of contribution of each individual parameter is given by a statistical measure within the parentheses in the decision tree. The first number in parentheses indicates the number of data points that can be classified using that set of parameters. The parameters appearing in the decision tree nodes are in descending order of importance.
At each decision node, the most useful parameter for classification can be selected using the appropriate estimation criteria. The criterion used to identify the best parameter invokes the concept of entropy and information gain, which is discussed in detail in the following subsections. The decision tree algorithm has two phases: construction and pruning. The construction phase is also known as the “growth phase”.
Algorithms require data for their operation; in fact, a greater availability of all historical data allows us to calibrate the model, thus improving the results. In addition, they allow us to include both categorical and continuous variables, so that we can enrich the results with dichotomous variables that respond to more qualitative aspects of the processes. Discretizing allows continuous variables to be converted into categories or ranges. This helps to simplify the rules that the decision tree uses to make predictions. Discretization makes decision trees more effective, interpretable, and efficient, which facilitates data-driven decision-making. The target variable (increase the price of unconventional water) can be described as a discrete variable, taking values between 1 and 10.
The variables that can influence the selling price of the example resource (non-conventional water) are then identified and discretized. The variables are divided into internal and external factors; this is due to the nature of the characteristics, with some depending on the internal functioning of the infrastructure and others on market-related aspects.
3.1.1. Internal Factors
The internal factors are defined by all variables related to the management and production of the infrastructure. First of all, in order to estimate market prices, it is necessary to know both the investment costs (CAPEX) and the treatment and operating costs (OPEX) of all the processes involved in obtaining the resources (reclaimed water, sludge, fertilizers, energy, etc.). In this case, we take as an example the non-conventional water that WWTPs can offer for other uses (industrial or agricultural). It is important to guarantee the recovery of all costs to ensure the sustainability of the process.
Capital and Operating Expenditures
Quality of the Influent (Organic Load)
Another consequence directly related to the quality of the incoming wastewater is the amount of resources that can be generated. For example, the amount of fertilizer depends directly on the concentration of nitrogen and phosphorus in the wastewater.
Precipitation in the Area
A decrease in rainfall might imply a reduction in the flow to be treated and therefore less non-conventional water. Reduced availability might justify an increase in the prices offered.
Leaks (Sanitation Network)
Quantities of Reclaimed Water to be Produced:
It is important to know at present the consumption of conventional water in the different sectors, mainly agricultural and industrial. In addition, the quality required for their processes will justify the use of a particular technology. An increase in the quantity demanded, whether of quality A or B, will justify an increase in price (EUR/m3). This is related to the availability of conventional water and the urban, agricultural, and industrial development of the study area. An increase in non-conventional water demand (last 6 months) allows for assigning a 10 to this characteristic. The choice of a six-month period justifies a trend in consumption. This variable is also explained as an external factor because water demand is determined by the users and the study area. In order to avoid duplication, it is quantified as an external factor.
3.1.2. External Factors
Conventional Water Quality (Aquifer, Network, Surface)
Availability of Conventional Water
The quality of service is an aspect that can be detrimental to industries that use the resource in their production process. For example, cardboard, paper, ceramics, and textile companies need to ensure the availability of water in their manufacturing processes. Thus, if the area suffers from water shortages, the WWTP can guarantee the continuous service of reclaimed water in order to meet the sector’s demand.
In the case of Spain, each municipality designs and sets its own freshwater tariffs. This may cause some differences depending on the geographical area assessed. We understand that non-conventional water is a substitute for freshwater, which can be used for both agricultural and industrial purposes. In this sense, users will normally opt for the resource (water) that offers the lowest price. In this case, the price of freshwater can be a restriction for the consumption of reclaimed water. Therefore, the proposed model must monitor the current freshwater prices to set a maximum reference value for both qualities (A and B). Beyond the maximum price constraint, the water situation in the Mediterranean area may also lead to continuous increases in the price of freshwater. In this situation, an increase in the price of the conventional resource should imply an increase in the non-conventional water resource offered to maximize the profits of the wastewater treatment plants. In the latter case, the variable takes dichotomous values; in the case of an increase in the price of freshwater, the discretized variable takes the value 10.
This work proposes to direct the potential of methods based on artificial intelligence (AI) toward aspects related to the supply and demand of resources generated in wastewater treatment plants (WWTPs). In the context of the circular economy, these infrastructures play a relevant role as they are able to produce, in addition to non-conventional water, fertilizers and electricity, among other resources. In this sense, WWTPs can be considered a source of production of non-conventional resources, with the capacity to satisfy different needs of productive sectors, such as agriculture and industry. Therefore, they are another economic actor in the market capable of competing with the current supply of resources from non-renewable sources. In this vein, in order to bring economic sustainability to the wastewater treatment sector, methodologies are needed to help operators optimize the tariffs offered for the resources produced, adapting prices to market conditions.
At present, the prices of these resources, whether reclaimed water or fertilizers, respond to fixed tariffs based solely on the costs incurred in the production processes. Setting sales prices (tariffs) solely on the basis of the costs incurred in the process can reduce the profits earned by the operators of these infrastructures. This is due to the fact that in the market there are a series of fluctuations that can affect the availability of natural resources, generating situations of scarcity that directly affect prices, increasing them. This situation generates a series of opportunities that can be exploited by the wastewater treatment sector. Therefore, new approaches are needed to monitor the market conditions and to adapt the price (supply) to the characteristics of the existing demand. In short, maximizing the profits generated will ultimately ensure the economic sustainability of these facilities.
The paper presents a practical application that uses a machine learning (ML) methodology to monitor and evaluate in real time the change in different variables, providing a dynamic solution in terms of tariff-setting for the resources generated in the wastewater treatment sector. The demonstration includes an empirical case that specifically addresses the marketing and distribution of reclaimed water. It identifies the variables that may influence the final price of the resource and classifies the possible changes they may undergo, addressing both aspects related to the infrastructure itself and those related to the market in which it operates. For ease of understanding, the result is expressed on a scale of 1 to 10, where 10 indicates an increase in the price of the resource. The results obtained include changes related to the availability and scarcity of natural resources, which are identified and taken into account in order to optimize the tariffs of unconventional resources produced in these facilities. This dynamic requires the use of methodologies capable of monitoring and evaluating in real time the impact of multiple variables on price generation.
The possibility of automating in real time the variables that influence prices represents a breakthrough in terms of the circular economy; beyond the relevance of the water resource, it is possible to supply society with other resources that are strategic for economic development. However, this approach can also have certain limitations: first of all, it is necessary to automate the process of obtaining data, both technical and economic. In this sense, the provision of results in real time depends on the updating process. Secondly, some external factors, such as the availability of the conventional resource to be evaluated, suggest the definition and discretization of the possible results, so some variables may be subject to the environment in which they operate and require a certain consensus that represents the reality of the market. Finally, there are certain social barriers that must be taken into account. The demand for these resources may represent a certain rejection due to the fact that they do not come from conventional sources.
This paper concludes by highlighting the strategic role that wastewater treatment plants play in the production of unconventional resources. The dynamic circumstances of the market generate commercial opportunities that operators can take advantage of. Having tools with which to update prices based on market characteristics makes it possible to maximize the profits obtained, thus ensuring the sustainability of the wastewater sector within the framework of the circular economy. The analysis of the different possibilities offered by the use of ML in the marketing channel serves as an illustrative and stimulating example for several companies in the sector, offering a point of view that can enrich the possibilities offered by AI.
Disasters Expo USA, is proud to be supported by Inergency for their next upcoming edition on March 6th & 7th 2024!
The leading event mitigating the world’s most costly disasters is returning to the Miami Beach
And in case you missed it, here is our ultimate road trip playlist is the perfect mix of podcasts, and hidden gems that will keep you energized for the entire journey-