Calculation method, system, server and storage medium for sludge treatment control parameters

Abstract

The embodiment of the present invention discloses a method, system, server and storage medium for calculating sludge treatment control parameters, wherein the method includes: obtaining status information of each component in the sludge treatment equipment; generating an initial component structure diagram based on the relationship between the components; using the convolution layer of the graph neural network to update and predict the nodes of the initial component structure diagram; using the random fluctuation layer of the graph neural network to form the uncertain variation of the equipment, and inputting the fluctuation feature matrix containing the deterministic variation and the uncertain variation into the deep reinforcement learning model to obtain the control parameters of the sludge treatment equipment. The uncertain variation is added to the deterministic variation of the sludge treatment equipment through the random fluctuation layer of the graph neural network, so that the graph neural network model can adapt to the influence of the external environment and emergencies to predict the update status of the sludge treatment equipment, obtain the control parameters to adjust the sludge treatment equipment, reduce the energy consumption of the equipment, and improve the sludge treatment efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of sludge treatment control, and in particular to a calculation method, system, server and storage medium for sludge treatment control parameters. Background Art

[0002] With the rapid development of industrialization and urbanization, sludge treatment has gradually become a key issue for environmental protection and sustainable development. Using electroosmosis and vacuum preloading methods to treat sludge can achieve a higher dehydration rate, but it will generate relatively high energy consumption. By introducing nanomaterials in the process of electroosmosis vacuum preloading sludge treatment, the pollutant removal efficiency can be effectively improved and the required energy consumption can be reduced. However, the application effect of nanomaterials will be affected by many factors, such as the voltage and current of electroosmosis, and the pressure of vacuum preloading.

[0003] The operating parameters of the electroosmotic vacuum preloading sludge treatment equipment can be adjusted according to the state to improve the sludge treatment efficiency and reduce the energy consumption of the sludge treatment equipment. However, due to the large number of influencing factors involved in the electroosmotic vacuum preloading sludge treatment equipment, it is necessary to deduce the control parameters used to adjust the operating parameters of the sludge treatment equipment according to the various states of each component in the sludge treatment equipment. By using an intelligent model to analyze the state of the sludge treatment equipment, the operating parameters can be controlled in time according to the various states of each component in the sludge treatment equipment. However, the intelligent model cannot predict the equipment state according to uncertain factors such as environmental climate change and equipment component failure during the sludge treatment process, resulting in the generated control parameters being unable to accurately adjust the operating parameters according to the state of the sludge treatment equipment. Summary of the invention

[0004] The embodiments of the present invention provide a method, system, server and storage medium for calculating sludge treatment control parameters to solve the technical problem of accurately controlling operating parameters in the sludge treatment process.

[0005] In a first aspect, an embodiment of the present invention provides a method for calculating sludge treatment control parameters, comprising:

[0006] Acquire status information of each component in the sludge treatment equipment, the status information including historical status information and current status information, generate a component feature matrix of each component based on the status information of each component, the component feature matrix reflects the multidimensional characteristics of each component, and the multidimensional characteristics represent the current state, change trend and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period;

[0007] Generate an initial component structure diagram based on the positional relationship and connection relationship between the components;

[0008] Inputting the initial component structure diagram into the trained graph neural network, and using the graph neural network to embed the component feature matrix into the node corresponding to the component;

[0009] Utilizing the convolutional layer of the graph neural network to aggregate component vectors between adjacent nodes for each node, updating and predicting the component feature vectors of the nodes, and obtaining a prediction feature matrix for each node;

[0010] Using the random fluctuation layer of the graph neural network to form a diffusion term based on the predicted feature matrix of each node and the diffusion coefficient, the diffusion term represents the random fluctuation when the state of the device is updated, and using the diffusion term and the Wiener process to form the uncertain change amount of the state of the device, and forming the state fluctuation feature matrix of the device based on the deterministic change amount and the uncertain change amount;

[0011] The state fluctuation characteristic matrix of the equipment is input into the trained deep reinforcement learning model to obtain the control parameters of the sludge treatment equipment.

[0012] Furthermore, the random fluctuation layer of the graph neural network is used to form a diffusion term according to the predicted feature matrix of each node and the diffusion coefficient, the diffusion term represents the random fluctuation when the device state is updated, and the diffusion term and the Wiener process are used to form the uncertain change of the device, and the state fluctuation feature matrix of the device is formed according to the deterministic change and the uncertain change of the device, including:

[0013] A drift term is formed according to the prediction feature matrix of each node and the drift coefficient, wherein the drift term represents a regular change when the device state is updated, and a deterministic change amount of the device is obtained by multiplying the drift term by a preset time increment;

[0014] The uncertain variation of the equipment is obtained by multiplying the diffusion term by a Wiener process of a preset time increment, wherein the Wiener process represents the uncertainty factor of the sludge treatment equipment;

[0015] According to the sum of the deterministic variation and the uncertain variation, a state fluctuation characteristic matrix of the device at a preset time increment is obtained, and the state fluctuation characteristic matrix reflects the variation of the device under the combined influence of regular changes and random fluctuations.

[0016] Furthermore, the convolutional layer of the graph neural network is used to aggregate component vectors between adjacent nodes for each node, and the component feature vectors of the node are updated and predicted to obtain a prediction feature matrix of each node, including:

[0017] The average component feature matrix of the neighboring nodes is formed using the neighboring nodes of each node and the aggregation function;

[0018] The predicted feature matrix of each node is obtained based on the average component feature matrix and the component vector of each node and adding random noise.

[0019] Furthermore, generating a component feature matrix of each component based on the status information of each component includes:

[0020] Using the status information of each component in the sludge treatment equipment, the historical status and current status of the sludge fluid and electric field at the location of each component are extracted;

[0021] forming a change trend and periodic characteristics of the sludge fluid and the electric field according to the historical state and the current state;

[0022] A component characteristic matrix is ​​formed according to the current state as well as the change trend and periodic characteristics.

[0023] Furthermore, the obtaining of status information of each component in the sludge treatment equipment includes:

[0024] Collect sensor information from various components in the sludge treatment equipment;

[0025] Eliminating abnormal values ​​and noise from the sensor information;

[0026] Perform data smoothing and standardization on sensor information.

[0027] Furthermore, the method further comprises:

[0028] The current state information of the sludge treatment equipment and the control parameters of the sludge treatment equipment output by the deep reinforcement learning model are used to perform gradient calculation to obtain the adjustment value of the strategy parameter;

[0029] The adjustment value is used to optimize the policy parameters of the deep reinforcement learning model.

[0030] Furthermore, generating an initial component structure diagram based on the positional relationship and connection relationship between the components includes:

[0031] The components are used as nodes, and the connection relationships between the components are used as edges. The edges reflect the physical connection states between the components, and an initial component structure diagram is generated based on the positional relationships and connection relationships between the components.

[0032] In a second aspect, an embodiment of the present invention further provides a system for calculating sludge treatment control parameters, comprising:

[0033] Data acquisition module, used to obtain status information of each component in the sludge treatment equipment;

[0034] A feature generation module, used for generating a component feature matrix of each component based on the status information of each component;

[0035] A diagram generation module, used to generate an initial component structure diagram based on the position relationship and connection relationship between the components;

[0036] A graph neural network module, used for generating a state fluctuation characteristic matrix of a device according to the initial component structure diagram;

[0037] A deep reinforcement learning module is used to generate control parameters of the sludge treatment equipment based on the state fluctuation characteristic matrix of the equipment.

[0038] In a third aspect, an embodiment of the present invention further provides a server, including:

[0039] one or more processors;

[0040] a storage device for storing one or more programs,

[0041] When the one or more programs are executed by the one or more processors, the one or more processors implement the method for calculating the sludge treatment control parameters provided in the above embodiment.

[0042] In a fourth aspect, an embodiment of the present invention further provides a storage medium comprising computer executable instructions, which, when executed by a computer processor, are used to execute the method for calculating the sludge treatment control parameters provided in the above embodiment.

[0043] The embodiments of the present invention provide a method, system, server and storage medium for calculating sludge treatment control parameters. Based on the prediction feature matrix of each node obtained by the convolution layer of the graph neural network through the random fluctuation layer of the graph neural network, the uncertain variation formed by the diffusion term and the Wiener process is added to the deterministic variation of the sludge treatment equipment using a random differential equation, so that the graph neural network model has the ability to consider unpredictable factors when predicting, can adapt to the impact of the external environment and emergencies on the sludge treatment equipment, can simulate the random dynamic changes of the sludge treatment equipment caused by unpredictable factors in reality, and then use the deep reinforcement learning model to generate the control parameters of the sludge treatment equipment, which can more accurately adjust the operating parameters of the sludge treatment equipment, thereby reducing the energy consumption of the sludge treatment equipment and improving the efficiency of the sludge treatment equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0045] Figure 1 This is a flow chart of a method for calculating sludge treatment control parameters according to Embodiment 1 of the present invention;

[0046] Figure 2 This is a flow chart of a method for calculating sludge treatment control parameters according to Embodiment 2 of the present invention;

[0047] Figure 3 This is a flow chart of a method for calculating sludge treatment control parameters according to Embodiment 3 of the present invention;

[0048] Figure 4 This is a flow chart of a method for calculating sludge treatment control parameters according to Embodiment 4 of the present invention;

[0049] Figure 5 It is a flow chart of a method for calculating a sludge treatment control parameter described in an embodiment of the present invention;

[0050] Figure 6 This is a flow chart of a method for calculating sludge treatment control parameters according to Embodiment 6 of the present invention;

[0051] Figure 7 This is a flow chart of a method for calculating sludge treatment control parameters according to Embodiment 7 of the present invention;

[0052] Figure 8 This is a schematic diagram of the structure of a system for calculating sludge treatment control parameters according to Embodiment 8 of the present invention;

[0053] Fig. 9 This is a schematic diagram of the structure of a server described in Embodiment 9 of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0055] Embodiment 1

[0056] Figure 1 This is a flow chart of a method for calculating a sludge treatment control parameter according to Embodiment 1 of the present invention. This embodiment takes into account the deterministic change and the uncertain change of the sludge treatment equipment, combines the deterministic change and the uncertain change to accurately predict the update state of the sludge treatment equipment, and generates a control parameter according to the predicted update state to adjust the operating parameters of the sludge treatment equipment. Specifically, the method includes the following steps:

[0057] S110, obtaining status information of each component in the sludge treatment equipment, wherein the status information includes historical status information and current status information, and generating a component feature matrix for each component based on the status information of each component, wherein the component feature matrix reflects multidimensional characteristics of each component, and the multidimensional characteristics represent the current state, change trend, and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period.

[0058] Electroosmotic vacuum preloading sludge treatment equipment is a device that uses electroosmotic action and preloading pressure to treat sludge. The electroosmotic vacuum preloading sludge treatment equipment is composed of a variety of electrodes that provide voltage, a pump that provides pressure, and pipes for sludge flow. In the electroosmotic vacuum preloading sludge treatment equipment, the effect of sludge treatment is affected by many factors. Therefore, various sensors for collecting information such as sludge fluid dynamics, electric field distribution, and sludge physical and chemical properties are provided at each component of the sludge treatment equipment, including voltage sensors, current sensors, pressure sensors, physical sensors, PH sensors, conductivity sensors, turbidity sensors, gas analysis sensors, etc., so as to obtain the voltage and current of the electroosmotic action at the location of each component in the sludge treatment equipment, the pressure of vacuum preloading, the fluid characteristics of the sludge, the water content, the density, and other information. As the sludge treatment equipment is running, the data collected by various sensors can be used to analyze the sludge state at the location of each component in the sludge treatment equipment, and the current sludge fluid characteristics, voltage and current, vacuum preload pressure and sludge physical and chemical properties at the location of each component can be obtained. The historical data of the sensor can also be analyzed to obtain the sludge state change trend and periodic characteristics at the location of each component. After the sludge treatment equipment has been running for a period of time, the received various sensor data are cleaned and preprocessed, and various features of each component in the sludge treatment equipment are extracted from them, and the component feature matrix of each component is formed using the various features, and the component feature matrix contains the features of the component in multiple dimensions.

[0059] S120, generating an initial component structure diagram based on the positional relationship and connection relationship between the components.

[0060] In order to better reflect the status of the sludge treatment equipment, an initial component structure diagram close to the structure of the sludge treatment equipment is established. The initial component structure diagram includes multiple nodes and edges. The relationship between the nodes reflects the positional relationship between the components in the sludge treatment equipment, and the edges reflect the interaction between the components in the sludge treatment equipment. The initial component structure diagram can express the change relationship between the components in the sludge treatment equipment and the mutual influence between the components.

[0061] S130, inputting the initial component structure diagram into the trained graph neural network, and using the graph neural network to embed the component feature matrix into the node corresponding to the component.

[0062] Graph neural networks have the ability to process data containing complex relationships and structures. Graph neural networks can be used to capture the complex relationships and interactions between nodes, and directly learn the representation of the status of sludge treatment equipment by nodes and graph structures from the initial component structure diagram. Using the correspondence between the nodes in the initial component structure diagram and the components in the sludge treatment equipment, the component feature matrix extracted from each component is embedded into the corresponding node in the initial component structure diagram, and the graph neural network uses the initial component structure diagram and the nodes representing the status of each component in the diagram to predict the status of the sludge treatment equipment.

[0063] In this embodiment, the graph neural network is trained using historical data and simulation data of the sludge treatment equipment. The historical data of the sludge treatment equipment includes multiple operation data of the sludge treatment equipment in the past. Each operation data can reflect the changes in the sludge state at the locations of each component in a single sludge treatment process. The graph neural network is trained using multiple operation data until the training is completed. The trained graph neural network can accurately predict the state changes of the sludge treatment equipment.

[0064] S140, utilizing the convolutional layer of the graph neural network to aggregate component vectors between adjacent nodes for each node, update and predict the component feature vectors of the nodes, and obtain a predicted feature matrix for each node.

[0065] The convolutional layer of the graph neural network is used to aggregate neighboring nodes to predict node updates. The trained graph neural network can accurately obtain the node update status and present it in a prediction feature matrix, indicating the component status of the sludge treatment equipment corresponding to the node. In the aggregation process, the aggregation function can be used to integrate the information of neighboring nodes, and according to the influence of neighboring nodes on the current node, the corresponding weight matrix is ​​assigned to the integrated neighboring node information and the current node information, and the current node is updated and predicted to obtain the prediction feature matrix of the current node.

[0066] In this embodiment, the weight matrix assigned to the integrated neighbor node information and the current node information is a learnable weight matrix. When training the graph neural network, the weight matrix can be optimized by minimizing the difference between the real-time data and the predicted data using the back propagation algorithm. Exemplarily, the cosine similarity is used to calculate the similarity between the current node and each neighbor node. The higher the similarity, the greater the influence of the neighbor node on the current node. The Softmax function is used to convert the similarity into a weight.

[0067] S150, using the random fluctuation layer of the graph neural network to form a diffusion term based on the predicted feature matrix of each node and the diffusion coefficient, the diffusion term represents the random fluctuation when the device state is updated, and using the diffusion term and the Wiener process to form the uncertain change of the device, and forming the state fluctuation feature matrix of the device based on the deterministic change and uncertain change of the device.

[0068] The convolution layer of the graph neural network can predict the update status of the current node based on the neighboring nodes. However, due to the large number of influencing factors involved in the sludge treatment equipment, although the convolution layer can predict based on the various states of each component in the sludge treatment equipment, the sludge treatment equipment will have unpredictable factors due to the dynamic changes in the environment and sudden failures of equipment components. The convolution layer cannot consider the unpredictable factors only by aggregating the information of neighboring nodes. Since the environment described by the sludge treatment equipment is dynamically changing, the sludge treatment equipment may be in a state of emergency due to weather reasons, equipment structure reasons, debris contained in the sludge, etc., as well as operating failures of components in the equipment, data collection failures of sensors, and transmission failures, etc., resulting in errors in local information in the equipment. The generation of unpredictable factors often has no strong correlation with neighboring nodes. Therefore, the convolution layer of the graph neural network cannot take into account the unpredictable factors in the sludge treatment equipment. It is necessary to introduce uncertain changes representing unpredictable factors to improve the prediction accuracy of the sludge treatment equipment.

[0069] In this embodiment, the diffusion term is formed by using the diffusion coefficient and the prediction feature matrix generated by the convolution layer. The diffusion term can only represent the random fluctuations when the equipment is updated. Combining the diffusion term with the Wiener process can reflect the uncertain changes when the equipment is updated. The time increment of the Wiener process can be determined according to the equipment update period that needs to be predicted, and then the uncertain changes generated when the equipment is updated during the update period are determined. The uncertain changes are combined with the deterministic changes generated when the equipment is updated during the update period by using a stochastic differential equation, and the state fluctuation feature matrix of the equipment containing the uncertain changes can be obtained. The state fluctuation matrix represents the overall state of the sludge inside the equipment after the sludge treatment equipment is updated. The state fluctuation feature matrix of the equipment includes consideration of unpredictable factors.

[0070] S160, inputting the state fluctuation characteristic matrix of the equipment into the trained deep reinforcement learning model to obtain the control parameters of the sludge treatment equipment.

[0071] The deep reinforcement learning model combines the perception ability of deep learning and the decision-making ability of reinforcement learning. It can make decisions based on the input fluctuation feature matrix representing the changing state of the equipment combined with its own trained strategy. In this embodiment, the decision made by the deep reinforcement learning model is the control parameter of the sludge treatment equipment. The control parameter is used to adjust the sludge treatment equipment. After the control parameter is input into the sludge treatment equipment, the operating parameters of the sludge treatment equipment can be adjusted, thereby adjusting the working state of each component in the sludge treatment equipment.

[0072] In this embodiment, a diffusion term representing random fluctuations is formed by a diffusion coefficient and a prediction feature matrix, and a stochastic differential equation is used to combine the uncertain change formed by the diffusion term and the Wiener process with the deterministic change when the sludge treatment equipment is updated, so that the model can consider the unpredictable factors of the equipment when predicting the state change of the sludge treatment equipment. The spatial perception ability of the graph structure and the time dynamic characteristics of the stochastic differential equation are used to accurately simulate the unpredictable dynamic changes of the sludge treatment equipment in reality, and then the deep reinforcement learning model is used to generate the control parameters of the sludge treatment equipment, which can more accurately adjust the operating parameters of the sludge treatment equipment, thereby reducing the energy consumption of the sludge treatment equipment and improving the efficiency of the sludge treatment equipment.

[0073] Embodiment 2

[0074] Figure 2 This is a flow chart of a method for calculating sludge treatment control parameters described in Example 2 of the present invention. This embodiment is optimized based on the above embodiment. In this embodiment, the random fluctuation layer of the graph neural network is used to form a diffusion term according to the predicted feature matrix and diffusion coefficient of each node. The diffusion term represents the random fluctuation when the device state is updated, and the diffusion term and the Wiener process are used to form the uncertain variation of the device. The state fluctuation feature matrix of the device is formed according to the deterministic variation and uncertain variation of the device. The specific optimization is:

[0075] A drift term is formed according to the prediction feature matrix of each node and the drift coefficient, wherein the drift term represents a regular change when the device state is updated, and a deterministic change amount of the device is obtained by multiplying the drift term by a preset time increment;

[0076] The uncertain variation of the equipment is obtained by multiplying the diffusion term by a Wiener process of a preset time increment, wherein the Wiener process represents the uncertainty factor of the sludge treatment equipment;

[0077] According to the sum of the deterministic variation and the uncertain variation, a device fluctuation characteristic matrix of the device at a preset time increment is obtained, and the device fluctuation characteristic matrix reflects the variation of the device under the combined influence of regular variation and random fluctuation.

[0078] Accordingly, the calculation method of the sludge treatment control parameters provided in this embodiment specifically includes:

[0079] S210, obtaining status information of each component in the sludge treatment equipment, wherein the status information includes historical status information and current status information, and generating a component feature matrix for each component based on the status information of each component, wherein the component feature matrix reflects multidimensional characteristics of each component, and the multidimensional characteristics represent the current state, change trend, and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period.

[0080] S220: Generate an initial component structure diagram based on the positional relationship and connection relationship between the components.

[0081] S230, inputting the initial component structure diagram into the trained graph neural network, and using the graph neural network to embed the component feature matrix into the node corresponding to the component.

[0082] S240, utilizing the convolutional layer of the graph neural network to aggregate component vectors between adjacent nodes for each node, update and predict the component feature vectors of the nodes, and obtain a predicted feature matrix for each node.

[0083] S250, forming a drift term according to the prediction feature matrix of each node and the drift coefficient, wherein the drift term represents the regular change when the device state is updated, and obtaining the deterministic change amount of the device by multiplying the drift term by a preset time increment.

[0084] The predicted update state of the sludge treatment equipment at the current moment represented by the prediction feature matrix of each node and the drift coefficient form a drift term. The drift coefficient describes the change trend of the sludge treatment equipment. Therefore, the drift term can represent the regular change trend of the sludge treatment equipment when it is updated. The change amount of the drift term in a certain time increment is obtained according to the regular change trend of the drift term, as the deterministic change amount of the sludge treatment equipment in a certain time increment.

[0085] S260, obtaining the uncertain variation of the equipment by multiplying the diffusion term by the Wiener process of a preset time increment, wherein the Wiener process represents the uncertainty factor of the sludge treatment equipment.

[0086] The predicted update state of the sludge treatment equipment at the current moment represented by the prediction feature matrix of each node and the diffusion coefficient form a diffusion term. The diffusion coefficient describes the irregular changes of the sludge treatment equipment under external influences. Therefore, the diffusion term can represent the unpredictable changes caused by external uncertain factors when the sludge treatment equipment is updated. External uncertain factors may be changes in the external environment of the sludge treatment equipment, local emergencies of the sludge treatment equipment, and component failures of the sludge treatment equipment. The Wiener process is a Brownian motion at a certain time increment. The Wiener process of the diffusion term at a certain time increment represents the random fluctuations of the diffusion term at a certain time increment in the manner of Brownian motion, as the uncertain change amount of the sludge treatment equipment at a certain time increment.

[0087] S270, obtaining a state fluctuation characteristic matrix of the device at a preset time increment based on the sum of the deterministic change amount and the uncertain change amount, wherein the state fluctuation characteristic matrix reflects the change amount of the device under the combined influence of regular changes and random fluctuations.

[0088] On the basis of the deterministic variation that represents the regular change trend, an uncertain variation that represents the unpredictable change is added to generate the equipment fluctuation characteristic matrix. The equipment characteristic matrix can reflect the variation that takes into account the unpredictable factors of the sludge treatment equipment and can more accurately represent the change state of the sludge treatment equipment. In this embodiment, the uncertain variation is added to the deterministic variation of the sludge treatment equipment using a stochastic differential equation. The calculation formula is as follows:

[0089]

[0090] in, is the device status, For time, is the drift coefficient, is the time increment, is the diffusion coefficient, is the Wiener process, is the change in equipment state, which is reflected as the equipment fluctuation characteristic matrix.

[0091] S280, inputting the equipment fluctuation characteristic matrix into the trained deep reinforcement learning model to obtain control parameters of the sludge treatment equipment.

[0092] This embodiment can more accurately predict the state of the sludge treatment equipment by adding uncertain changes representing unpredictable changes on the basis of the regular change trend of the sludge treatment equipment. The unpredictable factors in the working process of the sludge treatment equipment are taken into account in the prediction results. The control parameters obtained by using the equipment fluctuation characteristic matrix containing unpredictable factors and the deep reinforcement learning model can adapt to the fact that the sludge treatment equipment is affected by external uncertain factors in reality, and the operating parameters of the sludge treatment equipment can be adjusted more accurately, thereby improving the efficiency of the sludge treatment equipment and reducing energy consumption.

[0093] Embodiment 3

[0094] Figure 3 This is a flow chart of a method for calculating a sludge treatment control parameter according to Embodiment 3 of the present invention. This embodiment is optimized based on the above embodiment. In this embodiment, the convolutional layer of the graph neural network is used to aggregate the component vectors between adjacent nodes for each node, and the component feature vectors of the nodes are updated and predicted to obtain a prediction feature matrix of each node. The specific optimization is as follows:

[0095] The average component feature matrix of the neighboring nodes is formed using the neighboring nodes of each node and the aggregation function;

[0096] The predicted feature matrix of each node is obtained based on the average component feature matrix and the component vector of each node and adding random noise.

[0097] Accordingly, the calculation method of the sludge treatment control parameters provided in this embodiment specifically includes:

[0098] S310, obtaining status information of each component in the sludge treatment equipment, wherein the status information includes historical status information and current status information, and generating a component feature matrix for each component based on the status information of each component, wherein the component feature matrix reflects multidimensional characteristics of each component, and the multidimensional characteristics represent the current state, change trend, and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period.

[0099] S320: Generate an initial component structure diagram based on the positional relationship and connection relationship between the components.

[0100] S330, inputting the initial component structure diagram into the trained graph neural network, and using the graph neural network to embed the component feature matrix into the node corresponding to the component.

[0101] S340, using the neighboring nodes of each node and the aggregation function to form an average component feature matrix of the neighboring nodes.

[0102] In the convolutional layer of the graph neural network, the node prediction is updated by aggregating neighbor nodes, using the AGGREGATE aggregation function to collect information from neighbor nodes, and integrating the component features of neighbor nodes by weighted average to reflect the influence of neighbor nodes. Specifically, the component feature matrix of all neighbor nodes of the current node is averaged to obtain the average component feature matrix of the neighbor nodes of the current node.

[0103] S350, obtaining a prediction feature matrix of each node according to the average component feature matrix and the component vector of each node and adding random noise.

[0104] According to the influence of neighboring nodes on the current node, the corresponding weight matrix is ​​assigned to the component feature matrix of the neighboring nodes and the current node, the influence of different neighboring nodes is adjusted, and the previous information of the current node is retained. The node is predicted and updated, and Gaussian noise is added to simulate the random disturbance of the node. The above weighted aggregation results are nonlinearly transformed through the activation function to obtain the prediction feature matrix representing the node prediction update, completing this convolution process. The calculation formula is as follows:

[0105]

[0106] in, is the current node number, is the neighbor node number of the current node, For the current moment, is the activation function, For Node The set of neighbor nodes of and is the learnable weight matrix, is Gaussian noise, It is the feature representation of the node at a certain moment in time, specifically presented as a prediction feature matrix. The prediction feature matrix contains the interaction and influence of neighboring nodes on the current node, the change state of the current node itself, and the noise that the sensor set at the component corresponding to the node may collect during the working process.

[0107] In this embodiment, due to the different collection frequencies of various sensors in the sludge treatment equipment, when acquiring the sensor data of each component, some sensors of some components may not have collected data, resulting in the missing features corresponding to these sensors. Therefore, the component feature matrix corresponding to the component obtained lacks some features, but other components that are connected to the component may still collect the missing features of the component. The interaction between the component that collects the missing features of the component and the component can be used, and the aggregation operation of the graph neural network can be used to aggregate the features collected by other components to the node corresponding to the component using the relationship between nodes and edges, and reflect them in the updated prediction feature matrix of the node corresponding to the component. The principle is that when the graph neural network makes predictions, the predicted value obtained is only related to the historical status of the current node and the neighboring nodes, that is, only historical data is used for prediction. The aggregation operation in the above calculation can integrate the features of the neighboring nodes into the current node. While retaining the historical information of the current node, the average component features of the aggregated neighboring nodes are used to fill in some features of the current node that were not collected during the data collection phase, so that the updated state of the node contains the uncollected features. That is, the interaction and influence between the components are used to infer the influence of other components on the missing features of the current component, and then the effects of other components on this component are used to fill in the missing features of the current component.

[0108] In this embodiment, in order to obtain more comprehensive information about the sludge treatment equipment, the convolution layer can be multi-layer, and the above-mentioned convolution process is performed once in each layer to perform a prediction update on the current node. Through multiple iterations, the current node can be aggregated to more and more comprehensive features, making the updated prediction result of the node more accurate and able to fully consider the complex influences between the components in the sludge treatment equipment.

[0109] S360, using the random fluctuation layer of the graph neural network to form a diffusion term based on the predicted feature matrix of each node and the diffusion coefficient, the diffusion term represents the random fluctuation when the device state is updated, and using the diffusion term and the Wiener process to form the uncertain change of the device, and forming the device fluctuation feature matrix based on the deterministic change and uncertain change of the device.

[0110] S370, inputting the equipment fluctuation characteristic matrix into the trained deep reinforcement learning model to obtain control parameters of the sludge treatment equipment.

[0111] This embodiment aggregates neighbor nodes in the convolutional layer of the graph neural network, and assigns different weights to the aggregated average component feature matrix and the component features of the current node according to the degree of influence of the neighbor nodes on the current node. That is, the influence of the neighbor nodes is taken into account and the previous information of the current node is retained. At the same time, Gaussian noise is added to simulate the random disturbance of the node, so as to enhance the robustness of the model, so that the model can still maintain the performance to predict and update the node status in the face of noise. In addition, when some features are missing in the component feature matrix of the current node due to the different frequencies in the acquisition stage, the features aggregated from the neighbor nodes are used to fill in the missing features, so as to make a more comprehensive and accurate prediction of the update status of the sludge treatment equipment.

[0112] Embodiment 4

[0113] Figure 4 This is a flow chart of a method for calculating sludge treatment control parameters according to Embodiment 4 of the present invention. This embodiment is optimized based on the above embodiment. In this embodiment, the component characteristic matrix of each component is generated based on the status information of each component, and is specifically optimized as follows:

[0114] Using the status information of each component in the sludge treatment equipment, the historical status and current status of the sludge fluid and electric field at the location of each component are extracted;

[0115] forming a change trend and periodic characteristics of the sludge fluid and the electric field according to the historical state and the current state;

[0116] A component characteristic matrix is ​​formed according to the current state as well as the change trend and periodic characteristics.

[0117] Accordingly, the calculation method of the sludge treatment control parameters provided in this embodiment specifically includes:

[0118] S410, obtaining status information of each component in the sludge treatment equipment, wherein the status information includes historical status information and current status information.

[0119] Sensors in the sludge treatment equipment are used to collect various status information of the location of each component. The collected information includes the current status information of the sludge treatment equipment and the previous status information during the operation of the sludge treatment equipment. The various status information includes data such as voltage, current, pressure, moisture content, density, pH value, conductivity, turbidity and gas concentration of the sludge at the location of each component.

[0120] S420, using the status information of each component in the sludge treatment equipment, extracting the historical status and current status of the sludge fluid and the electric field at the location of each component.

[0121] The sludge state information of each component location is extracted from the various state information obtained by the sensor. The sludge state information includes sludge fluid characteristics and sludge electric field characteristics. The sludge fluid characteristics include the velocity field and pressure distribution of the sludge fluid. The Navier-Stokes equation is used for calculation. The formula is as follows:

[0122]

[0123] in, is the density, is the symbol of partial derivative, is the velocity field, For pressure, is the dynamic viscosity, is the external force, represents the fluid velocity at time The rate of change on the fluid. Fluid velocity is proportional to the pressure gradient. A larger pressure gradient will push the fluid to flow at a higher speed. Fluid velocity is inversely proportional to viscosity. The higher the viscosity, the greater the flow resistance and the lower the fluid velocity. In addition, fluid velocity is also related to density and the shape of the flow pipe. The greater the fluid density, the greater the flow inertia and the lower the flow velocity. The smaller the diameter of the pipe, the greater the flow resistance and the lower the flow velocity.

[0124] The electric field characteristics of sludge include electric field intensity and electric field distribution, which are calculated using Maxwell equations. The formula is as follows:

[0125]

[0126] in, is the electric field strength, is the charge density, The electric field strength indicates the magnitude of the electric field force, which affects the movement speed of charged particles and the fluid speed. In the electroosmosis process, the fluid speed is proportional to the electric field strength. The increase in the electric field strength will increase the movement speed of charged particles and fluids, thereby increasing the overall speed of the fluid.

[0127] S430, forming a change trend and periodic characteristics of the sludge fluid and the electric field according to the historical state and the current state.

[0128] The sensor data also contains time series features corresponding to various state information. The short-term fluctuations of the time series are smoothed, and the sliding mean and standard deviation of the time series are used to capture the dynamic change trend of the time series at past time points. The frequency information of the sensor data is analyzed using time series and fast Fourier transform to identify the periodic change characteristics of the sensor data.

[0129] S440, forming a component characteristic matrix according to the current state, change trend and periodic characteristics. The component characteristic matrix reflects the multidimensional characteristics of each component, and the multidimensional characteristics represent the current state, change trend and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period.

[0130] The extracted sludge fluid characteristics and sludge electric field characteristics are combined with the change trends and periodic characteristics extracted from the time series and frequency information to form a component feature matrix. The component feature matrix contains the characteristics of the sludge in multiple dimensions at the location of each component, specifically the current state of the sludge fluid and electric field, as well as the change trends and change cycles of the sludge fluid and electric field in the previous period of time.

[0131] S450: Generate an initial component structure diagram based on the positional relationship and connection relationship between the components.

[0132] S460, inputting the initial component structure diagram into the trained graph neural network, and using the graph neural network to embed the component feature matrix into the node corresponding to the component.

[0133] S470, using the convolutional layer of the graph neural network to aggregate component vectors between adjacent nodes for each node, update and predict the component feature vectors of the nodes, and obtain a predicted feature matrix for each node.

[0134] S480, using the random fluctuation layer of the graph neural network to form a diffusion term based on the predicted feature matrix of each node and the diffusion coefficient, the diffusion term represents the random fluctuation when the device state is updated, and using the diffusion term and the Wiener process to form the uncertain change of the device, and forming the device fluctuation feature matrix based on the deterministic change and uncertain change of the device.

[0135] S490, inputting the equipment fluctuation characteristic matrix into the trained deep reinforcement learning model to obtain control parameters of the sludge treatment equipment.

[0136] This embodiment collects sensor data, extracts the sludge fluid characteristics and electric field characteristics of each component location, and uses the time series in the sensor data for analysis to obtain the change trend and change cycle of the sludge fluid and electric field. The component feature matrix is ​​used to reflect the multi-dimensional characteristics including the current state of the sludge fluid and electric field as well as the change trend and change cycle.

[0137] Embodiment 5

[0138] Figure 5This is a flow chart of a method for calculating a sludge treatment control parameter according to Embodiment 5 of the present invention. This embodiment is optimized based on the above embodiment. In this embodiment, the state information of each component in the sludge treatment equipment is obtained, and is specifically optimized as follows:

[0139] Collect sensor information from various components in the sludge treatment equipment;

[0140] Eliminating abnormal values ​​and noise from the sensor information;

[0141] Perform data smoothing and standardization on sensor information.

[0142] Accordingly, the calculation method of the sludge treatment control parameters provided in this embodiment specifically includes:

[0143] S510, collecting sensor information of each component in the sludge treatment equipment.

[0144] In order to fully understand the changes in the sludge treatment process, various sensors are installed on each component in the sludge treatment equipment to collect various data at the location of each component. The various data can be used for analysis to extract the required features to represent the state and changes of the sludge.

[0145] S520: Eliminate abnormal values ​​and noise from the sensor information.

[0146] Since the data acquired by the sensor may contain the influence of factors such as noise, it is necessary to clean the data acquired by the sensor, remove unreasonable outliers and noise, and extract reliable required features based on the cleaned data. For example, the Z-score method can be used to verify the data. Specifically, the mean and standard deviation of the data set are first calculated, and then the Z-score of each data point is calculated, that is, each point is subtracted from the mean and divided by the standard deviation, and the data points whose Z-score exceeds the preset threshold are regarded as outliers and are removed, thereby improving the data quality.

[0147] S530, performing data smoothing and standardization processing on the sensor information.

[0148] The time series in the acquired sensor data needs to be smoothed to reduce the impact of short-term fluctuations and noise. The time series and the frequency features obtained based on the time series through fast Fourier transform are standardized to eliminate dimensional differences and reduce the impact on the model. Exemplarily, when processing time series data of temperature, pressure and chemical properties, a sliding average method is used for smoothing. When normalizing data, the minimum-maximum normalization can be used. By subtracting the minimum value from the data of each data point and dividing it by the difference between the maximum and the minimum value, the data of all data points are compressed to between 0 and 1, eliminating the dimensional differences between different features, which can improve the training effect and stability of the model in subsequent use.

[0149] The status information includes historical status information and current status information. A component feature matrix of each component is generated based on the status information of each component. The component feature matrix reflects the multidimensional characteristics of each component. The multidimensional characteristics represent the current state, change trend and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period.

[0150] In this embodiment, the features can also be screened to improve the representativeness and importance of the features. For example, PCA screening can be used to map high-dimensional data to low-dimensional space through linear transformation, retain important variation information in the data, and screen out combinations of features including voltage, current, pressure, etc., and then analyze to obtain features representing processing efficiency or energy consumption. Polynomials can also be used to combine features to generate new combined features, capture nonlinear relationships between features, effectively reduce redundant information, and improve the expressiveness of the model.

[0151] S540: Generate an initial component structure diagram based on the positional relationship and connection relationship between the components.

[0152] S550, inputting the initial component structure diagram into the trained graph neural network, and using the graph neural network to embed the component feature matrix into the node corresponding to the component.

[0153] S560, using the convolutional layer of the graph neural network to aggregate component vectors between adjacent nodes for each node, update and predict the component feature vectors of the nodes, and obtain a predicted feature matrix for each node.

[0154] S570, using the random fluctuation layer of the graph neural network to form a diffusion term based on the predicted feature matrix of each node and the diffusion coefficient, the diffusion term represents the random fluctuation when the device state is updated, and using the diffusion term and the Wiener process to form the uncertain change of the device, and forming the device fluctuation feature matrix based on the deterministic change and uncertain change of the device.

[0155] S580, inputting the equipment fluctuation characteristic matrix into the trained deep reinforcement learning model to obtain control parameters of the sludge treatment equipment.

[0156] This embodiment cleans and preprocesses the data, removes abnormal data, eliminates dimensional differences between different features, ensures the accuracy and reliability of the data, and extracts features based on accurate and reliable data to improve the prediction accuracy and stability of the model.

[0157] Embodiment 6

[0158] Figure 6 This is a flow chart of a method for calculating a sludge treatment control parameter according to Embodiment 6 of the present invention. This embodiment is optimized based on the above embodiment. In this embodiment, the following steps may be added:

[0159] The current state information of the sludge treatment equipment and the control parameters of the sludge treatment equipment output by the deep reinforcement learning model are used to perform gradient calculation to obtain the adjustment value of the strategy parameter;

[0160] The adjustment value is used to optimize the policy parameters of the deep reinforcement learning model.

[0161] Accordingly, the calculation method of the sludge treatment control parameters provided in this embodiment specifically includes:

[0162] S610, obtaining status information of each component in the sludge treatment equipment, wherein the status information includes historical status information and current status information, and generating a component feature matrix for each component based on the status information of each component, wherein the component feature matrix reflects multidimensional characteristics of each component, and the multidimensional characteristics represent the current state, change trend, and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period.

[0163] S620: Generate an initial component structure diagram based on the positional relationship and connection relationship between the components.

[0164] S630, inputting the initial component structure diagram into the trained graph neural network, and using the graph neural network to embed the component feature matrix into the node corresponding to the component.

[0165] S640, using the convolutional layer of the graph neural network to aggregate component vectors between adjacent nodes for each node, update and predict the component feature vectors of the nodes, and obtain a predicted feature matrix for each node.

[0166] S650, using the random fluctuation layer of the graph neural network to form a diffusion term based on the predicted feature matrix of each node and the diffusion coefficient, the diffusion term represents the random fluctuation when the device state is updated, and using the diffusion term and the Wiener process to form the uncertain change of the device, and forming the device fluctuation feature matrix based on the deterministic change and uncertain change of the device.

[0167] S660, inputting the equipment fluctuation characteristic matrix into the trained deep reinforcement learning model to obtain control parameters of the sludge treatment equipment.

[0168] The control parameters generated by the deep reinforcement learning model can adjust the operating parameters of the sludge treatment equipment. For example, when the electroosmosis efficiency is detected to be reduced, which is specifically manifested as a decrease in current, the deep reinforcement learning model generates control parameters for increasing the current accordingly; when the moisture content of the sludge is detected to be higher than the set threshold, the deep reinforcement learning model generates control parameters for reducing the flow rate to enhance the dehydration effect of the sludge; when the sludge treatment temperature is detected to be lower than the optimal temperature range, the deep reinforcement learning model generates heating control parameters for increasing the temperature, and vice versa.

[0169] S670, using the current state information of the sludge treatment equipment and the control parameters of the sludge treatment equipment output by the deep reinforcement learning model to perform gradient calculation to obtain the adjustment value of the strategy parameter.

[0170] In the application process of the deep reinforcement learning model, the policy parameters of the model can be optimized to maximize the expected reward of the model and improve the adjustment accuracy of the control parameters output by the deep reinforcement learning model on the sludge treatment equipment. For example, the expected reward of the strategy is calculated by the action taken by the strategy on the current state and the advantage estimate of the action, and then the gradient of the expected reward is calculated to obtain the derivative of the policy parameter with respect to the expected reward, that is, the adjustment value of the policy parameter, which indicates the adjustment direction of the policy parameter to increase the expected reward.

[0171] S680: Optimize the strategy parameters of the deep reinforcement learning model using the adjustment value.

[0172] The old strategy parameters are optimized using the adjusted values ​​of the strategy parameters and the learning step size of the model to obtain new strategy parameters. The calculation formula is as follows:

[0173]

[0174] in, are the new strategy parameters, are the old policy parameters, is the learning step size, is the derivative of the strategy parameter with respect to the expected reward, reflecting the optimization direction of the strategy parameter. is the behavior of the current strategy, is the current state, For action, Estimated advantage.

[0175] This embodiment optimizes the policy parameters of the deep reinforcement learning model by combining the current state of the sludge treatment equipment obtained by the deep reinforcement learning model and the actions taken by the deep reinforcement learning model based on the current state to maximize the expected reward of the policy. The optimized policy can better adapt to different environmental changes and operating conditions, thereby improving the flexibility and response capabilities of the system, enabling the deep reinforcement learning model to make more efficient decisions, and the generated control parameters can improve the efficiency of the sludge treatment equipment and reduce the energy consumption of the sludge treatment equipment when adjusting the sludge treatment equipment.

[0176] Example 7

[0177] Figure 7 This is a flow chart of a method for calculating sludge treatment control parameters described in Example 6 of the present invention. This embodiment is optimized based on the above embodiment. In this embodiment, an initial component structure diagram is generated based on the positional relationship and connection relationship between the components. The specific optimization is: using the components as nodes, using the connection relationship between the components as edges, the edges reflect the physical connection status between the components, and generating the initial component structure diagram based on the positional relationship and connection relationship between the components.

[0178] Accordingly, the calculation method of the sludge treatment control parameters provided in this embodiment specifically includes:

[0179] S710, obtaining status information of each component in the sludge treatment equipment, wherein the status information includes historical status information and current status information, and generating a component feature matrix for each component based on the status information of each component, wherein the component feature matrix reflects multidimensional characteristics of each component, and the multidimensional characteristics represent the current state, change trend, and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period.

[0180] S720, using the components as nodes, using the connection relationships between the components as edges, where the edges reflect the physical connection states between the components, and generating an initial component structure diagram based on the positional relationships and connection relationships between the components.

[0181] In order to better reflect the relationship and interaction between the components in the sludge treatment equipment, the components in the sludge treatment equipment are used as nodes, and the connection relationship between the components is used as the edge between the connecting nodes. The edge reflects the physical connection relationship such as the electrical connection between the components and the flow connection of the sludge. The initial component structure diagram generated based on the position relationship and connection relationship of the components can better reflect the interaction between the nodes and more realistically reflect the situation of the sludge treatment equipment.

[0182] S730, input the initial component structure diagram into the trained graph neural network, and use the graph neural network to embed the component feature matrix into the node corresponding to the component.

[0183] S740, using the convolutional layer of the graph neural network to aggregate component vectors between adjacent nodes for each node, update and predict the component feature vectors of the nodes, and obtain a predicted feature matrix for each node.

[0184] S750, using the random fluctuation layer of the graph neural network to form a diffusion term based on the predicted feature matrix and diffusion coefficient of each node, the diffusion term represents the random fluctuation when the device state is updated, and using the diffusion term and the Wiener process to form the uncertain change of the device, and forming the device fluctuation feature matrix based on the deterministic change and uncertain change of the device.

[0185] S760, inputting the equipment fluctuation characteristic matrix into the trained deep reinforcement learning model to obtain control parameters of the sludge treatment equipment.

[0186] This embodiment uses the positions of the components in the sludge treatment equipment and the physical connections between the components as nodes and edges to generate an initial component structure diagram, which can better reflect the interactions between the nodes established through physical connections, more realistically reflect the situation of the sludge treatment equipment, and provide accurate data for model input.

[0187] Embodiment 8

[0188] Figure 8 FIG. 1 is a schematic diagram of a system for calculating sludge treatment control parameters according to Embodiment 8 of the present invention. Figure 8 As shown, the system comprises:

[0189] The data acquisition module 810 is used to obtain status information of each component in the sludge treatment equipment;

[0190] A feature generation module 820, configured to generate a component feature matrix of each component based on the status information of each component;

[0191] A diagram generation module 830, for generating an initial component structure diagram based on the positional relationship and connection relationship between the components;

[0192] A graph neural network module 840, for generating a state fluctuation characteristic matrix of the device according to the initial component structure diagram;

[0193] The deep reinforcement learning module 850 is used to generate control parameters of the sludge treatment equipment according to the state fluctuation characteristic matrix of the equipment.

[0194] The calculation system of the sludge treatment control parameters described in this embodiment obtains various state information of each component in the sludge treatment equipment, extracts the sludge fluid electric field characteristics of each component according to the various state information of each component, establishes an initial component structure diagram based on the positional relationship and connection relationship between the components, uses a graph neural network to generate a device fluctuation feature matrix representing the device state according to the initial component structure diagram, and uses a deep reinforcement learning model to generate the control parameters of the sludge treatment equipment according to the device fluctuation feature matrix. When using the graph neural network to update and predict the state of the sludge treatment equipment, the uncertain influence of external unpredictable factors on the sludge treatment equipment is added, and a prediction feature matrix containing uncertain changes can be obtained. The generated control parameters of the sludge treatment equipment can be combined with the actual emergency situation to adjust the operating parameters of the sludge treatment equipment, so that the sludge treatment equipment improves efficiency and reduces energy consumption.

[0195] Based on the above embodiments, the graph neural network module includes:

[0196] A graph convolution unit is used to form a drift term according to the prediction feature matrix of each node, and use the drift term to obtain a deterministic change amount at a preset time increment;

[0197] The random fluctuation unit is used to form a diffusion term according to the prediction feature matrix of each node, and use the diffusion term and the Wiener process to obtain the uncertain change amount at the preset time increment.

[0198] Based on the above embodiments, the graph convolution unit includes:

[0199] an aggregation subunit, for forming an average component feature matrix of adjacent nodes according to adjacent nodes of each node and an aggregation function;

[0200] The updating subunit is used to add random noise according to the average component feature matrix and the node's own component vector to obtain the predicted feature matrix of each node.

[0201] Based on the above embodiments, the feature generation module includes:

[0202] A physical feature extraction unit, used to extract the sludge fluid and electric field characteristics at the location of each component using the status information of each component in the sludge treatment equipment;

[0203] A statistical feature extraction unit, used for forming a change trend and periodic characteristics of the sludge fluid electric field according to the historical state and current state of the sludge fluid and the electric field;

[0204] The feature splicing unit is used to splice the sludge fluid and electric field characteristics with the change trend and periodic characteristics to obtain the component feature matrix of each component.

[0205] Based on the above embodiments, the data acquisition module includes:

[0206] A data acquisition unit for collecting sensor information of various components in the sludge treatment equipment;

[0207] A data cleaning unit, used to remove abnormal values ​​and noise from the sensor information;

[0208] The data preprocessing unit is used to perform data smoothing and standardization processing on sensor information.

[0209] Based on the above embodiments, the deep reinforcement learning module further includes:

[0210] An adjustment value calculation unit, used to generate an adjustment value of a strategy parameter;

[0211] The strategy parameter optimization unit is used to optimize the strategy parameters according to the strategy parameter adjustment value.

[0212] The calculation system of the sludge treatment control parameters provided in the embodiment of the present invention can execute the calculation method of the sludge treatment control parameters provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0213] Embodiment 9

[0214] Fig. 9 This is a schematic diagram of the structure of a server described in Embodiment 9 of the present invention. Fig. 9 A block diagram of an exemplary server 12 suitable for use in implementing embodiments of the present invention is shown. Fig. 9 The server 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0215] like Fig. 9 As shown, the server 12 is represented in the form of a general-purpose computing device. The components of the server 12 may include, but are not limited to: one or more processors or processing units;

[0216] System memory, bus 18 that connects various system components including system memory 28 and processing unit 16 .

[0217] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0218] The server 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the server 12, including volatile and non-volatile media, removable and non-removable media.

[0219] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The server 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Fig. 9 not shown, usually called a "hard drive"). Although Fig. 9 Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.

[0220] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28, such program modules 42 including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0221] The server 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), may communicate with one or more devices that enable a user to interact with the device / system / server 12, and / or may communicate with any device that enables the server 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the server 12 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the server 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the server 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0222] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system memory 28, such as implementing the calculation method of the sludge treatment control parameters provided in the embodiment of the present invention.

[0223] Embodiment 10

[0224] Embodiment 10 of the present invention further provides a storage medium comprising computer executable instructions, wherein the computer executable instructions, when executed by a computer processor, are used to execute the method for calculating the sludge treatment control parameters provided in the above embodiments.

[0225] The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device.

[0226] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0227] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0228] Computer program code for performing the operations of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0229] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for calculating sludge treatment control parameters, characterized in that: include: Acquire status information of each component in the sludge treatment equipment, the status information including historical status information and current status information, generate a component feature matrix of each component based on the status information of each component, the component feature matrix reflects the multidimensional characteristics of each component, and the multidimensional characteristics represent the current state, change trend and periodic characteristics of the sludge fluid electric field at the location of each component within a preset time period; Generate an initial component structure diagram based on the positional relationship and connection relationship between the components; Inputting the initial component structure diagram into the trained graph neural network, and using the graph neural network to embed the component feature matrix into the node corresponding to the component; Utilizing the convolutional layer of the graph neural network to aggregate component vectors between adjacent nodes for each node, updating and predicting the component feature vectors of the nodes, and obtaining a prediction feature matrix for each node; A drift term is formed according to the prediction feature matrix of each node and the drift coefficient, wherein the drift term represents a regular change when the device state is updated, and a deterministic change amount of the device is obtained by multiplying the drift term by a preset time increment; The uncertain variation of the equipment is obtained by multiplying the diffusion term by a Wiener process of a preset time increment, wherein the Wiener process represents the uncertainty factor of the sludge treatment equipment; According to the sum of the deterministic variation and the uncertain variation, a state fluctuation characteristic matrix of the device at a preset time increment is obtained, wherein the state fluctuation characteristic matrix reflects the variation of the device under the combined influence of regular variation and random fluctuation; The state fluctuation characteristic matrix of the equipment is input into the trained deep reinforcement learning model to obtain the control parameters of the sludge treatment equipment.

2. The method according to claim 1, characterized in that The convolutional layer of the graph neural network is used to aggregate component vectors between adjacent nodes for each node, and the component feature vectors of the node are updated and predicted to obtain a prediction feature matrix of each node, including: The average component feature matrix of the neighboring nodes is formed using the neighboring nodes of each node and the aggregation function; The predicted feature matrix of each node is obtained based on the average component feature matrix and the component vector of each node and adding random noise.

3. The method according to claim 1, characterized in that The generating a component feature matrix of each component based on the status information of each component includes: Using the status information of each component in the sludge treatment equipment, the historical status and current status of the sludge fluid and electric field at the location of each component are extracted; forming a change trend and periodic characteristics of the sludge fluid and the electric field according to the historical state and the current state; A component characteristic matrix is ​​formed according to the current state as well as the change trend and periodic characteristics.

4. The method according to claim 1, characterized in that: The obtaining of status information of each component in the sludge treatment equipment includes: Collect sensor information from various components in the sludge treatment equipment; Eliminating abnormal values ​​and noise from the sensor information; Perform data smoothing and standardization on sensor information.

5. The method according to claim 1, characterized in that The method further comprises: The current state information of the sludge treatment equipment and the control parameters of the sludge treatment equipment output by the deep reinforcement learning model are used to perform gradient calculation to obtain the adjustment value of the strategy parameter; The adjustment value is used to optimize the policy parameters of the deep reinforcement learning model.

6. The method according to claim 1, characterized in that An initial component structure diagram is generated based on the positional relationship and connection relationship between the components, including: The components are used as nodes, and the connection relationships between the components are used as edges. The edges reflect the physical connection states between the components, and an initial component structure diagram is generated based on the positional relationships and connection relationships between the components.

7. A calculation system for sludge treatment control parameters, characterized in that: include: Data acquisition module, used to obtain status information of each component in the sludge treatment equipment; A feature generation module, used for generating a component feature matrix of each component based on the status information of each component; A diagram generation module, used to generate an initial component structure diagram based on the position relationship and connection relationship between the components; A graph neural network module, used for generating a state fluctuation characteristic matrix of the device according to the initial component structure diagram; The graph neural network module includes: a graph convolution unit, which is used to form a drift term according to the prediction feature matrix of each node, and use the drift term to obtain a deterministic change amount at a preset time increment; a random fluctuation unit, which is used to form a diffusion term according to the prediction feature matrix of each node, and use the diffusion term and the Wiener process to obtain an uncertain change amount at a preset time increment; and a state fluctuation feature matrix is ​​generated using the deterministic change amount and the uncertain change amount of the initial component structure diagram; A deep reinforcement learning module is used to generate control parameters of the sludge treatment equipment based on the state fluctuation characteristic matrix of the equipment.

8. A server, characterized in that: The server comprises: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method for calculating the sludge treatment control parameters as described in any one of claims 1-6.

9. A storage medium comprising computer executable instructions, wherein the computer executable instructions are used to execute the method for calculating sludge treatment control parameters as claimed in any one of claims 1 to 6 when executed by a computer processor.

Images