Pyrolysis plant and operating method thereof

Abstract

The present invention describes a pyrolysis plant for the pyrolysis of industrial and urban organic waste including apparatus such as means for receiving and conditioning waste, pyrolysis reactors, combustion chambers for supplying thermal energy, a rectification tower, a gas condensation apparatus, a non-condensed gas scrubbing tower, cooling apparatus, and tanks for oil and wastewater. The plant is provided with a control system with sensors to monitor input and output variables, and mechanisms to regulate the operation of the apparatus. It integrates two analysers based on artificial neural networks: a first artificial neural network adjusts the temperature of the combustion chambers, and a second artificial neural network controls the temperature of the rectification towers, working in tandem to optimise the energy efficiency and performance of the process.

Description

[0001] DESCRIPTION

[0002] Pyrolysis Plant and Operating Procedure

[0003] TECHNICAL SECTOR

[0004] The present invention falls within the field of devices, systems and procedures for pyrolytic oils from the valorization of industrial and urban plastic waste.

[0005] BACKGROUND OF THE INVENTION

[0006] Pyrolysis plants represent an advanced technological solution for the treatment and valorization of waste, transforming it into high-value-added products through a thermochemical process. Pyrolysis, which takes place in the absence of oxygen, breaks down complex organic materials into simpler fractions, generating primarily three products: combustible gases (such as hydrogen and methane), liquids like bio-oil, and carbon-rich solid residues known as biochar or coke.

[0007] These plants are especially relevant for managing plastic waste, used tires, and biomass—sectors where traditional disposal methods, such as landfills or incineration, generate significant environmental problems. By reducing reliance on landfills and decreasing emissions associated with combustion, pyrolysis positions itself as a sustainable and efficient alternative within the circular economy.

[0008] Furthermore, pyrolysis plants offer economic benefits by recovering materials and generating energy, contributing to a cleaner development model. However, their implementation requires addressing technical, regulatory, and economic challenges to maximize their potential in integrated waste management.

[0009] Currently, advanced control systems are essential for the efficient operation of pyrolysis plants, especially when real-time process modeling is required. However, real-time control of the various units that make up a pyrolysis plant presents certain complexities that conventional methods, such as temperature control using PLC (Programmable Logic Controller) systems, can hardly manage optimally due to the nature of the process and the heterogeneity of the input materials. It is in this context that artificial intelligence methods demonstrate their ability to overcome the limitations of traditional approaches, offering more adaptive and effective solutions.

[0010] In this context, artificial neural networks have acquired increasing importance in recent years in the field of Chemical Engineering, where they have been widely applied to optimize various processes and improve decision-making.

[0011] Essentially, a neural network is composed of a set of interconnected artificial neurons, or perceptrons. An artificial neuron (AN) is a processing unit that attempts to mimic the behavior of a biological neuron, such as those in the human brain. Each neuron receives multiple inputs, which it combines, usually through addition. This addition is adjusted using a transfer function, and the resulting value is sent directly as the output of the processing element.

[0012] In a neural network (NN), each input is multiplied by its weight, similar to the synaptic connection between biological neurons. The weighted inputs are summed, and the neuron's activation level is calculated. The NN's output connects to other neurons via weighted connections. In a typical NN, neurons are organized in layers, with connections between adjacent layers to form a sequential structure.

[0013] There are two layers that connect to the external environment: the input layer, which acts as a buffer where data is presented to the network, and the output layer, which stores the network's response to that input. The other layers are known as hidden layers. In the development of an artificial neural network (ANN), the network learns the rules for processing information by adjusting the weights associated with the connections between the neurons in its various layers. The goal of training an ANN is to ensure that, for a specific set of inputs, the network generates a set of outputs that closely approximate the expected results or, at least, are consistently acceptable. This training process is based on the sequential application of various datasets or input vectors, adjusting the interconnection weights according to a predetermined learning algorithm.During training, the weights are progressively updated, gradually converging towards optimal values ​​that allow the network to produce the desired outputs for each input provided.

[0014] Artificial neural network training algorithms are divided into two categories: supervised and unsupervised. In supervised training, each input vector is paired with its desired output vector. The process involves presenting an input vector to the network, calculating its output, comparing it to the expected output, and using the difference (error) to adjust the connection weights. This adjustment follows an algorithm that seeks to minimize the error. The vector pairs in the training set are processed sequentially and iteratively. Weight adjustment continues until the error across the entire training set is sufficiently small and acceptable.

[0015] A key feature of artificial neural networks is that they learn their own rules from the examples provided during the training phase. This learning is achieved through a rule that adjusts the connection weights based on the input examples and, optionally, the desired outputs.

[0016] Another important aspect of neural networks is how they store information. Their memory is distributed across the weights of all the connections in the network. Furthermore, some neural networks are "associative": they can recognize partial inputs, search for the most similar one stored in their memory, and generate the output corresponding to the complete input. Backpropagation networks are especially useful in this technical field. These networks use a supervised training method, in which pairs of patterns are presented: an input pattern and its corresponding desired output. With each presentation, the weights are adjusted to reduce the error between the desired output and the network's response. The backpropagation algorithm includes a forward propagation phase and a backward propagation phase, which are performed for each pattern during training.

[0017] The forward propagation phase begins when a pattern is present in the input layer. Each input unit takes the corresponding value from the pattern, and the output level of the first layer is calculated. The remaining layers then perform forward propagation to determine their respective activation levels. Once this phase is complete, the correction or backward propagation begins. Calculations to adjust the connection weights start in the output layer and work backward through all layers to the input layer. The weight adjustments are divided into two groups: those corresponding to the units in the output layer and those corresponding to the units in the hidden layers.

[0018] Adjusting the weights in the output layer is straightforward, as the desired value for each unit is known. Each unit generates an actual value, which is compared to the expected value in the training set pattern.

[0019] Regarding the adjustment of hidden layer weights, these layers do not have a vector of desired outputs, so it is not possible to apply the error propagation method used in the output layer processing units. The error value calculated for these types of units is obtained from mathematical equations, which are not detailed in this description, that allow the definition of a variable q, known as the learning coefficient. This coefficient, generally between 0.25 and 0.75, reflects the degree of learning of the network.

[0020] The present invention relates to a pyrolysis plant equipped with a control system based on the combined use of artificial neural networks. Once trained with the process input and output variables, this system is capable of real-time temperature control in critical plant units, such as the combustion chamber that supplies thermal energy to the reactor. This ensures optimized plant resource utilization at all times during operation.

[0021] EXPLANATION OF THE INVENTION

[0022] The present invention, in its first aspect, presents a pyrolysis plant for industrial and municipal plastic waste, comprising at least the following equipment: means for receiving and conditioning the waste, at least one pyrolysis reactor, at least one combustion chamber configured to supply thermal energy to the pyrolysis reactor(s), at least one rectification tower, at least one gas condensation unit, at least one non-condensed gas scrubbing tower, at least one indirect contact cooling unit, at least one tank configured for receiving the light and heavy oil fractions, and at least one tank for receiving wastewater, wherein the plant has a control system characterized in that it comprises:

[0023] - at least one installation with sensors and measuring instruments associated with the plant equipment so that they detect the values ​​of the input and output variables of said equipment;

[0024] - at least one facility for regulating or controlling the operation of the plant's equipment;

[0025] - a first analyzer of the temperature of the combustion chamber(s) (T-06) in the form of a first artificial neural network (ANN1), so that it emits control magnitudes from the measured values ​​to the plant's equipment regulation and control facilities to adjust T-06 to a temperature lower than 670°C;

[0026] - a second temperature analyzer of the rectification tower(s) (T-01) in the form of a second artificial neural network (ANN2), so that it emits control magnitudes from the measured values ​​to the plant's equipment regulation and control facilities to adjust T-01 to a temperature lower than 315°C;

[0027] where the second analyzer is activated to operate in tandem with the first analyzer when T-06 exceeds a threshold value (Tu).

[0028] In its second aspect, the invention relates to a pyrolysis process for industrial and urban plastic waste using a pyrolysis plant according to the first aspect of the invention, which comprises the operational stages of waste treatment, waste pyrolysis, and condensation of the reactor outlet gases, characterized in that the temperature of the combustion chamber(s) (T-06) and the temperature of the rectification tower(s) (T-01) are controlled by artificial neural networks (ANNs), and said process comprises:

[0029] - acquire input variable values ​​from at least one operational stage of the process using sensors and measuring instruments;

[0030] - record values ​​of temperature T-06 and temperature T-01;

[0031] - present to the RNA1 the history of the values ​​of the input variables and the temperature T-06, as an output variable, and train them by adjusting the weights of the neuron connections until they gradually converge towards the values ​​that make each value of input variable produce the corresponding theoretical or desired value of temperature T-06;

[0032] - present to the RNA2 the history of the values ​​of the input variables and temperature T-01, as an output variable, and train them by adjusting the weights of the neuron connections until they gradually converge towards the values ​​that make each value of input variable produce the corresponding theoretical or desired value of temperature T-01;

[0033] - execute the ANNs, trained by a learning algorithm, for the estimation of the values ​​of T-01 and T06 from the values ​​of the input variables of at least one operational stage of the ongoing process, where ANN2 is activated to operate in tandem with ANN1 when T-06 exceeds the threshold value (Tu); and

[0034] - readjust the variables of at least one operational stage of the process until the estimated T-06 and T-01 temperature values ​​are adjusted to the theoretical or desired values ​​with an error lower than indicated.

[0035] The pyrolysis processes of municipal and industrial waste are characterized by their nonlinear, multivariable, and non-steady-state nature, primarily due to the heterogeneity of the waste being treated. This characteristic results in a process gas with a variable composition, making it difficult to maintain a constant and suitable temperature in the combustion chamber(s) responsible for supplying thermal energy to the pyrolysis reactor.

[0036] The present invention overcomes the limitations of the prior art by addressing this problem efficiently through an innovative control system. This system employs two artificial neural networks (ANNs) that operate in tandem to control and optimize the process, achieving significant energy savings compared to conventional control systems.

[0037] BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The above and other advantages and features will be more fully understood from the following detailed description of some exemplary embodiments with reference to the accompanying drawings, which are to be considered illustrative and not limiting, in which:

[0039] Figure 1 shows a schematic diagram of the operation of an artificial neural network. In this case, the network consists of four layers of neurons: the input layer, two hidden layers with a specific number of neurons (O1 and O2). n The input layer consists of a set of input neurons (En), whose number corresponds to the number of input variables (Ven). These variables are weighted by the weights of the connections between neurons, and through a transfer function, an output signal from the network (Vs) is generated.

[0040] Figure 2 shows a schematic diagram of the pyrolysis plant of the invention.

[0041] Figure 3 shows a schematic diagram of the control process of the pyrolysis plant of the invention.

[0042] List of references

[0043] H-001 - Feed screw

[0044] H-002 - Feed hopper assembly

[0045] H-003 - Incoming waste conveyor belts

[0046] H-101 - Combustion chamber

[0047] R-001 - Pyrolysis reactor

[0048] K-001 - Rectification Tower 1

[0049] K-002 - Rectification Tower 2

[0050] W-001 - Primary Capacitor

[0051] W-002 - Secondary Capacitor

[0052] W-003 - Tertiary Capacitor

[0053] K-003 - Heavy fraction separator

[0054] K-004 - Condensate Trap

[0055] D-001 - Heavy fraction decanter

[0056] D-002 - Light fraction decanter

[0057] P-002 - External reflux pump

[0058] D-005 - Safety vent treatment

[0059] D-006 - Gas cleaning D-003 - Vacuum system 1

[0060] D-004 - Vacuum System 2

[0061] P-005 - Recirculation Pump

[0062] VE-01 - Venturi

[0063] K-008 - Chimney flue

[0064] H-102 - Afterburner of combustion gases and vents

[0065] PREFERRED EMBODIMENT OF THE INVENTION

[0066] In its first aspect, the invention relates to a pyrolysis plant for industrial and urban plastic waste, comprising at least the following equipment:

[0067] - means for receiving and conditioning the waste,

[0068] - at least one pyrolysis reactor,

[0069] - at least one combustion chamber configured to supply thermal energy to the pyrolysis reactor(s),

[0070] - at least one rectification tower,

[0071] - at least one gas condensation unit,

[0072] - at least one non-condensed gas scrubbing tower,

[0073] - at least one indirect contact cooling unit,

[0074] - at least one tank configured for receiving light and heavy oil fractions,

[0075] - at least one tank for receiving wastewater and

[0076] - at least one afterburner for the combustion gases,

[0077] where the plant has a control system characterized in that it comprises:

[0078] - at least one installation with sensors and measuring instruments associated with the plant equipment so that they detect the values ​​of the input and output variables of said equipment;

[0079] - at least one installation for regulating or controlling the operation of the plant equipment; - a first analyzer of the temperature of the combustion chamber(s) (T-06) in the form of a first artificial neural network (ANN1), so that it emits control magnitudes from the measured values ​​to the plant equipment's regulation and control installations to adjust T-06 to a temperature lower than 670°C;

[0080] - a second temperature analyzer of the rectification tower(s) (T-01) in the form of a second artificial neural network (ANN2), so that it emits control magnitudes from the measured values ​​to the plant equipment's regulation and control facilities to adjust T-01 to a temperature equal to or less than 315°C;

[0081] where the second analyzer is activated to operate in tandem with the first analyzer when T-01 exceeds a threshold value (Tu).

[0082] The term “industrial and municipal plastic waste” refers to discarded polymeric materials from various sources. Industrial plastic waste originates from manufacturing and handling processes in industrial settings, including offcuts, defective products, and industrial packaging. In contrast, municipal plastic waste results from everyday consumption in residential, commercial, and municipal environments, and includes packaging, bags, bottles, and other single-use products.

[0083] In the preferred embodiment of the pyrolysis plant, the plant comprises a PLC system that controls the combustion temperature (T-06) at the start of the pyrolysis process. This system only controls the start of the process and is replaced by the tandem ANN system.

[0084] In the preferred embodiment of the pyrolysis plant, the threshold value (Tu) for the activation of the second analyzer form of a second artificial neural network (ANN2) is 290°C.

[0085] In the preferred embodiment of the pyrolysis plant, it further comprises a reflux system configured to reintroduce a flow of condensed light pyrolysis oils into the gas condensation equipment. This reflux system is activated by the control system when the set temperature is exceeded in the combustion chamber (T-06) or in the rectification columns (T-01), thus enabling precise and optimal temperature control in the rectification column.

[0086] In the preferred embodiment of the pyrolysis plant, the combustion chamber(s) comprise two independent temperature sensors. These sensors are used to obtain a more precise combustion chamber temperature (T-06).

[0087] In the preferred embodiment of the pyrolysis plant, the means for receiving and conditioning the waste comprise at least one sealed hopper where the heat from the reactor's combustion gases is used to pre-dry the waste. The use of at least one sealed hopper allows the combustion gases to be used for pre-drying the waste as it is introduced into the pyrolysis reactor, resulting in significant energy savings.

[0088] In the preferred embodiment of the pyrolysis plant, the plant comprises at least one rectification column integrated with an internal reflux.

[0089] In the preferred embodiment of the pyrolysis plant, at least one gas condensation unit comprises a plurality of condensers, each including a pH sensor configured for chlorine detection. This gas condensation unit, which includes a pH sensor configured for chlorine detection, allows for the identification of chlorine in the pyrolysis products. Since chlorine is an undesirable component, its early detection triggers an alert, facilitating modifications to the process conditions, either by adjusting the incoming waste or correcting any deficiencies in one of the units.

[0090] In its second aspect, the invention relates to a pyrolysis process for industrial and urban plastic waste using a pyrolysis plant according to the first aspect of the invention, which comprises the operational stages of waste treatment, pyrolysis of the waste, and condensation of the reactor outlet gases, characterized in that the temperature of the combustion chamber(s) (T-06) and the temperature of the rectification tower(s) (T-01) are controlled by artificial neural networks (ANNs), and said process comprises:

[0091] - acquire input variable values ​​from at least one operational stage of the process using sensors and measuring instruments;

[0092] - record values ​​of temperature T-06 and temperature T-01;

[0093] - present to the RNA1 the history of the values ​​of the input variables and the temperature T-06, as an output variable, and train them by adjusting the weights of the neuron connections until they gradually converge towards the values ​​that make each value of input variable produce the corresponding theoretical or desired value of temperature T-06;

[0094] - present to the RNA2 the history of the values ​​of the input variables and temperature T-01, as an output variable, and train them by adjusting the weights of the neuron connections until they gradually converge towards the values ​​that make each value of input variable produce the corresponding theoretical or desired value of temperature T-01;

[0095] - execute the ANNs, trained by a learning algorithm, to estimate the values ​​of T-01 and T06 from the values ​​of the input variables of at least one operational stage of the ongoing process, where ANN2 is activated to operate in tandem with ANN1 when T-06 exceeds the threshold value (Tu); and

[0096] - readjust the variables of at least one operational stage of the process until the estimated T-06 and T-01 temperature values ​​are adjusted to the theoretical or desired values ​​with an error lower than indicated.

[0097] In the preferred embodiment of the pyrolysis process, the theoretical value of T-06 is less than 670°C, the theoretical value of T01 is equal to or less than 315°C, and the threshold value (Tu) is 290°C.

[0098] In the preferred embodiment of the pyrolysis process, the input variables of RNA1 comprise:

[0099] Process gas flow rate,

[0100] LPG flow rate,

[0101] Airflow,

[0102] % of O2 in combustion gases,

[0103] % moisture in feed materials,

[0104] % of mineral fillers in food-grade plastics,

[0105] Weight of material fed,

[0106] Total volume of pyrolysis oils generated,

[0107] Total volume of wastewater generated,

[0108] Total volume of process gas generated,

[0109] Current weight of the reactor load,

[0110] Cabal average of the material fed in the last hour,

[0111] Internal reactor temperature, and

[0112] Medium N2 Cabal fed the last 5 min.

[0113] In the preferred embodiment of the pyrolysis process, the input variables of the RNA2 comprise:

[0114] Combustion chamber temperature (T-06)

[0115] % moisture in feed materials,

[0116] Percentage of plastics in food materials,

[0117] % of mineral fillers in food-grade plastics,

[0118] Weight of material fed,

[0119] Total volume of pyrolysis oils generated,

[0120] Total volume of wastewater generated,

[0121] Total volume of process gas generated,

[0122] Current weight of reactor load, Average weight of material fed in the last hour,

[0123] Internal temperature of the reactor,

[0124] External reflux flow rate,

[0125] Light condensate tank temperature, and

[0126] - Process gas flow rate,

[0127] The use of these input variables, in both RNA1 and RNA2, allows for optimal operation of the pyrolysis process without needing to include quality parameters of the resulting products in the control model. This facilitates real-time process measurement and control, using parameters that are relatively easy to obtain in a procedure of this nature.

[0128] In the preferred embodiment of the pyrolysis process, the neural networks are of the backpropagation type with at least four layers comprising one input layer, two hidden layers, and one output layer. In this preferred embodiment, the hidden layers comprise a maximum of 30 perceptrons or neurons. The learning algorithm employed in the neural networks is the Levenberg-Marquardt algorithm.

Claims

CLAIMS 1. A pyrolysis plant for industrial and municipal plastic waste, comprising at least the following equipment: means for receiving and conditioning the waste, at least one pyrolysis reactor, at least one combustion chamber configured to supply thermal energy to the pyrolysis reactor(s), at least one rectification tower, at least one gas condensation unit, at least one non-condensed gas scrubbing tower, at least one indirect contact cooling unit, at least one tank configured for receiving the light and heavy oil fractions, at least one tank for receiving wastewater, and at least one afterburner for the combustion gases, wherein the plant has a control system characterized in that it comprises: - at least one installation with sensors and measuring instruments associated with the plant equipment so that they detect the values ​​of the input and output variables of said equipment; - at least one facility for regulating or controlling the operation of the plant's equipment; - a first analyzer of the temperature of the combustion chamber(s) (T-06) in the form of a first artificial neural network (ANN1), so that it emits control magnitudes from the measured values ​​to the plant's equipment regulation and control facilities to adjust T-06 to a temperature lower than 670°C; - a second temperature analyzer of the rectification tower(s) (T-01) in the form of a second artificial neural network (ANN2), so that it emits control magnitudes from the measured values ​​to the plant equipment's regulation and control facilities to adjust T-01 to a temperature equal to or less than 315°C; where the second analyzer is activated to operate in tandem with the first analyzer when T-01 exceeds a threshold value (Tu).

2. The pyrolysis plant according to claim 1 characterized in that the threshold value (Tu) is 290°C.

3. The pyrolysis plant according to claim 1, characterized in that it further comprises a reflux system configured to reintroduce a flow of condensed light pyrolysis oils into the gas condensation equipment.

4. The pyrolysis plant according to claim 1 characterized in that the combustion chamber(s) comprise within them two independent temperature sensors.

5. The pyrolysis plant according to claim 1 characterized in that the means for receiving and conditioning the waste comprise at least one sealed hopper where the heat from the reactor's combustion gases is used in pre-drying the waste.

6. The pyrolysis plant according to claim 1 characterized in that the plant comprises at least one rectification column integrated with an internal reflux.

7. The pyrolysis plant according to claim 1 characterized in that at least one gas condensation unit comprises a plurality of condensers including a pH sensor configured for chlorine detection.

8. A pyrolysis process for industrial and urban plastic waste using a pyrolysis plant according to any of the preceding claims, comprising the operational stages of waste treatment, waste pyrolysis, and condensation of the reactor outlet gases, characterized in that the temperature of the combustion chamber(s) (T-06) and the temperature of the rectification tower(s) (T-01) are controlled by artificial neural networks (ANNs), and said process comprises: - acquire input variable values ​​from at least one operational stage of the process using sensors and measuring instruments; - record values ​​of temperature T-06 and temperature T-01; - present to the RNA1 the history of the values ​​of the input variables and the temperature T-06, as an output variable, and train them by adjusting the weights of the neuron connections until they gradually converge towards the values ​​that make each value of input variable produce the corresponding theoretical or desired value of temperature T-06; - present to the RNA2 the history of the values ​​of the input variables and temperature T-01, as an output variable, and train them by adjusting the weights of the neuron connections until they gradually converge towards the values ​​that make each value of input variable produce the corresponding theoretical or desired value of temperature T-01; - execute the ANNs, trained by a learning algorithm, to estimate the values ​​of T-01 and T06 from the values ​​of the input variables of at least one operational stage of the ongoing process, where ANN2 is activated to operate in tandem with ANN1 when T-01 exceeds the threshold value (Tu); and - readjust the variables of at least one operational stage of the process until the estimated T-06 and T-01 temperature values ​​are adjusted to the theoretical or desired values ​​with an error lower than indicated.

9. Production process according to the preceding claim, characterized in that the theoretical value of T-06 is less than 670°C, the theoretical value of T01 is equal to or less than 315°C and the threshold value (Tu) is 290°C.

10. Production process according to any of claims 8 to 9, characterized in that the input variables of RNA1 comprise: - Process gas flow rate, - LPG flow rate, - Airflow, - % of O2 in combustion gases, - % moisture in feed materials, - % of mineral fillers in food-grade plastics, - Weight of material fed, - Total volume of pyrolysis oils generated, - Total volume of wastewater generated, - Total volume of process gas generated, - Current weight of the reactor load, - Average Cabal of the material fed in the last hour, - Internal temperature of the reactor (T-09), and - Medium N2 Cabal fed the last 5 min.

11. Production process according to any of claims 8 to 10, characterized in that the input variables of the ANN2 comprise: - Combustion chamber temperature (T-06) - % moisture in feed materials, - % of plastics in food material, - % of mineral fillers in food-grade plastics, - Weight of material fed, - Total volume of pyrolysis oils generated, - Total volume of wastewater generated, - Total volume of process gas generated, - Current weight of the reactor load, - Average Cabal of the material fed in the last hour, - Internal temperature of the reactor, - External reflux flow rate, - Light condensate tank temperature, and - Process gas flow rate,12. Production process according to any of claims 8 to 11, characterized in that the RNAs are of the backpropagation type with at least 4 layers comprising; 1 input layer, 2 hidden layers and 1 output layer.

13. Production process according to the preceding claim, characterized in that the hidden layers comprise a maximum of 30 perceptrons or neurons.

14. Production process according to any of claims 8 to 13, characterized in that the learning algorithm is the Levenberg-Marquardt algorithm.

Images