ES2553679B2 - Plant and process for the production of diesel from industrial and urban waste - Google Patents

Abstract

The present invention relates to a plant for the production of diesel from the pyrolysis of industrial and urban waste, which includes sensors and measuring instruments and equipment control facilities and at least one artificial neural network (RNA) and also refers to a diesel production process in said plant, which is controlled by RNA; and includes the recording of the input and output values of the operational stages of the process, the training and validation of the RNAs, which once executed adjust the variables of the operational stages of the process until a completely controlled and optimized process is achieved.

Description

5

10

fifteen

twenty

25

30

35

DESCRIPTION

Plant and process for the production of diesel from industrial and urban waste. OBJECT OF THE INVENTION

The present invention belongs to the field of energy recovery and refers to a plant for the production of diesel from the pyrolysis of industrial and urban waste, which comprises means for the reception and pretreatment of waste, at least one reactor , a plurality of condensation towers followed each by a heat exchanger and a decanter, sensors and measuring instruments associated with the plant equipment so that at least one installation of regulation or control of the operation of the equipment of the plant and at least one state analyzer in the form of at least one artificial neural network (RNA).

The present invention also relates to an RNA-controlled diesel production process which comprises the operative stages of waste treatment, pyrolysis of the waste and condensation of the reactor exhaust gases and said process additionally comprises from the registration of values input and output of the process stages through the sensors adapted to the equipment, the training and validation of the RNAs, which once executed allow to readjust the variables of at least one operational stage of the process until a completely controlled and optimized process is achieved .

BACKGROUND OF THE INVENTION

Directive 2008/98 / EC of the European Parliament establishes a legal framework for the treatment of waste in the European Union. Its objective is to protect the environment and human health by preventing the harmful effects of production and waste management. This Directive calls on the Member States to eliminate the use of controlled landfills as a finalist technology for urban and industrial waste, promoting prevention, reuse and recycling. When this is not possible, energy recovery is the alternative chosen by the Directive.

Within the energy valorization there are several possible technologies, such as incineration, biomethanization, gasification or pyrolysis. Among these, gasification and pyrolysis are the technologies that obtain the highest energy efficiency since

2

5

10

fifteen

twenty

25

30

35

They convert waste into fuels that can be used beyond simple combustion to obtain thermal energy. In both cases, the waste is heated, directly or indirectly at temperatures high enough to initiate thermal cracking of the molecules that compose them, in a reducing atmosphere.

In the case of gasification, the operating temperature is between 800 ° C and 1200 ° C, favoring the conversion of the solid residue into a combustible gas, consisting of H2, CO, CO2, H2O and non-condensable hydrocarbons, minimizing the formation of condensable hydrocarbons and residual coke.

On the other hand, pyrolysis works with more moderate operating temperatures, between 300 ° C and 600 ° C, so that the production of condensable and coke hydrocarbons increases. The fuels produced can be used in internal combustion engines generating work, electricity and heat, significantly increasing the energy efficiency obtained from the waste compared to that of the incineration process.

The main problem of dealing with CSR (Recovered Solid Fuel) is the heterogeneity of the waste, which makes it difficult to gasify, so if you want to gasify the waste, you usually opt for a pyrolysis prior to the gasification process (patent ES2337442T3), so that gasification is carried out on heterogeneous effluents such as coke and non-condensable hydrocarbons that have been generated during pyrolysis.

EP2674472A1 and EP1538191B1 describe organic solid waste pyrolysis processes that involve the use of catalysts and the invention patent ES2166945T3 describes a process for the treatment of used rubber and plastics, where the treated waste is decomposed in a pyrolytic reactor without the presence of a catalyst. . The reactor is heated from the outside and operates in discontinuous mode. In a discontinuous mode of operation, the reactor always works in a non-stationary regime, thus varying the properties of the system throughout the reaction process. This mode of operation has disadvantages, which involves high operating costs and variation in the quality of The output products. It is not appropriate for large productions, such as those required for waste treatment.

A problem associated with this type of process is the heterogeneity of the waste to be treated, so that generally the raw material is restricted to a type of waste, for example

3

5

10

fifteen

twenty

25

30

35

In patents ES2220180T3, EP0592057B1 and ES2078743T3, pyrolysis is limited to plastic derivatives and rubber to maintain the homogeneity of the input residue.

Notably, patent application ES2503442A1 describes an integral waste valorization system for the production of diesel fuel which uses two different pyrolysis systems according to the characteristics of the residue. On the one hand, it uses a tubular reactor, only for the fraction of the polypropylene present in residue, in order to ensure homogeneity, while the rest of the material is pyrolyzed by the KDV system (patent EP1538191B1) which uses a catalyst.

Once the pyrolysis has been carried out, the condensation stage of the generated pyroltic oils (where non-condensable hydrocarbons are found) is crucial to obtain a quality fuel. Patents EP1190014B1 and EP0297424B1 describe two-stage condensation processes, first recovering components with a rocker point higher than water and then components with a lower rocker point.

On the other hand, the control of the operation variables in the process is essential to achieve the desired quality of the final fuel. In this type of process, the selection and implementation of the optimal control strategy is complex due to the heterogeneity of the waste to be treated.

The control systems most used in this type of plants are those based on a classic control widely known as proportional-integral-derivative PID which requires for its operation that at least one sensor, one controller and one actuator. The sensor determines the state of the system and provides an analog or digital signal to a controller. The controller reads an external signal that represents the value to be achieved. This signal is called the set point (or reference point), which is of the same nature and has the same range of values as the signal provided by the sensor. The controller subtracts the current point signal to the setpoint signal, thus obtaining! the error signal, which determines at every moment the difference between the desired value (setpoint) and the measured value. The error signal is used by each of the 3 components of the PID controller. The 3 signals added together make up the output signal that the controller will use to govern the actuator. The signal resulting from the sum of these three is called the manipulated variable and is transformed to be compatible with the actuator used.

5

10

fifteen

twenty

25

30

35

Currently, there are advanced control systems that play a decisive role in the control of industrial plants, especially if real-time process modeling is required. In this sense, in recent years, modeling with artificial neural networks has gained importance within the field of Chemical Engineering where it has been exploited in multiple applications.

An artificial neuron (NA) is a unit of calculation or processing element that attempts to model the behavior of a natural neuron, such as those that constitute the human brain, which has several inputs and combines them, usually with a basic sum. The sum of the inputs is modified by a transfer function and the value of the output of this transfer function is passed directly to the output of the processor element.

Figure 1 shows the detailed scheme of the operation of an NA.

In each NA, each input or input vector is multiplied by its corresponding "weight" or weight, analogous to the degree of synapse connection (degree of union between neurons). All weighted inputs are added together and the level of excitation or activation of the neuron is determined.

The output or output vector of the NA can be connected to the inputs of other NAs through weighted connections corresponding to the efficiency of the synapse of the neural connections.

An RNA consists of a set of NAs connected in a specific way, they are usually organized into groups called levels or layers. A typical network consists of a sequence of layers with connections between consecutive adjacent layers.

There are two layers with connections to the outside world. An input layer, input buffer, where data is presented to the network, and an output buffer layer that maintains the network response to an input. The rest of the layers are called hidden layers.

In the development of a neural network, unlike other computer systems such as expert systems, neither knowledge nor the rules of knowledge processing must be programmed. The neural network learns the rules of knowledge processing by adjusting the weighted connections between neurons of different layers of the network.

5

10

fifteen

twenty

25

30

35

The objective of the training of an RNA is to ensure that a given application, for a set of inputs, produces the set of desired or minimally consistent outputs. The training process consists of the sequential application of different sets or input vectors so that the weights of the interconnections are adjusted according to a predetermined procedure. During the training session the weights gradually converge towards the values that make each input produce the desired output vector.

Training algorithms or procedures for adjusting the values of RNA connections can be classified into two groups: Supervised and Non-Supervised.

Supervised training are algorithms that require the matching of each input vector with its corresponding output vector. The training consists of presenting a network input vector, calculating the network output, comparing it with the desired output, and the resulting error or difference is used to feedback the network and change the weights according to an algorithm that tends to Minimize the error. The pairs of vectors in the training set are applied sequentially and cyclically. The error and the adjustment of the weights for each pair is calculated until the error for the whole training set is a small and acceptable value.

Another characteristic of RNAs is that they generate their own rules by learning from the examples shown in the training phase. Learning is achieved through a learning rule that adapts or changes the weights of the connections in response to the input examples, and optionally also in response to the desired outputs. This characteristic of RNAs is what makes it possible to say that neural networks "learn from experience."

Another important feature of RNAs is the way information is stored. The memory of the RNAs is distributed throughout all the weighted connections of the network. Some RNAs are "associative", that is to say that for a partial entry the network will choose the most similar entry in memory and generate an output that corresponds to the full entry.

Backward-type networks have a supervised training method. Through this method, the network is presented with pairs of patterns, a paired input pattern

6

5

10

fifteen

twenty

25

30

35

with a desired output pattern. For each presentation the weights are adjusted so as to reduce the error between the desired output and the network response. The retropropagation learning algorithm involves a forward propagation phase and another backward propagation phase, where both phases are carried out for each pattern presented in the training session.

The forward propagation phase begins when a pattern is presented in the input layer of the network. Each unit of the entry corresponds to an element of the input pattern vector. The input units take the value of their corresponding element from the input pattern and the activation value or output level of the first layer is calculated. Then the other layers will carry out the forward propagation phase that determines the level of activation of the other layers. Once the forward propagation phase is completed, the correction phase or backward propagation phase begins. Calculations of the modifications of all the weights of the connections start at the output layer and continue backwards through all the layers of the network to the input layer. Within the types of weight adjustment two groups can be classified, adjustment of processing units of the output layer and adjustment of processing units of the hidden layers.

The adjustment of these weights of the output layer is relatively simple because it exists and the desired value is known for each of the units of the output layer. Each unit of the output layer produces a real number as output and is compared with the desired value specified in the training set pattern.

As for the adjustment of weights of the hidden layers, these layers do not have a desired output vector and therefore the error propagation method mentioned in the case of processing units of the output layer cannot be followed. The calculated error value for this type of processing units is obtained from the mathematical equations that we will not describe in the present description which allow defining a variable known as the learning coefficient. This coefficient, usually between 0.25 and 0.75 reflects the degree of network learning. Sometimes this coefficient is modified from a high initial value to progressively lower values during the training session in order to achieve better learning.

During the training or learning process of Back Propagation it is common to measure quantitatively the learning by means of the RMS (Root Mean Square) value of the error

7

5

10

fifteen

twenty

25

30

35

of the network. This measure reflects the way in which the network is achieving correct answers; As the network learns, its RMS value decreases. Because the network output values and the desired output values are real values, it is necessary to define a cut-off parameter or a threshold value of the RMS value of the network error that allows to say that the network approaches the output desired and consider that the answer is correct.

Particularly in the present invention, a control system based on the use of artificial neural networks of an industrial and urban waste pyrolysis plant has been developed for the first time; Once trained with the input and output variables of the process, it is able to predict what the properties of the effluents from the reactor will be, indicating the temperature setpoint values in the five condensation stages of the pyrolytic oils. Together, a second system of artificial neural networks has been developed that is responsible for controlling the temperature at the head of the "Quench" towers of the condensation stages. That is, the first network determines what the temperature in the head of each column and the second regulates the flow of refrigerant to maintain the assigned temperature.

DESCRIPTION OF THE INVENTION

The present invention presents a plant for the production of diesel oil by pyrolysis of industrial and urban organic waste, which comprises the following equipment; means for receiving and conditioning the waste, at least one reactor, a plurality of condensation towers linked each to a heat exchanger and a decanter, at least one non-condensed gas scrubber tower, at least one cooling equipment indirect contact, a deposit for the reception of wastewater and leachate and a treatment station (WWTP) and additionally includes;

- 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 installation of regulation or control of the operation of the equipment of the plant.

5

10

fifteen

twenty

25

30

35

- at least one state analyzer in the form of at least one artificial neural network (RNA), so that it emits from the measured values, control quantities towards the regulation and control facilities of the plant equipment.

The present invention also describes a process for the production of diesel oil in a plant as described above, which comprises the operative stages of waste treatment, pyrolysis of waste and condensation of reactor exhaust gases and is controlled by neural networks. artificial (RNA), and said process additionally comprises:

- acquire values of the input variables of at least one operational stage of the process by means of sensors and measuring instruments,

- record values of the output variables of at least one operational stage of the process,

- analyze the values of the output variables and choose those with the highest statistical incidence with respect to the values of the input variables, which are considered the theoretical or desired values,

- present to the RNAs the history of the chosen values of the input and output variables of at least one operational stage which you want to model and train by adjusting the weights of the neuron connections until they gradually converge towards the values that make each input variable value produce the corresponding theoretical or desired value of the output variable,

- execute the RNAs, trained by means of a learning algorithm, for the estimation of the values of the output variables from the values of the input variables of at least one operational stage of the ongoing process and,

- readjust the variables of at least one operative stage of the process until the values of the estimated output variables are adjusted to the theoretical or desired output values with a lower error than indicated.

5

10

fifteen

twenty

25

30

35

The pyrolysis process of urban and industrial waste is a non-linear, multivariable, and non-stationary process, which is complex to control due to the great heterogeneity of the waste to be treated. The present invention overcomes the state of the art by solving this problem in an efficient way by using RNA to control and optimize a process that operates in a continuous mode and without the need to use catalysts, resulting in high quality products.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT OF THE INVENTION

The following describes a preferred embodiment of the invention but not exclusive in relation to the attached figures. Particularly but not limited to, Figure 2 shows the distribution of equipment of a plant for the production of diesel by pyrolysis of industrial and urban organic waste, as well as the input and output materials to the operational stages according to the following nomenclature :

 one)

 Reactor

 2)

 Particle filter

 3)

 "Quench" tower type espay

 4)

 Heat exchanger

 5)

 Decanter

 6)

 "Quench" tower with padding

 7)

 Heat exchanger

 8)

 Decanter

 9)

 "Quench" tower with padding

 10)

 Heat exchanger

 eleven)

 Decanter

 12)

 "Quench" tower with padding

 13)

 Heat exchanger

 14)

 Decanter

 fifteen)

 "Quench" tower with padding

 16)

 Heat exchanger

 17)

 Decanter

 18)

 Scrubber gas cleaning tower

 19)

 Water tank with salts

 twenty)

 Deposit with NaOH solution

 twenty-one)

 Cold equipment

5

10

fifteen

twenty

25

30

35

22) Wastewater tank (WWTP)

Input materials and output products of the operational stages:

CSR - Solid Fuel Recovered

C - Waste

G - Diesel

SG - Singas

L1 - Light 1

L2 - Light 2

FP - Heavy fraction

CON - Condensates

The plant additionally comprises,

- measuring instruments, more specifically temperature probes located in different reactor heating zones and at the head of condensation towers

- PLC control facilities for reactor operation and condensation towers,

- and state analyzers in the form of three RNAs, so that they emit from the measured values, control quantities towards the regulation and control facilities (PLC) of the reactor and the condensation towers.

Preferably but not limitatively, the reactor operates with indirect heating, provided by 8 gas burners located outside the rotary cylinder.

Preferably but not limitatively, of the five "" Quench "" towers, the first is of the "spray" type and the next four are filling towers with "Pall" rings.

Preferably but not limitatively, gas scrubbing occurs in a tower of type "Scrubber"

The present invention also refers to the process for the production of diesel oil in a plant such as the one described above which comprises the following operational stages:

5

10

fifteen

twenty

25

30

35

Waste treatment; Depending on its characteristics, it can be directly stored to maintain a constant feed of the reactor or must be conditioned, by crushing and drying, before being stored. To maximize the performance of the process, the residue must enter the reactor with a humidity equal to or less than 10% and a particle size equal to or less than 50 mm, although it is not an indispensable condition. There is no limitation in terms of content of metals and inert materials although these penalize the energy efficiency of the process. An amount of chlorine is recommended, on a dry basis, less than 1% by weight to guarantee the useful life of the facilities, as well! as a concentration of sulfur, on a dry basis, less than 0.8% by weight to ensure that the fuel obtained complies with Royal Decree 1088/2010 that limits the concentration of sulfur by 1,000 ppm, which can be modified by adapting to the modifications of the law or future needs.

In a preferred embodiment of the invention, the pyrolysis reactor is fed with 500 kg / hr of CSR with a humidity of 10%, particle size of 50 mm and a PCI of 22.5 MJ / Kg. The average composition is 40% in varied plastics (PE, PP, PET and PS), 50% in paper, cardboard, fabrics and FORM and the rest inert material.

The pyrolysis of the waste is produced in a "Kiln" horizontal rotary cylindrical reactor. It is fed continuously by means of two hoppers in series that ensure the tightness of the feed. with the residue, ensuring the inert atmosphere inside the reactor. The continuous rotary design with an inclination of up to 5 °, causes the material to circulate inside the same direction of the inclination of the drum causing a progressive heating ramp of the material through its internal translation of the drum and taking advantage of the natural thermal gradient that exists inside it, which results in a maximum optimization of the energy efficiency of the material (up to 100% more than in the classic intermittent furnaces). The regulation and control of the treatment cycles is carried out through temperature probes in each zone that give the signal to the table and Electrical control for internal temperature control of each combustion chamber (at no time there is direct contact between the burner flames and the interior of the reactor) by means of a PLC, and electronic frequency inverters to control the speed of rotation of the pumps . By adjusting the two variables, temperature and speed (residence time), the necessary pyrolysis conditions for the material are obtained. The regulation and control of the treatment cycles are carried out through a PLC that governs 4 independent temperature regulation zones, each with two

5

10

fifteen

twenty

25

30

35

gas burners The available working temperature range is 450 ° C to 900 ° C. The pyrolysis gases composed of pyrolltic oils and non-condensable gases are extracted by the higher rear part of the pyrolysis drum and directed to the condensation stage, passing first through a filter, where the fine particles that can be dragged are collected. The residual solid phase, formed by carbon and inert material is cooled in a rotating cylinder, which, like that of pyrolysis, is driven by a compact group formed by a motor-reducer governed through an electronic frequency variator that allows varying the speed without staggering. Once cooled, this material is constantly dosed in "big-bags".

At the exit of the reactor, 126 kg / h of carbon with a high ash content and a PCI of 15 MJ / Kg are obtained. Of the pyrolysis gases obtained, 92 kg / h are non-condensable gases with a PCI of 25 MJ / Kg and are used to heat the reactor.

Preferably but not limiting, condensation occurs in five stages to separate products according to their boiling point. Each stage is composed of a "Quench" tower and a decanter. The first two towers cool the gases above the water temperature to avoid water condensation and use the condensed pyroltic oils in the second "Quench" tower as coolant The third and fourth tower use water as a cooling fluid and allow a large part of the water contained in the gases to condense. Finally, the fifth tower "Quench" uses pyroltic oils, as a cooling fluid, to be able to condense the lighter fuel fraction thanks to its thermodynamic affinity. The first tower is a spray type, with two rows of sprinklers, while the other four are filled with rings "Pall".

The heaviest fraction of the oils is obtained from the first tower, which corresponds to a boiling point greater than 350 ° C. From the second tower, the largest fraction that will be used for the formulation of class C diesel oil is obtained, which comprises pyroltic oils with boiling points between 180 ° C and 350 ° C. In the third and fourth tower, mostly water and polar compounds related to water are condensed together with a light fraction of hydrocarbons. Finally, in the fifth column the lighter fraction of the pyroltic oils is condensed, with boiling points below 150 ° C.

The different fractions of pyrolytic oils, obtained at each stage of condensation, are mixed until the desired quality is achieved for class C diesel.

13

5

10

fifteen

twenty

25

30

35

heavier fraction is mixed with the inlet residue and pyrolyzed again. The surplus of the light fraction is stored for sale as gasoline. The aqueous fraction is derived to a water treatment plant.

The non-condensable gases are then cleaned and adapted for use in the burners of the external chamber of the reactor. These gases have already been partially treated through the pyrolysis and condensation zones, and the larger suspended particles have been trapped in the basket filter, and through the five stages of condensation they have been "washed" by an organic current and one of water. The cleaning of the gases takes place preferably in a Scrubber tower where the gases are washed in contact with a water stream with a slightly basic pH in case the presence of acids in the gas stream is detected. The water stream will be purged from precipitates (salts) periodically and its pH will be adjusted according to needs. At the exit of the Scrubber tower, the gas is cooled with an indirect contact cold equipment at -10 ° C to condense the water and hydrocarbons that are absorbed in the gas and avoid condensation before being used in the reactor burners .

The condensable gases obtained are composed of 130 kg / h of the aqueous phase (partly formed during the pyrolysis and partly contained in the moisture of the inlet residue) and 152 Kg / h of pyrolytic oils of which 105 kg / h serve to formulate diesel for heating and the rest is a fraction of light hydrocarbons. The pyrolytic oils obtained have an average PCI of 36 MJ / Kg.

Preferably, three RNAs are used to control the operation variables in the process, RNA1 to determine the operation variables of the pyrolysis reactor, RNA2 to determine the head temperatures of the “Quench” towers and RNA3 to optimize the system of control of the temperatures of head of the towers “Quench”. Said control process includes:

- acquire values of the input variables to the reactor and the condensation towers

by sensors and measuring instruments,

- record values of the output variables of the reactor and the towers of

condensation,

5

10

fifteen

twenty

25

30

35

- analyze the values of the output variables and choose those with the highest statistical incidence with respect to the values of the input variables, which are considered the theoretical or desired values,

- Present the chosen values of the input and output variables to the reactor and the condensation towers to the RNAs 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 produces the corresponding theoretical or desired value of the output variable,

- execute the RNAs, trained by means of a learning algorithm, for the estimation of the values of the output variables from the values of the input variables of the ongoing process and,

- readjust the variables of at least one operative stage of the process until the values of the estimated output variables are adjusted to the theoretical or desired output values with a lower error than indicated.

RNA1 encompasses the pyrolysis reactor, where it, considered as a black box, receives a series of inputs (characteristics of the residue and variables of operation of the reactor) and obtains outputs (characteristics of the products obtained).

The input variables in the present preferred embodiment, with a total of 17, are:

Characteristics of the residue:% of plastics,% of materials with high cellulosic content,% of inert materials, PCI, humidity, average particle size, mass flow and elemental analysis (carbon, nitrogen, oxygen and hydrogen).

Operation variables: residence time of the gas, residence time of the solid and temperatures of the four heating zones (T1, T2, T3 and T4)

The output variables in the present preferred embodiment, with a total of 9, are:

% solid (carbon),% non-condensable gas,% condensable gas (% aqueous phase,% organic phase with a boiling point above 360 ° C,% organic phase with a boiling point between 360 ° C and 180 ° C and% organic phase with a boiling point

fifteen

5

10

fifteen

twenty

25

30

35

below 180 ° C), solid PCI, non-condensable gas PCI and condensable organic phase PCI.

A 4-layer backpropagation network is used (1 input, 2 hidden and 1 output). The first layer has 17 neurons that correspond to the 17 input variables, the second and third layers have a maximum of 30 neurons and their number is adjusted during the training of the network and the last layer has 9 neurons, which correspond to The output variables.

RNA1, once trained with historical data on the operation of the pyrolysis reactor, is able to predict the 9 output variables, from the 17 input variables. In this way, once the batch of waste to be pyrolyzed is characterized, RNA1 is executed for different values of the reactor operating variables and those that maximize the production of diesel oil are chosen.

RNA2 covers the 5 "Quench" columns, where condensable hydrocarbons and the aqueous phase are recovered. The objective of this RNA is to determine the head temperatures of the "Quench" towers that maximize the separation of condensable gases into five well differentiated effluents, always favoring the effluent that includes the boiling points between 360 ° C and 180 ° C, which correspond to diesel. Five sub-networks are made in series, one for each column (RNA21, RNA22, RNA23, RNA24, RNA25). The five sub-networks share structure and variables of input and output, changing only the data used for their training. Five 3-layer backpropagation networks are used (1 input, 1 hidden and 1 output)

The input variables in the present preferred embodiment, with a total of 7, are:

Gas flow,% of non-condensable gas,% of condensable gas (% of aqueous phase,% of organic phase with a boiling point greater than 360 ° C,% of organic phase with a boiling point between 360 ° C and 180 ° C and% organic phase with a boiling point below 180 ° C) and tower head temperature.

The output variables in the present preferred embodiment, with a total of 3, are:

Flow of gases that have condensed,% of organic phase and temperature at which 10% of the condensed organic phase evaporates.

16

5

10

fifteen

twenty

25

30

35

The RNA2 sub-networks are run in series, following the RNA1 network, until the head temperatures of the "Quench" towers are found that optimize the production of the different qualities of hydrocarbons for the final formulation of diesel.

The RNA3 consists of five sub-networks, one for each tower "Quench", and is responsible for controlling the temperature of the head of the towers, replacing the classic control with a PID controller. The advantages of using an RNA in Instead of a PID controller, the control is more robust with an RNA, it adapts faster to disturbances and is not as sensitive to downtime as PIDs. Five 3-layer backpropagation networks are used (1 input, 1 hidden and 1 output)

The input variables in the present preferred embodiment, with a total of 4, are:

Tower head temperature at time n, tower head temperature at time n-1, tower head temperature at time n-2 and gas flow.

The output variable in the present preferred embodiment is:

% Overture of the control valve

Once RNA1, RNA2 and RNA3 have been trained and validated, they are able to adjust, optimize and control the entire production process more efficiently than adjusting the variables afterwards, from the analysis of the products obtained.

BRIEF DESCRIPTION OF THE FIGURES

To complement the description that is being carried out and in order to help a better understanding of the characteristics of the invention, it is accompanied as an integral part of this description, a set of drawings, where with an illustrative and non-limiting nature, what has been represented next:

Figure 1 shows the detailed scheme of the functioning of a neuron. In this case, the neuron receives the signal of two inputs (x1 and x2). These are weighted according to the weights (p11 and p12) of the connections and through a transfer function (f (h))

5

10

fifteen

twenty

25

gets an output signal from the neuron (y1). The weights of the connections are modified during training until the desired signal is achieved.

Figure 2 shows the distribution of the equipment and the input and output materials of the operational stages of a diesel production plant.

Figure 3 shows the simplified scheme of the process controlled by the three interconnected RNAs. RNA1 predicts what the properties of the reactor effluents will be and determines the temperature inside the reactor, RNA2 determines what the temperature at the head of each column should be and RNA3 regulates the flow of refrigerant to maintain the assigned temperature.

Figure 4 shows the scheme of RNA1. The 17 input variables feed the 17 neurons of the input layer (En). These connect with the first hidden layer (O1n) of neurons that in turn connect with the second hidden layer of neurons (O2n) and these with the 9 neurons of the output layer (Sn)

Figure 5 shows the scheme of one of the RNA2. The 7 input variables feed the 7 neurons of the input layer (En), which in turn are connected to the neurons of the hidden layer (On) and these with the 3 neurons of the output layer (Sn)

Figure 6 shows the scheme of one of the RNA3. The 4 input variables feed the 4 neurons of the input layer (En), which in turn, are connected with the neurons of the hidden layer (On) and these with the neuron of the output layer (Sn).

Claims (20)

5 10 fifteen twenty 25 30 1. Plant for the production of diesel by pyrolysis of industrial and urban organic waste, which includes the following equipment; means for receiving and conditioning the waste, at least one reactor, a plurality of condensation towers linked each to a heat exchanger and a decanter, at least one non-condensed gas scrubber tower, at least one cooling equipment of indirect contact, a deposit for the reception of wastewater and leachate and a treatment station (WWTP) and, - 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 installation of regulation or control of the operation of the equipment of the plant. - at least one state analyzer in the form of at least one artificial neural network (RNA), so that it emits from the measured values, control quantities towards the regulation and control facilities of the plant equipment. 2. Plant for the production of diesel oil according to revindication 1 characterized in that the reactor where the pyrolysis occurs is a "Kiln" rotary horizontal cylindrical reactor. 3. Plant for the production of diesel according to revindication 1 characterized in that it comprises five "Quench" type condensation towers located in series, of which the first one is of the "spray" type and the next four are filling towers. 4. Plant for the production of diesel oil according to revindication 1 characterized in that it comprises a scrubber scrubber tower for non-condensable gases. 5 10 fifteen twenty 25 30 35 5. Process of production of diesel oil in a pyrolysis plant according to any of the preceding claims which comprises the operative stages of waste treatment, pyrolysis of the waste and condensation of the reactor exit gases characterized in that it is controlled by networks artificial neuronal (RNA), and said process additionally comprises: - acquire values of the input variables of at least one operational stage of the process by means of sensors and measuring instruments, - record values of the output variables of at least one operational stage of the process, - analyze the values of the output variables and choose those with the highest statistical incidence with respect to the values of the input variables, which are considered the theoretical or desired values, - present to the RNAs the history of the chosen values of the input and output variables of at least one operational stage which you want to model and train by adjusting the weights of the neuron connections until they gradually converge towards the values that make each input variable value produce the corresponding theoretical or desired value of the output variable, - execute the RNAs, trained by means of a learning algorithm, for the estimation of the values of the output variables from the values of the input variables of at least one operational stage of the ongoing process and, - readjust the variables of at least one operative stage of the process until the values of the estimated output variables are adjusted to the theoretical or desired output values with a lower error than indicated. 6. Diesel oil production process in a pyrolysis plant according to revindication 5 characterized in that the RNAs are trained and validated to adjust, optimize and control the entire production process. 7. Process of production of diesel in a pyrolysis plant according to claims 5 and 6, characterized in that an RNA1 determines the operation variables in the pyrolysis stage of the waste. 8. Process of production of diesel in a pyrolysis plant according to revindication 7, characterized in that RNA1, operates with input variables and output variables, where the input variables are, - the characteristics of the waste to be treated comprising at least one within the group of; % of plastics,% of materials with high cellulosic content,% of inert materials, PCI, humidity, average particle size, mass flow and elemental analysis (carbon, nitrogen, oxygen and hydrogen) or combinations thereof. - the reactor operating variables comprising at least one within the group of; the residence time of the gas, residence time of the solid and temperatures of the heating zones (T1, T2, T3 and T4) or combinations thereof. And the output variables are those comprising at least one within the group consisting of: -% solid (carbon),% non-condensable gas,% condensable gas,% aqueous phase,% organic phase with a boiling point above 360 ° C,% organic phase with a boiling point between 360 ° C and 180 ° C and% of organic phase with a boiling point below 180 ° C, solid PCI, non-condensable gas PCI and condensable organic phase PCI or combinations thereof. 9. Production process of diesel oil in a pyrolysis plant according to claims 7 and 8, characterized in that RNA1 is of the retropropagation type with at least 4 layers comprising; 1 input layer, 2 hidden layers and 1 output layer. 10. Diesel oil production process in a pyrolysis plant according to revindication 9, characterized in that the first layer has at least 17 neurons that correspond to the input variables, the second and the third image 1 layer, they have at least a maximum of 30 neurons and their number is adjusted during the training of the network and the last layer has at least 9 neurons, which correspond to the output variables. 11. Diesel oil production process in a pyrolysis plant according to claims 5 and 6, characterized in that an RNA2 determines the head temperatures of the condensation towers of the gases leaving the pyrolysis reactor. 12. Diesel oil production process in a pyrolysis plant according to revindication 11, characterized in that RNA2, operates with input variables and output variables, where the input variables are: - the characteristics of the condensable gases and the aqueous phase, comprising at least one within the group consisting of the gas flow rate,% non-condensable gas,% condensable gas,% aqueous phase,% organic phase with a point boiling above 360 ° C,% organic phase with a boiling point between 360 ° C and 180 ° C and% organic phase with a boiling point below 180 ° C) and tower head temperature or combinations of them. And the output variables are: - the characteristics of the condensates obtained, comprising at least one within the group consisting of, flow of gases that have condensed,% of organic phase and temperature at which 10% of the condensed organic phase evaporates or combinations thereof. 13. Process of production of diesel oil in a pyrolysis plant according to claims 11 and 12, characterized in that RNA2 in turn comprises five artificial neuronal sub-networks (RNA21, RNA22, RNA23, RNA24, RNA25) of the retropropagation type, of at least 3 layers each, comprising 1 input layer, 1 hidden layer and 1 output layer. 14. Process of production of diesel oil in a pyrolysis plant according to claims 11 to 13, characterized in that the RNA2 sub-networks each correspond to a condensation column and are executed in series. image2 15. Process of production of diesel oil in a pyrolysis plant according to claims 5 to 14, characterized in that an RNA3 controls the head temperatures of the gas condensation towers. 5 16. Process of production of diesel oil in a pyrolysis plant according to the revindication 15, characterized in that, the input variables to RNA3 are: - head temperature of the tower at time n, head temperature of the tower at time n-1, head temperature of the tower at time n-2 and flow of gases or combinations thereof. And the output variable comprises: 15 -% overture of the control valve 17. Production process of diesel oil in a pyrolysis plant according to claims 15 and 16, characterized in that RNA3 uses five artificial neural sub-networks of the backpropagation type, one for each column of 20 condensation, of at least 3 layers each comprising at least 1 input layer, 1 hidden layer and 1 output layer. 18. Production process of diesel oil in a pyrolysis plant according to the preceding claims, characterized in that the pyrolysis is carried out continuously, within a temperature range between 450 ° C and 900 ° C in an oxygen-free atmosphere and in the absence of catalysts. 19. Production process of diesel oil in a pyrolysis plant according to the preceding claims, characterized in that the condensation of the gases at the exit of the reactor occurs in five stages, where: - the first stage of condensation produces the heaviest fraction of oils with boiling temperature (Te) above 350 ° C, - the second stage produces the pyrolytic oils with (Te) between 180 ° C and 350 ° C, 5 10 fifteen twenty 25 30 - the third and fourth tower stages produce a light fraction of hydrocarbons, water and polar compounds - the fifth stage produces the lightest fraction of the hydrocarbons with (Te) below 150 ° C. 20. Production process of diesel oil in a pyrolysis plant according to the preceding claims, characterized in that the non-condensable gases are washed with a stream of water in a washing tower, cooled to -10 ° C in a cooling equipment by indirect contact and the reactor burners are recirculated once free of condensates. 21. Process of production of diesel oil in a pyrolysis plant according to the preceding claims, characterized in that the wastewater and leachate from the process is derived to a water treatment plant.