Full-automatic intelligent control system and method for fermentation equipment

Abstract

The invention discloses a full-automatic intelligent control system and method for fermentation equipment, and relates to the technical field of automatic control of the fermentation equipment. The method comprises the following steps: acquiring time sequence data of state parameters and controlled parameters in a historical fermentation process, classifying fermentation states in fermentation equipment according to the state parameter data, acquiring a state space, defining the adjustment of the controlled parameters as one action, classifying the action according to the adjustment amount of the controlled parameters, and acquiring an action space; extracting a state parameter time sequence data sequence when any action in the action space is executed, constructing a Markov chain model of fermentation states in the fermentation equipment, defining a conversion value between any two fermentation states, calculating a conversion value expectation after k time steps at the t moment when the action is executed, and selecting a state parameter time sequence data sequence in the k time step; the action corresponding to the maximum value in the value expectation calculation result is converted to serve as one action of the kth time step, and the controlled parameter of the kth time step is adjusted.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic control of fermentation equipment, and in particular to a fully automatic intelligent control system and method for fermentation equipment. Background Art

[0002] Aerobic composting and anaerobic composting are two common forms of composting. Anaerobic composting has a simple process and low compost temperature, but a long composting period and strong odor, so it is less used at present. Aerobic composting is carried out under ventilation and feeding conditions, with higher compost temperature, shorter fermentation cycle and higher degree of harmlessness. Therefore, many composting plants at home and abroad use aerobic composting. Aerobic composting process of organic waste is an effective means to achieve harmlessness, stabilization and resource utilization of organic solid waste, and has become one of the key solutions to solve ecological and environmental problems.

[0003] Natural composting has problems such as incomplete degradation, unbalanced nutrition, slow fermentation speed, long composting time and odor pollution. The use of composting reactors can effectively solve the problems of waste treatment and discharge, thereby improving fermentation efficiency, shortening production cycles and realizing mechanized production, making the composting process more environmentally friendly and controllable. At present, common types of aerobic composting reactors include: tower reactors, silo reactors, tunnel kiln composting reactors, drum reactors, tank reactors, etc. In the actual production process, it can be selected according to the production scale and production needs. In the process of industrial composting, the fermentation of organic matter is essentially the action process of microorganisms. Relevant studies have shown that controlling the temperature, humidity and oxygen supply of the fermentation process is the key to the whole process. Therefore, the above parameters need to be precisely controlled to ensure the efficient fermentation process.

[0004] At present, the control method used for the temperature and oxygen flow of aerobic composting reactors is still the traditional PID control, and there are two main control methods: closed-loop control and compensation control. The existing control system is a nonlinear time-delay system with multiple inputs, large disturbances, and strong coupling. The conventional PID control algorithm can hardly make the system reach a stable state. Therefore, in order to reduce the overshoot, differential and intermediate feedback methods are generally used for adjustment, but the control effect is not ideal, the controlled quantity still has a large overshoot, and the response speed is very slow.

[0005] In order to solve the problems existing in conventional PID control, experts and scholars at home and abroad have proposed methods that can make the parameters of the PID control system change with the system operating environment. Among them, the more effective and simple ones are the fuzzy PID control method and the neural network PID control method. The error signal of the neural network learning system is used to automatically adjust the PID parameters, so that the system parameters have the ability to self-adjust with the changes in the operating environment. Although the above control method can optimize the control effect of the traditional PID control algorithm, in the actual production process, different adjustment actions often have an impact on a variety of controlled parameters. For example, when the oxygen flow rate is adjusted, the temperature inside the reactor will also be affected and changed with the change of the oxygen flow rate. Then the system needs to further adjust the temperature. In this cycle, the formulation of the adjustment strategy and the execution of the adjustment action are always in a passive state during the entire control process, and the controlled parameters cannot be actively regulated according to the real-time status and change trend inside the reactor. To this end, we propose a fully automatic intelligent control system and method for fermentation equipment. Summary of the invention

[0006] The main purpose of the present invention is to provide a fully automatic intelligent control system and method for fermentation equipment, which can effectively solve the problems in the background technology.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] A fully automatic intelligent control method for fermentation equipment, comprising:

[0009] Acquire time series data of state parameters and controlled parameters in the historical fermentation process; wherein the state parameter includes at least one of the temperature, oxygen content, and moisture content in the fermentation equipment; and the controlled parameter is the ventilation volume in the fermentation equipment.

[0010] The fermentation state in the fermentation equipment is classified according to the state parameter data to obtain a state space S={s1, s2, ..., s u},s u It is represented as the u-th type of fermentation state in the fermentation equipment; the adjustment of the controlled parameter is defined as an action, and the actions are classified according to the adjustment amount of the controlled parameter to obtain the action space A = {a1, a2, ..., a v}, a v It is represented as the vth type of action;

[0011] The fermentation state classification process comprises the following steps:

[0012] Step S11: Set the optimal value of the state parameter of the rth category at time t to The qth sampling data value of the rth type of state parameter in the historical fermentation process is

[0013] Step S12: Calculate the q sampling data values ​​of the state parameters respectively The optimal value of the state parameter The distance between The calculation formula is: To obtain Construct a data sample set, denoted as

[0014] Step S13: Use the Sigmoid function to adjust the mapping of each element in the acquired data sample set to between [0,1], and obtain the Sigmoid function value corresponding to each element Using function values The elements are classified according to the following principles:

[0015] when hour, For the first category;

[0016] when hour, For the second category;

[0017] when hour, For the third category;

[0018] And so on.

[0019] when hour, It is the μth category;

[0020] in, They are the minimum and maximum values ​​of the Sigmoid function of the elements in the data sample set;

[0021] Step S14: According to The classification of the state parameter is used to determine the qth state parameter sampling data value. The corresponding fermentation states are as follows:

[0022] when When it is the first type, the state parameter The corresponding fermentation state is the first type;

[0023] when For the second type, the state parameter The corresponding fermentation state is the second type;

[0024] when When it is the third type, the state parameter The corresponding fermentation state is the third category;

[0025] And so on.

[0026] when When it is the μth type, the state parameter The corresponding fermentation state is class μ.

[0027] The action classification process includes the following steps:

[0028] Step S21: Set the controlled parameter at time t to cp t , the adjustment step size is △cp;

[0029] Step S22: the adjustment of the controlled parameter is divided into three categories: increase, remain unchanged, and decrease. When the adjustment amount is within an adjustment step length △cp, it is regarded as an adjustment to keep the controlled parameter unchanged. When the adjustment step length △cp is increased, it is regarded as an adjustment to increase the controlled parameter. When the adjustment step length △cp is decreased, it is regarded as an adjustment to decrease the controlled parameter.

[0030] Step S23: taking any adjustment of keeping the controlled parameter unchanged, increasing the controlled parameter, or decreasing the controlled parameter in step S22 as one action, and obtaining a classification result of the action;

[0031] Step S24: Collect all the actions once and obtain the action space A = {a1, a2, ..., a v}.

[0032] Extract any action a in the action space A λ State parameter time series data sequence when executed Using data series A Markov chain model of the fermentation state in the fermentation equipment is constructed, and the model is defined as: Represented as action a λ When executed, the probability that the fermentation state in the fermentation equipment is transferred from the i-th category to the j-th category in one step, where i, j∈u; λ∈v; S t Represented as action a λ The fermentation state in the fermentation equipment at the time t being executed; S t+1 Represented as action a λ The fermentation state in the fermentation equipment at time t+1 after a time step is executed; probability The calculation formula is:

[0033]

[0034] In the formula, Represented as a data sequence In the case of action a λWhen executed, the frequency at which the fermentation state in the fermentation equipment is transferred from the i-th category to the j-th category in one step.

[0035] Defined in the state space S, any two of the fermentation states s i 、s j The conversion value between i →s j ), calculate action a λ The expected value of the transformation after k time steps at time t is G t+k , the calculation formula is: In the formula, γ represents the attenuation factor, and γ∈(0,1); R t+k (s i →s j ) is represented as the conversion value at time step t+k;

[0036] Select the kth time step, the expected transformation value G t+k The maximum value G in the calculation result (t+k)max The corresponding action is taken as an action at the k-th time step, and the determined action is executed to adjust the controlled parameter at the k-th time step.

[0037] A fully automatic intelligent control system for fermentation equipment, comprising:

[0038] A data acquisition module, used to obtain the time series data of state parameters and controlled parameters in the historical fermentation process;

[0039] A first data processing module is used to classify the fermentation state in the fermentation equipment according to the state parameter data to obtain a state space S;

[0040] A second data processing module is used to define the adjustment of the controlled parameter as an action, classify the actions according to the adjustment amount of the controlled parameter, and obtain an action space A;

[0041] Model building module, used to extract any action a in the action space A λ State parameter time series data sequence when executed Using data series Construct a Markov chain model of the fermentation state in the fermentation equipment;

[0042] The conversion value expectation calculation module is used to define in the state space S, any two of the fermentation states s i 、s j The conversion value between i →s j ), and calculate action a λ The expected value of the transformation after k time steps at time t is Gt+k ;

[0043] An action determination module is used to select the expected conversion value G in the kth time step t+k The maximum value G in the calculation result (t+k)max The corresponding action is taken as an action at the kth time step;

[0044] The action execution module is used to execute a determined action and adjust the controlled parameter at the kth time step.

[0045] The system also includes a memory, a processor, and a computer program stored on the memory and executable on the processor.

[0046] The present invention has the following beneficial effects:

[0047] Compared with the prior art, the present invention obtains the time series data of the state parameters and the controlled parameters in the historical fermentation process, classifies the fermentation state in the fermentation equipment according to the state parameter data, obtains the state space S, defines the adjustment of the controlled parameter as an action, classifies the actions according to the adjustment amount of the controlled parameter, obtains the action space A, and extracts any action a in the action space A. λ State parameter time series data sequence when executed Using data series Construct a Markov chain model of the fermentation state in the fermentation equipment, and define in the state space S that any two of the fermentation states s i 、s j The conversion value between i →s j ), calculate action a λ The expected value of the transformation after k time steps at time t is G t+k , select the kth time step, the expected conversion value G t+k The maximum value G in the calculation result (t+k)max The corresponding action is taken as an action at the kth time step, and a determined action is executed to adjust the controlled parameter at the kth time step. The statistical model is used to comprehensively consider the influence of the adjustment action on various controlled parameters, and the controlled parameters are actively regulated according to the real-time state and change trend inside the reactor to improve the control effect and the control accuracy of the controlled parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of a process of a fully automatic intelligent control method for fermentation equipment of the present invention;

[0049] Figure 2 This is a structural schematic diagram of a fully automatic intelligent control system for fermentation equipment of the present invention;

[0050] Figure 3 It is a schematic diagram of the calculation process of the expected conversion value in the technical solution of the present invention. DETAILED DESCRIPTION

[0051] The present invention will be further described below in conjunction with specific implementation methods, wherein the accompanying drawings are only used for exemplary descriptions and represent only schematic diagrams rather than actual drawings, and should not be understood as limiting the present invention. In order to better illustrate the specific implementation methods of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the size of the actual product.

[0052] The specific implementation process of the technical solution of the present invention includes the following steps:

[0053] Step 1: Obtain time series data of state parameters and controlled parameters in the historical fermentation process; wherein the state parameters include at least one of the temperature, oxygen content, and moisture content in the fermentation equipment; and the controlled parameter is the ventilation volume in the fermentation equipment.

[0054] Step 2: Classify the fermentation state in the fermentation equipment according to the state parameter data and obtain the state space S = {s1, s2, ..., s u},s u It is represented as the u-th fermentation state in the fermentation equipment;

[0055] The fermentation status classification process includes the following steps:

[0056] Step S21: Set the optimal value of the rth state parameter at time t to The qth sampling data value of the rth type of state parameter in the historical fermentation process is

[0057] Step S22: Calculate the q sampling data values ​​of the state parameters respectively The optimal value of the state parameter The distance between The calculation formula is: To obtain Construct a data sample set, denoted as

[0058] Step S23: Use the Sigmoid function to adjust the mapping of each element in the acquired data sample set to between [0,1], and obtain the Sigmoid function value corresponding to each element Using function values The elements are classified according to the following principles:

[0059] when hour, For the first category;

[0060] when hour, For the second category;

[0061] when hour, For the third category;

[0062] And so on.

[0063] when hour, It is the μth category;

[0064] in, They are the minimum and maximum values ​​of the Sigmoid function of the elements in the data sample set;

[0065] Step S24: According to The classification of the state parameter is used to determine the qth state parameter sampling data value. The corresponding fermentation states are as follows:

[0066] when When it is the first type, the state parameter The corresponding fermentation state is the first type;

[0067] when For the second type, the state parameter The corresponding fermentation state is the second type;

[0068] when When it is the third type, the state parameter The corresponding fermentation state is the third category;

[0069] And so on.

[0070] when When it is the μth type, the state parameter The corresponding fermentation state is class μ.

[0071] Step 3: Define the adjustment of the controlled parameter as an action, classify the actions according to the adjustment amount of the controlled parameter, and obtain the action space A = {a1, a2, ..., a v}, a v It is represented as the vth type of action;

[0072] The action classification process includes the following steps:

[0073] Step S31: Set the controlled parameter at time t to cp t , the adjustment step size is △cp;

[0074] Step S32: the adjustment of the controlled parameter is divided into three categories: increase, remain unchanged, and decrease. When the adjustment amount is within an adjustment step length △cp, it is regarded as an adjustment to keep the controlled parameter unchanged. When the adjustment step length △cp is increased, it is regarded as an adjustment to increase the controlled parameter. When the adjustment step length △cp is decreased, it is regarded as an adjustment to decrease the controlled parameter.

[0075] Specifically,

[0076] When the controlled parameter is changed from cp t Adjust to the interval [cp t -△cp,cp t +△cp], it is an adjustment to keep the controlled parameter unchanged, which is a single action;

[0077] When the controlled parameter is changed from cp t Adjust to the interval (cp t +△cp,cp t +2×△cp], it is to increase the adjustment step length of the controlled parameter, which is another action;

[0078] When the controlled parameter is changed from cp t Adjust to the interval (cp t +2×△cp,cp t +3×△cp], it is another action to increase the adjustment step length of the controlled parameter;

[0079] And so on.

[0080] When the controlled parameter is changed from cp t Adjust to the interval (cp t +p×△cp,cp t +(p+1)×△cp], it is to increase the adjustment step length of p to the controlled parameter, which is another action;

[0081] Similarly,

[0082] When the controlled parameter is changed from cp t Adjust to the interval (cp t -2×△cp,cp t -△cp], it is an adjustment to reduce the controlled parameter by one adjustment step, which is another action;

[0083] When the controlled parameter is changed from cp t Adjust to the interval (cp t -3×△cp,cp t -2×△cp], it is another action to reduce the controlled parameter by two adjustment steps;

[0084] And so on.

[0085] When the controlled parameter is changed from cp t Adjust to the interval (cp t -(p+1)×△cp,cp t -p×△cp], it is another action to reduce the controlled parameter by p adjustment steps.

[0086] Step S33: taking any adjustment of keeping the controlled parameter unchanged, any adjustment of increasing the controlled parameter, and any adjustment of decreasing the controlled parameter in step S32 as one action, and obtaining the classification result of the action;

[0087] Step S34: Collect all the actions once and obtain the action space A = {a1, a2, ..., a v}, in this embodiment, 2p+1=v.

[0088] Step 4: Extract any action a in the action space A λ State parameter time series data sequence when executed

[0089]

[0090] Step 5: Utilize the data series A Markov chain model of the fermentation state in the fermentation equipment is constructed, and the model is defined as: Represented as action a λ When executed, the probability that the fermentation state in the fermentation equipment is transferred from the i-th category to the j-th category in one step, where i, j∈u; λ∈v; S t Represented as action a λ The fermentation state in the fermentation equipment at the time t being executed; S t+1 Represented as action a λ The fermentation state in the fermentation equipment at time t+1 after a time step is executed; where the probability The calculation formula is:

[0091]

[0092] In the formula, Represented as a data sequence In the case of action a λ When executed, the frequency at which the fermentation state in the fermentation equipment is transferred from the i-th category to the j-th category in one step.

[0093] Step 6: Define any two fermentation states s in the state space S i 、s j The conversion value between i →s j ),like Figure 3The calculation process of expected conversion value is shown in Figure 1. λ The expected value of the transformation after k time steps at time t is G t+k , the calculation formula is: In the formula, γ represents the attenuation factor, and γ∈(0,1); R t+k (s i →s j ) is represented as the conversion value at time step t+k;

[0094] Step 7: Select the expected conversion value G in the kth time step t+k The maximum value G in the calculation result (t+k)max The corresponding action is taken as an action at the kth time step, and a determined action is executed to adjust the controlled parameters at the kth time step.

[0095] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A fully automatic intelligent control method for fermentation equipment, characterized in that: include: Obtain the time series data of state parameters and controlled parameters in the historical fermentation process; The fermentation state in the fermentation equipment is classified according to the state parameter data to obtain a state space S={s1, s2, ..., s u },s u It is represented as the u-th type of fermentation state in the fermentation equipment; the adjustment of the controlled parameter is defined as an action, and the actions are classified according to the adjustment amount of the controlled parameter to obtain the action space A = {a1, a2, ..., a v }, a v It is represented as the vth type of action; Extract any action a in the action space A λ State parameter time series data sequence when executed Using data series A Markov chain model of the fermentation state in the fermentation equipment is constructed, and the model is defined as: Represented as action a λ When executed, the probability that the fermentation state in the fermentation equipment is transferred from the i-th category to the j-th category in one step, where i, j∈u; λ∈v; S t Represented as action a λ The fermentation state in the fermentation equipment at the time t being executed; S t+1 Represented as action a λ The fermentation state in the fermentation equipment at time t+1 after a time step is executed; Defined in the state space S, any two of the fermentation states s i 、s j The conversion value between i →s j ), calculate action a λ The expected value of the transformation after k time steps at time t is G t+k , the calculation formula is: In the formula, γ represents the attenuation factor, and γ∈(0,1); R t+k (s i →s j ) is represented as the conversion value at time step t+k; Select the kth time step, the expected transformation value G t+k The maximum value G in the calculation result (t+k)max The corresponding action is taken as an action at the k-th time step, and the determined action is executed to adjust the controlled parameter at the k-th time step.

2. A fully automatic intelligent control method for fermentation equipment according to claim 1, characterized in that: The state parameter includes at least one of the temperature, oxygen content and moisture content in the fermentation equipment; the controlled parameter is the ventilation volume in the fermentation equipment.

3. A fully automatic intelligent control method for fermentation equipment according to claim 1, characterized in that: The fermentation state classification process comprises the following steps: Step S11: Set the optimal value of the state parameter of the rth category at time t to The qth sampling data value of the rth type of state parameter in the historical fermentation process is Step S12: Calculate the q sampling data values ​​of the state parameters respectively The optimal value of the state parameter The distance between The calculation formula is: To obtain Construct a data sample set, denoted as Step S13: Use the Sigmoid function to adjust the mapping of each element in the acquired data sample set to between [0,1], and obtain the Sigmoid function value corresponding to each element Using function values The elements are classified according to the following principles: when hour, For the first category; when hour, For the second category; when hour, For the third category; And so on. when hour, It is the μth category; in, They are the minimum and maximum values ​​of the Sigmoid function of the elements in the data sample set; Step S14: According to The classification of the state parameter is used to determine the qth state parameter sampling data value. The corresponding fermentation states are as follows: when When it is the first type, the state parameter The corresponding fermentation state is the first type; when For the second type, the state parameter The corresponding fermentation state is the second type; when When it is the third type, the state parameter The corresponding fermentation state is the third category; And so on. when When it is the μth type, the state parameter The corresponding fermentation state is class μ.

4. A fully automatic intelligent control method for fermentation equipment according to claim 1, characterized in that: The action classification process includes the following steps: Step S21: Set the controlled parameter at time t to cp t , the adjustment step size is △cp; Step S22: the adjustment of the controlled parameter is divided into three categories: increase, remain unchanged, and decrease. When the adjustment amount is within an adjustment step length △cp, it is regarded as an adjustment to keep the controlled parameter unchanged. When the adjustment step length △cp is increased, it is regarded as an adjustment to increase the controlled parameter. When the adjustment step length △cp is decreased, it is regarded as an adjustment to decrease the controlled parameter. Step S23: taking any adjustment of keeping the controlled parameter unchanged, increasing the controlled parameter, or decreasing the controlled parameter in step S22 as one action, and obtaining a classification result of the action; Step S24: Collect all the actions once and obtain the action space A = {a1, a2, ..., a v }.

5. A fully automatic intelligent control method for fermentation equipment according to claim 1, characterized in that: Action a λ When executed, the probability that the fermentation state in the fermentation equipment is transferred from the i-th category to the j-th category in one step The calculation formula is: In the formula, Represented as a data sequence In the case of action a λ When executed, the frequency at which the fermentation state in the fermentation equipment is transferred from the i-th category to the j-th category in one step.

6. A fully automatic intelligent control system for fermentation equipment, characterized in that: The system is used to implement the steps of a fully automatic intelligent control method for fermentation equipment according to any one of claims 1 to 5, including: A data acquisition module, used to obtain the time series data of state parameters and controlled parameters in the historical fermentation process; A first data processing module is used to classify the fermentation state in the fermentation equipment according to the state parameter data to obtain a state space S; A second data processing module is used to define the adjustment of the controlled parameter as an action, classify the actions according to the adjustment amount of the controlled parameter, and obtain an action space A; Model building module, used to extract any action a in the action space A λ State parameter time series data sequence when executed Using data series Construct a Markov chain model of the fermentation state in the fermentation equipment; The conversion value expectation calculation module is used to define in the state space S, any two of the fermentation states s i 、s j The conversion value between i →s j ), and calculate action a λ The expected value of the transformation after k time steps at time t is G t+k ; An action determination module is used to select the expected conversion value G in the kth time step t+k The maximum value G in the calculation result (t+k)max The corresponding action is taken as an action at the kth time step; The action execution module is used to execute a determined action and adjust the controlled parameter at the kth time step.

7. A fully automatic intelligent control system for fermentation equipment according to claim 6, characterized in that: The system also includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can implement the steps of a fully automatic intelligent control method for fermentation equipment according to any one of claims 1 to 5 when executing the program.

Images