Air compression station flow supply and demand gap calculation method based on multi-source data and dynamic game

Abstract

The invention discloses an air compression station flow supply and demand gap calculation method based on multi-source data and a dynamic game, and belongs to the technical field of industrial air compression station intelligent control. Constructing a pipeline pressure drop dynamic model based on the multi-source real-time data; establishing a workshop demand dynamic response function based on the multi-source real-time data, and further calculating an air compression station flow supply and demand gap; a non-cooperative game model of multiple air compressor units and multiple workshops is constructed, and the minimum total energy consumption of air compression stations serves as a target function; based on the air compression station flow supply and demand gap, an improved Shapley value method is adopted to distribute game benefits, a Q-learning strategy updating mechanism is combined to realize dynamic game parameter adaptive adjustment, and a gas production distribution scheme of multiple units is optimized, so that the method can reduce pressure drop calculation errors, improve supply and demand gap prediction accuracy, solve the problem of large matching errors of a traditional method, and improve the air compression station flow supply and demand gap prediction efficiency. And the comprehensive energy consumption of the air compression station is high.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of intelligent control of industrial air compression stations, and particularly relates to an air compression station flow supply-demand gap calculation method based on multi-source data and dynamic game. BACKGROUND

[0002] In industrial production, as a key power supply unit, the air compression station directly affects the production continuity and energy cost in terms of the matching accuracy of gas production and gas demand. However, the existing technology has many deficiencies:

[0003] Traditional air compression stations generally use single-point pressure monitoring and fixed pressure drop models, without considering the dynamic changes of local pipe resistance and the characteristics of multi-unit collaborative gas supply, resulting in large matching errors of gas production and gas demand; the existing flow prediction methods lack mechanism model constraints and have poor adaptability to complex pipe topologies, and cannot realize dynamic compensation of multi-workshop demand; at the same time, the existing resource allocation strategies ignore pressure energy loss and unit energy consumption characteristics, resulting in dynamic adjustment lag of supply-demand gap and low energy efficiency.

[0004] Related patents also have defects: the patent document with the application number CN202311030363.5 and the name "Integrated air compression station intelligent cloud control system and control method" only relies on sensor data acquisition, which is difficult to realize accurate matching of supply and demand; the patent document with the application number CN202110967818.0 and the name "Multi-air compression station dynamic collaborative optimization control system and method" does not consider the pressure drop superposition effect when multiple units supply gas collaboratively, resulting in large pressure drop calculation errors and affecting the control effect. SUMMARY

[0005] To solve the problems raised in the background art, the application provides an air compression station flow supply-demand gap calculation method based on multi-source data and dynamic game, to solve the problems of large matching errors of gas production and gas demand, inability to realize dynamic compensation of multi-workshop demand, and low energy efficiency in the prior art.

[0006] To achieve the above-mentioned purposes, the application provides the following technical solutions:

[0007] The air compression station flow supply-demand gap calculation method based on multi-source data and dynamic game comprises the following steps:

[0008] S1: Collecting multi-source real-time data of the gas production end and the gas demand end, the multi-source real-time data of the gas production end including the gas supply pressure and the gas supply amount of the air compression unit of the air compression station, and the multi-source real-time data of the gas demand end including the demand pressure P req , the environmental pressure P amb , and the basic data of the pipeline for flow transportation between the gas production end and the gas demand end, the basic data including the resistance coefficient, the topological structure and the roughness;

[0009] S2: Based on multi-source real-time data, a pipeline pressure drop dynamic model is constructed, the pipeline pressure drop dynamic model adopts Colebrook formula, an LSTM network is used to predict the resistance coefficient λ along the pipeline, the local resistance coefficient ξ is dynamically corrected based on the pipeline topological structure and roughness attenuation model, the fluid characteristic parameters and pipeline structure parameters are combined to calculate the dynamic constraint of gas production rate;

[0010] S3: Based on the demand pressure P req and the environmental pressure P amb , a workshop demand dynamic response function is established, the workshop demand dynamic response function considers the flow-pressure relationship of the pressure-sensitive equipment in the flow demand equipment in the workshop, and the real-time flow demand Q demand of the workshop is quantified based on the workshop demand dynamic response function;

[0011] S4: Based on the real-time flow demand Q demand of the workshop and the dynamic constraint of gas production rate, a pressure energy loss coefficient is introduced to calculate the flow supply-demand gap of the air compression station;

[0012] S5: A non-cooperative game model of multiple air compression units and multiple workshops is constructed, taking the minimum total energy consumption of the air compression station as the objective function, and the objective function includes the energy consumption item of the air compression unit and the pressure energy loss penalty item;

[0013] S6: Based on the flow supply-demand gap of the air compression station, an improved Shapley value method is used to allocate the game income, and a dynamic game parameter adaptive adjustment is realized by combining the Q-learning strategy update mechanism to optimize the gas production allocation scheme of multiple units.

[0014] Preferably, the expression of the pipeline pressure drop dynamic model in S2 is: ;

[0015] Wherein, ΔP is the pipeline pressure drop, Q is the volume flow rate in the pipeline, ρ is the fluid density, L is the pipeline length, D is the pipeline inner diameter, and A is the pipeline cross-sectional area.

[0016] Preferably, the expression of the workshop demand dynamic response function in S3 is:

[0017] ;

[0018] Wherein, k P is the pressure-flow coefficient, and m is the pressure-sensitive index.

[0019] Preferably, the calculation expression of the flow supply-demand gap of the air compression station in S4 is:

[0020] ;

[0021] Wherein, Let t be the flow supply and demand gap. For the real-time flow demand of the j-th workshop, Let L be the actual air supply of the i-th air compressor unit at the gas production end, α be the pressure energy loss coefficient, and L be the total length of the pipeline from the gas production end to the gas consumption end. i,j Let be the length of the pipeline segment from the i-th air compressor unit at the gas production end to the j-th workshop. This characterizes the impact of pipeline pressure drop on the effective gas supply from the corresponding unit to the workshop.

[0022] Preferably, the expression for the objective function in S5 is:

[0023] ;

[0024] Where N is the number of air compressor units, a i b i Let Q be the energy consumption characteristic parameter of the i-th unit. i Let be the gas production of the i-th unit, λ be the pressure energy loss penalty factor, M be the number of workshops, and P be the gas production of the i-th unit. j Let P be the actual pressure in the j-th workshop. target This is to create target pressure for the workshop.

[0025] Preferably, S1 further includes preprocessing the collected multi-source real-time data, the preprocessing including data cleaning, outlier removal and data standardization.

[0026] Preferably, in S2, after predicting the friction coefficient λ using an LSTM network and dynamically correcting the local resistance coefficient ξ based on the pipeline topology and roughness attenuation model, the friction coefficient λ and the local resistance coefficient ξ are further corrected using a Kalman filter algorithm.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] Significantly improved accuracy: The dynamic model of pipeline pressure drop, which integrates mechanism and data, combined with the dual correction of LSTM network and Kalman filter, reduces the error in pressure drop calculation; the dynamic response function of workshop demand accurately quantifies the flow demand of pressure-sensitive equipment, improves the accuracy of supply and demand gap prediction, and solves the problem of large matching error in traditional methods;

[0029] Significantly optimized energy consumption: Based on a non-cooperative game model involving multiple units and multiple workshops, with the goal of minimizing total energy consumption, and combining an improved Shapley value method and Q-learning strategy, dynamic optimization of unit gas production allocation is achieved, thereby reducing the overall energy consumption of the air compressor station. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating the process of this application. Detailed Implementation

[0031] To facilitate understanding of the technical content of this invention by those skilled in the art, the invention will be further described in detail below with reference to the accompanying drawings and specific examples. It should be understood that the specific examples described herein are merely illustrative and not intended to limit the scope of the invention.

[0032] Example 1

[0033] A method for calculating the supply and demand gap of air compressor station flow based on multi-source data and dynamic game theory, such as Figure 1 As shown, it includes the following steps:

[0034] S1: Deploy corresponding pressure sensors, flow sensors, and other detection devices through an edge computing gateway to collect multi-source real-time data from both the gas production and consumption ends. The multi-source real-time data from the gas production end includes the gas supply pressure and gas supply volume of all air compressor units in the air compressor station, while the multi-source real-time data from the gas consumption end includes the demand pressure P of each workshop. req Environmental pressure P amb The basic data of the pipeline used for flow transportation between the gas production end and the gas consumption end includes the initial resistance coefficient, topology and pipeline roughness.

[0035] S2: Based on multi-source real-time data, a dynamic model of pipeline pressure drop is constructed. The dynamic model of pipeline pressure drop adopts the Colebrook formula, predicts the friction coefficient λ through LSTM network, dynamically corrects the local resistance coefficient ξ based on pipeline topology and roughness attenuation model, and calculates the dynamic constraint of gas production by combining fluid characteristic parameters and pipeline structural parameters.

[0036] S3: Based on workshop demand pressure P req and environmental pressure P amb A dynamic demand response function for the workshop is established, which considers the flow-pressure relationship of pressure-sensitive equipment among the flow-demand devices in the workshop. Based on the dynamic demand response function, the real-time flow demand Q of the workshop is quantified. demand ;

[0037] S4: Based on the real-time flow demand Q of the workshop demand With dynamic constraints on gas production, a pressure energy loss coefficient is introduced to calculate the supply and demand gap of air compressor station flow.

[0038] S5: Construct a non-cooperative game model of multiple air compressor units and multiple workshops, with the objective function being to minimize the total energy consumption of the air compressor station. The objective function includes the energy consumption term of the air compressor unit and the pressure energy loss penalty term.

[0039] S6: Based on the supply and demand gap of air compressor station flow, the improved Shapley value method is used to allocate game payoffs, and the Q-learning strategy update mechanism is combined to realize the dynamic adaptive adjustment of game parameters and optimize the gas production allocation scheme of multiple units.

[0040] In this embodiment, this application reduces the pressure drop calculation error by using a mechanism-data fusion pipeline pressure drop dynamic model, combined with dual correction of LSTM network and Kalman filter; the workshop demand dynamic response function accurately quantifies the flow demand of pressure-sensitive equipment, improving the accuracy of supply and demand gap prediction and solving the problem of large matching error in traditional methods; based on a non-cooperative game model of multiple units and multiple workshops, this application aims to minimize total energy consumption, and combines the improved Shapley value method and Q-learning strategy to achieve dynamic optimization of unit gas production allocation, thereby reducing the overall energy consumption of the air compressor station.

[0041] Example 2

[0042] The difference between this embodiment and Embodiment 1 is that the dynamic model of pipeline pressure drop in S2 is expressed as follows:

[0043] ;

[0044] Where ΔP is the pressure drop in the pipeline, Q is the volumetric flow rate in the pipeline, ρ is the fluid density, L is the pipeline length, D is the pipeline inner diameter, and A is the pipeline cross-sectional area.

[0045] In this embodiment, the pipeline pressure drop dynamic model can accurately calculate the pressure loss caused by friction resistance. At the same time, a local resistance coefficient ξ is introduced to consider the local pressure loss caused by pipeline accessories (such as elbows, valves, tees, etc.). The two are superimposed to obtain the total pipeline pressure drop ΔP. This model predicts the friction resistance coefficient through a data-driven method (LSTM network) and corrects the local resistance coefficient by combining the mechanism model (pipeline topology and roughness attenuation model), so as to realize the dynamic and accurate calculation of pressure drop and provide reliable pressure drop data support for subsequent stages.

[0046] Example 3

[0047] The difference between this embodiment and embodiment 2 is that the expression for the workshop demand dynamic response function in S3 is:

[0048] ;

[0049] Where, k P Here, m is the pressure-flow coefficient, and k is the pressure sensitivity index. PIt is a parameter characterizing the proportional characteristics of the flow-pressure relationship of pressure-sensitive equipment in the workshop, reflecting the basic flow output capacity of the equipment under a unit pressure difference; m is the pressure sensitivity index, used to describe the sensitivity of the equipment flow to pressure changes. The larger the value of m, the more significant the change in equipment flow with pressure.

[0050] Example 4

[0051] The difference between this embodiment and embodiment 3 is that the calculation expression for the air compressor station flow supply and demand gap in S4 is as follows:

[0052] ;

[0053] in, Let t be the flow supply and demand gap. For the real-time flow demand of the j-th workshop, Let L be the actual air supply of the i-th air compressor unit at the gas production end, α be the pressure energy loss coefficient, and L be the total length of the pipeline from the gas production end to the gas consumption end. i,j Let be the length of the pipeline segment from the i-th air compressor unit at the gas production end to the j-th workshop. This characterizes the impact of pipeline pressure drop on the effective gas supply from the corresponding unit to the workshop.

[0054] In this embodiment, during gas transmission through pipelines, pressure drop leads to energy loss, resulting in an effective gas supply to the workshop that is lower than the actual gas supply output from the generator unit. Furthermore, the pipeline segment lengths from different generator units to different workshops vary, causing differences in the degree of pressure loss. This embodiment addresses this issue through an attenuation factor. Quantifying the impact of pressure loss on the effective gas supply along different unit-workshop paths makes the calculation of total effective gas supply more consistent with actual transmission scenarios; the supply-demand gap is the difference between the total demand flow and the total effective gas supply, which can accurately reflect the real-time matching status of gas production and consumption.

[0055] Example 5

[0056] The difference between this embodiment and embodiment 4 is that the expression for the objective function in S5 is:

[0057] ;

[0058] Where N is the number of air compressor units, a i b i Let Q be the energy consumption characteristic parameter of the i-th unit. iLet be the gas production of the i-th unit, and λ be the pressure energy loss penalty factor, a weighted parameter balancing unit energy consumption and pressure energy loss. It penalizes the deviation between the actual pressure and the target pressure in the workshop. The value of λ directly affects the focus of the optimization objective: the larger λ is, the more severe the penalty for pressure deviation, and the higher the priority is to ensure pressure stability during the optimization process; the smaller λ is, the higher the priority is to reduce unit energy consumption during the optimization process. M represents the number of workshops, and P... j Let P be the actual pressure in the j-th workshop. target This is to create target pressure for the workshop.

[0059] Example 6

[0060] The difference between this embodiment and embodiment 5 is that S6 specifically includes:

[0061] Define the game participants and payoff function: The game participants are all air compressor units at the gas production end, forming a participant set N={1,2,...,n} (n is the number of air compressor units); the payoff function is defined as "the total energy saving of the air compressor station after the units participate in gas supply", i.e., R. i =E0−E i Where E0 is the total energy consumption of the air compressor station without using the optimization method of this invention, E i The total energy consumption of the air compressor station after the i-th unit participates in the gas supply;

[0062] Improved Shapley value method for allocating revenue:

[0063] ① Calculate the characteristic function of all subsets: For all non-empty subsets S of the participant set N, calculate the total energy consumption of the air compressor station corresponding to each subset, and then obtain the characteristic function v(S)=E0−E of subset S. S (E) S (This represents the total energy consumption after the combined gas supply to the units in subset S).

[0064] ② Introducing an energy consumption weighting factor to improve the Shapley value: To reflect the differences in energy consumption characteristics among different units, an energy consumption characteristic parameter a for each unit is introduced. i As a weight, the improved formula for calculating the Shapley value is:

[0065] ;

[0066] Where |S| represents the number of units in subset S, n represents the total number of units, and v(S\{i}) is the characteristic function corresponding to the i-th unit after removing it from subset S. The revenue distribution result of each unit is calculated using this formula.

[0067] Q-learning strategy update mechanism:

[0068] ① Define the state space:

[0069] The real-time flow demand deviation ΔQ of each workshop j =Q demand,j -Q supply,j Discretization is performed to obtain all state levels in the state space, Q. supply,j This represents the current gas supply allocated to the j-th workshop.

[0070] ② Define the action space:

[0071] The action space is the set of gas production adjustment amounts for each unit. By setting reasonable adjustment steps (such as ±1 or ±2 units of flow), a finite set of actions is formed.

[0072] ③ Define the reward function: The reward function comprehensively considers the total energy consumption of the air compressor station and the supply-demand gap, and its expression is:

[0073] ;

[0074] Where E min Q represents the theoretical minimum energy consumption, where β and γ are weighting coefficients. total For the total demand flow of all workshops, ΔQ t Estimate the supply-demand gap at the current moment;

[0075] ④ Strategy Update: Initialize the Q-table (State-Action Value Table), and iteratively update the Q-value based on the collected real-time data. The update formula is:

[0076] ;

[0077] Where α is the learning rate, γ is the discount factor, s is the current state, a is the current action, s′ is the next state after executing action a, and a′ is the optimal action in the next state. The iteration continues until the Q-table converges, resulting in a stable policy.

[0078] Gas production allocation scheme optimization: Combining the revenue allocation results of the improved Shapley value with the optimization strategy after Q-learning convergence, the gas production of each unit is adjusted to form the final multi-unit gas production allocation scheme.

[0079] In this embodiment, the non-cooperative game model clearly defines the goal of "minimizing total energy consumption," sets constraints such as dynamic constraints on gas production and unit power limits, builds an optimization top-level framework, calculates the energy savings of each unit using the improved Shapley value method, allocates the total demand flow according to the proportion of the savings, determines the basic gas production of each unit, and then uses the Q-learning strategy to learn the real-time workshop demand deviation pattern, outputs the gas production adjustment of each unit, superimposes the basic gas production and the dynamic adjustment, and after correction, ensures that the constraints of the game model are met, thus obtaining the final gas production allocation scheme.

[0080] Example 7

[0081] The difference between this embodiment and Embodiment 1 is that S1 further includes preprocessing the collected multi-source real-time data, the preprocessing including data cleaning, outlier removal and data standardization.

[0082] Example 8

[0083] The difference between this embodiment and embodiment 1 is that in S2, after predicting the friction coefficient λ using an LSTM network and dynamically correcting the local resistance coefficient ξ based on the pipe topology and roughness attenuation model, a Kalman filter algorithm is used to perform a second correction on the friction coefficient λ and the local resistance coefficient ξ. Specifically:

[0084] Initial drag coefficient acquisition:

[0085] Initial value of friction coefficient: First, collect relevant data such as pipe roughness, flow rate, and pressure. After standardizing and preprocessing the data, input it into the LSTM network. By setting an appropriate number of network layers, hidden layer units, and training iterations, the network learns the mapping relationship between the data during training, and finally outputs the initial value of friction coefficient.

[0086] Initial value of local resistance coefficient: The initial value of local resistance coefficient is obtained by dynamically correcting the roughness through a roughness attenuation model, taking into account the pipeline topology (including the type, quantity and arrangement of accessories such as elbows and valves) and the influence of factors such as pipeline service life and medium corrosion on roughness.

[0087] Kalman filter initialization: Define a state vector containing the friction coefficient and local friction coefficient, set the variance parameters of the state transition matrix, process noise and observation noise, and determine the initial covariance matrix.

[0088] Iterative correction: Based on the actual pressure drop measurement data of the pipeline, the observed value of the resistance coefficient is back-calculated, and the resistance coefficient is continuously corrected through the prediction, gain calculation and state update steps of Kalman filtering.

[0089] Convergence judgment: Repeat the iteration until the change in drag coefficient between two consecutive iterations meets the preset threshold, and obtain the friction drag coefficient and local drag coefficient after secondary correction.

[0090] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of this application, and should be understood as not limiting the scope of protection of this application to such specific statements and embodiments. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for calculating the supply and demand gap of air compressor station flow based on multi-source data and dynamic game theory, characterized in that, Includes the following steps: S1: Collect multi-source real-time data from both the gas production and consumption ends. The multi-source real-time data from the gas production end includes the air supply pressure and volume of the air compressor units at the air compressor station. The multi-source real-time data from the gas consumption end includes the workshop's demand pressure P. req Environmental pressure P amb The basic data of the pipeline used for flow transportation between the gas production end and the gas consumption end includes the resistance coefficient, topology and roughness. S2: Based on multi-source real-time data, a dynamic model of pipeline pressure drop is constructed. The dynamic model of pipeline pressure drop adopts the Colebrook formula, predicts the friction coefficient λ through LSTM network, dynamically corrects the local resistance coefficient ξ based on pipeline topology and roughness attenuation model, and calculates the dynamic constraint of gas production by combining fluid characteristic parameters and pipeline structural parameters. S3: Based on workshop demand pressure P req and environmental pressure P amb A dynamic demand response function for the workshop is established, which considers the flow-pressure relationship of pressure-sensitive equipment among the flow-demand devices in the workshop. Based on the dynamic demand response function, the real-time flow demand Q of the workshop is quantified. demand ; S4: Based on the real-time flow demand Q of the workshop demand With dynamic constraints on gas production, a pressure energy loss coefficient is introduced to calculate the supply and demand gap of air compressor station flow. S5: Construct a non-cooperative game model of multiple air compressor units and multiple workshops, with the objective function being to minimize the total energy consumption of the air compressor station. The objective function includes the energy consumption term of the air compressor unit and the pressure energy loss penalty term. S6: Based on the supply and demand gap of air compressor station flow, the improved Shapley value method is used to allocate game payoffs, and the Q-learning strategy update mechanism is combined to realize the dynamic adaptive adjustment of game parameters and optimize the gas production allocation scheme of multiple units.

2. The method for calculating the supply and demand gap of air compressor station flow based on multi-source data and dynamic game theory according to claim 1, characterized in that, The dynamic model of pipeline pressure drop in S2 is expressed as follows: ； Where ΔP is the pressure drop in the pipeline, Q is the volumetric flow rate in the pipeline, ρ is the fluid density, L is the pipeline length, D is the pipeline inner diameter, and A is the pipeline cross-sectional area.

3. The method for calculating the supply and demand gap of air compressor station flow based on multi-source data and dynamic game theory according to claim 2, characterized in that, The expression for the dynamic response function of workshop demand in S3 is: ； Where, k P is the pressure-flow coefficient, and m is the pressure sensitivity index.

4. The method for calculating the supply and demand gap of air compressor station flow based on multi-source data and dynamic game theory according to claim 3, characterized in that, The expression for calculating the supply and demand gap of air compressor station flow rate mentioned in S4 is as follows: ； in, Let t be the flow supply and demand gap. For the real-time flow demand of the j-th workshop, Let L be the actual air supply of the i-th air compressor unit at the gas production end, α be the pressure energy loss coefficient, and L be the total length of the pipeline from the gas production end to the gas consumption end. i,j Let be the length of the pipeline segment from the i-th air compressor unit at the gas production end to the j-th workshop. This characterizes the impact of pipeline pressure drop on the effective gas supply from the corresponding unit to the workshop.

5. The method for calculating the supply and demand gap of air compressor station flow based on multi-source data and dynamic game theory according to claim 4, characterized in that, The expression for the objective function in S5 is: ； Where N is the number of air compressor units, a i b i Let Q be the energy consumption characteristic parameter of the i-th unit. i Let be the gas production of the i-th unit, λ be the pressure energy loss penalty factor, M be the number of workshops, and P be the gas production of the i-th unit. j Let P be the actual pressure in the j-th workshop. target This is to create target pressure for the workshop.

6. The method for calculating the supply and demand gap of air compressor station flow based on multi-source data and dynamic game theory according to claim 1, characterized in that, S1 also includes preprocessing the collected multi-source real-time data, the preprocessing including data cleaning, outlier removal and data standardization.

7. The method for calculating the supply and demand gap of air compressor station flow based on multi-source data and dynamic game theory according to claim 1, characterized in that, In S2, the friction coefficient λ is predicted by an LSTM network, and the local resistance coefficient ξ is dynamically corrected based on the pipeline topology and roughness attenuation model. Then, the friction coefficient λ and the local resistance coefficient ξ are corrected a second time by a Kalman filter algorithm.

Images