Multi-press line coupling energy consumption optimization method and system

Abstract

The invention discloses a multi-press line coupling energy consumption optimization method and system. The method comprises the following steps: collecting real-time operation state data and real-time power data of each device of the stamping production line; establishing a hybrid logic dynamic model based on the data, and describing a relation between a discrete working state of the equipment and a continuous energy consumption dynamic state and a material flow sequence constraint; performing event-triggered or periodic-triggered system-level rolling optimization by using an upper coordinator to minimize the total system electricity charge cost including time-sharing electricity-degree electricity charge and demand electricity charge as a target, and solving an optimal working state sequence of each device in a future time domain; issuing the sequence to a local controller of the equipment; and the local controller generates and executes a smooth control instruction based on the equipment dynamics model. The system comprises a sensing layer, an optimization decision-making layer and an execution layer. On the premise that the production takt is guaranteed, the energy consumption and demand peak value of the whole line are reduced, and the energy efficiency management level is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of industrial automation and intelligent manufacturing, and relates to a multi-machine stamping line coupling energy consumption optimization method and system. BACKGROUND

[0002] A multi-machine stamping production line is usually composed of a destacker, multiple presses, a feeding and discharging robot, a conveyor belt and other equipment according to a strict process sequence, and belongs to a typical discrete-continuous hybrid system. Its running characteristics are high single-machine power, frequent start-stop and a strong coupling material flow dependence relationship between devices. The current electricity fee structure of industrial users generally adopts a two-part electricity price, that is, a time-of-use electricity fee charged according to the electricity consumption and a demand electricity fee charged according to the maximum demand in the billing period. The latter often accounts for of the total electricity fee, and once the demand contract value is exceeded, a high penalty will be triggered.

[0003] However, the existing stamping line control system mainly realizes fixed-beat or event-driven local control based on a programmable logic controller (PLC), and the energy management is in a subordinate position. The energy-saving strategies of various devices are independent of each other and cannot be adjusted in coordination according to the real-time production state of the whole line, the electricity price signal or the demand usage. In the stable production stage, if the strokes of multiple presses are synchronized, the total power of the system will be temporarily superimposed, the average power peak value in the 15-minute window will be raised, and the demand electricity fee will be directly pushed up.

[0004] Although some researches attempt to introduce a scheduling algorithm to reduce energy consumption, the following deficiencies generally exist: first, there is a lack of an accurate model that can uniformly describe the relationship between the discrete working state of the device (such as running, idle standby and shutdown) and the continuous power dynamics, resulting in that the optimization result deviates from the reality. Second, the material flow sequence constraint between devices is not explicitly modeled, and the generated scheduling instructions may violate the process logic. Third, the optimization target only focuses on the total electricity consumption, and does not associate the demand electricity fee cost with the predicted system power peak value, so it cannot realize the comprehensive minimization of the electricity fee; fourth, the system architecture lacks the online correction ability of the model parameters, and it is difficult to adapt to the device aging or working condition changes.

[0005] Therefore, a multi-machine stamping line coupling energy consumption optimization method and system are needed, which can integrate hybrid system modeling, consider time-of-use electricity price and dynamic demand penalty, meet process timing constraints and support closed-loop self-learning, so as to realize the coordinated reduction of electricity fee and demand electricity fee on the premise of guaranteeing production efficiency. SUMMARY

[0006] To solve the problems in the background art, the application provides a multi-machine stamping line coupling energy consumption optimization method and system.

[0007] To achieve the above object, the technical scheme adopted by the present application is as follows: a multi-machine stamping line coupling energy consumption optimization method, comprising the following steps: Collecting real-time running state data and real-time power data of each device in the stamping production line; Based on the collected data, a hybrid logic dynamic model is established for the stamping production line, which describes the relationship between the discrete working state of each device and the continuous energy consumption dynamics, as well as the material flow sequence constraints between devices; Based on the hybrid logic dynamic model, system-level rolling optimization is performed through an upper coordinator to minimize the total electricity cost of the system including time-of-use electricity charges and demand charges, and to solve the optimal working state sequence of each device in the future prediction time domain, wherein the demand charge cost is associated with the predicted system total power peak; The optimal working state sequence is downloaded to the corresponding device local controller; The device local controller generates specific device control instructions based on the received optimal working state sequence and executes them; Wherein, the trigger conditions for system-level rolling optimization include event trigger and periodic trigger, and the event trigger is based on the received production plan change instruction or device fault signal.

[0008] Specifically, in the hybrid logic dynamic model, the discrete working state of each device includes at least running state, idle standby state and shutdown state; The hybrid logic dynamic model represents the discrete state of the device through binary variables, and represents the power consumption and state switching constraints of the device in each state through linear or quadratic relationships.

[0009] Specifically, the objective function of the total electricity cost of the system is represented as: The sum of the product of the system total power at each time in the prediction time domain and the corresponding time-of-use electricity price, plus a demand penalty term; Wherein, the calculation method of the demand penalty term is: the product of the predicted system total power peak and the dynamically adjusted demand penalty coefficient.

[0010] Specifically, the adjustment mechanism of the demand penalty coefficient includes: Obtain the actual maximum demand that has occurred in the current billing period; Calculate the proportion of the actual maximum demand and the preset demand contract value; When the proportion exceeds the first preset threshold, increase the demand penalty coefficient according to the preset rule; When the proportion is lower than the second preset threshold, decrease the demand penalty coefficient according to the preset rule.

[0011] Specifically, the device local controller generates specific device control instructions, including: Based on the kinetic model of the device itself, the received discrete working state sequence is converted into a continuous smooth motion trajectory setting value or a power system control instruction.

[0012] Specifically, before establishing the hybrid logic dynamic model, a model parameter identification step is further included: By operating each device to run in different working states, power data is collected, and system identification algorithms are used to determine power parameters and state switching parameters in the hybrid logic dynamic model.

[0013] Specifically, the system-level rolling optimization specifically includes: The hybrid logic dynamic model and the objective function are constructed as a mixed integer programming problem, and a mixed integer programming solver is called to solve it to obtain the optimal working state sequence.

[0014] Specifically, the optimal working state sequence is issued to the corresponding device local controller by implementing a publish-subscribe message protocol, and each device local controller subscribes to its exclusive command topic.

[0015] On the basis of the above-mentioned multi-machine stamping line coupling energy consumption optimization method, the present application further proposes a multi-machine stamping line coupling energy consumption optimization system, comprising: The perception layer includes state sensors and power measuring devices deployed on each device of the stamping production line, for collecting real-time running state data and real-time power data; The optimization decision layer includes an edge computing device, which is configured to: Based on the data collected by the perception layer, a hybrid logic dynamic model of the stamping production line is established and maintained; Based on event triggering or periodic triggering, system-level rolling optimization is performed to solve the optimal working state sequence; The optimal working state sequence is issued through a communication network; The execution layer includes local controllers associated with each device, which are configured to: Receive the optimal working state sequence; Generate device control instructions based on the optimal working state sequence and drive the device to execute.

[0016] Specifically, the edge computing device is further configured to: Periodically correct the parameters of the hybrid logic dynamic model according to historical running data to update the hybrid logic dynamic model.

[0017] Compared with the prior art, the present application has the following beneficial effects: by constructing a combined optimization target including time-sharing kilowatt-hour electricity charge and demand electricity charge, reducing the comprehensive electricity cost of the stamping line and reducing the average energy consumption of the whole line. A hybrid logic dynamic model is used to uniformly describe the energy consumption characteristics of each device in the running, idle standby and shutdown states, and material flow sequence constraints are embedded to realize multi-device collaborative energy saving and avoid invalid energy consumption caused by uncoordinated device states. The system supports an optimization mechanism combining event triggering and periodic triggering, which can quickly respond to production disturbances and reduce invalid idle energy consumption. The device local controller converts discrete state instructions into smooth control signals based on the dynamic model, ensuring smooth operation and prolonging the service life of the device. The edge computing device also has model self-correction capability and can update parameters regularly to adapt to changes in working conditions, ensuring reliable long-term optimization effect. BRIEF DESCRIPTION OF DRAWINGS

[0018] Fig. 1 is a flow chart of a multi-machine stamping line coupled energy consumption optimization method of the present application; Fig. 2 is a multi-machine stamping line coupled energy consumption optimization system architecture diagram of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0020] As shown in Figs. 1-2 The technical solutions adopted by the present application are as follows: a multi-machine stamping line coupled energy consumption optimization method, comprising the following steps: S1: collecting real-time running state data and real-time power data of each device in the stamping production line.

[0021] Through the state sensors and power measuring devices deployed in the sensing layer on each device in the stamping production line, the real-time running state data and real-time power data of each device at the current time are synchronously obtained.

[0022] The status sensor is used to detect the operating mode of the equipment, including running status, idle standby status, or shutdown status. The power measurement device is a three-phase smart meter used to measure the voltage, current, and active power in the equipment's power distribution circuit at a high sampling frequency (e.g., 100ms period) to obtain accurate real-time power data. This data acquisition process aggregates data to a data concentrator via an industrial communication network (e.g., Profinet RT / IRT or EtherCAT), and the optimization decision layer subscribes to and receives the data via the OPC UA protocol, serving as the basic input for subsequent establishment of a hybrid logic dynamic model and execution system-level rolling optimization.

[0023] S2: Based on the collected data, a hybrid logic dynamic model is established for the stamping production line. The hybrid logic dynamic model describes the relationship between the discrete working state of each device and the continuous energy consumption dynamics, as well as the material flow sequence constraints between devices.

[0024] Specifically, in the hybrid logic dynamic model, the discrete operating states of each device include at least the running state, the idle standby state, and the power-off state.

[0025] The hybrid logic dynamic model uses binary variables to represent the discrete states of the device and uses linear or quadratic relationships to represent the power consumption and state switching constraints of the device in each state.

[0026] The hybrid logic dynamic model is a modeling tool that uses collected real-time operating status data and real-time power data to provide a unified mathematical description of the energy consumption behavior and collaborative logic of all equipment in the stamping production line.

[0027] The hybrid logic dynamic model first provides each device with... Define three mutually exclusive discrete operating states: running state, idle standby state, and power-off state, each represented by a binary variable. , and This indicates that the constraint relationship is satisfied at any given time: ; equipment At any moment Real-time power consumption Described by the following expression: ; in, , , respectively equipment The reference power parameters for the corresponding state are obtained through offline parameter identification. (Function) Represents the variable power components related to the current process load, using linear... Or quadratic relationship modeling ; Load-power relationship parameters of the equipment obtained through system identification . Process noise term, including measurement noise or unmodeled dynamics, regarded as zero-mean disturbance in actual optimization, not involved in decision variable modeling, used to characterize measurement uncertainty.

[0028] The hybrid logic dynamic model further introduces state switching constraints, including minimum running time, minimum shutdown time and state transition energy consumption, which are transformed into linear inequality constraints through the big M method and embedded in the optimization problem to ensure physical feasibility.

[0029] In addition, the hybrid logic dynamic model depicts the material flow sequence constraints between devices, i.e. for any device pair whose state execution timing needs to satisfy: ; Wherein, indicates the process completion flag (integer variable) of device at time , is the material transmission delay step number from device to device , is an auxiliary binary variable, is a large enough constant, and the inequality set ensures that downstream device can only start after upstream device completes the corresponding process and goes through the transmission delay, thereby ensuring the process correctness of the production process.

[0030] In summary, the hybrid logic dynamic model uniformly represents discrete states through binary variables, associates continuous power dynamics through linear or quadratic functions, and integrates state switching logic and material flow sequence constraints, forming a hybrid system mathematical model that can be used for system-level rolling optimization.

[0031] Specifically, before establishing the hybrid logic dynamic model, a model parameter identification step is included: by operating each device under different working states, power data is collected, and system identification algorithm is used to determine the power parameters and state switching parameters in the hybrid logic dynamic model.

[0032] Before establishing the hybrid logic dynamic model, the system performs an offline model parameter identification step to ensure that the physical parameters in the model have actual engineering accuracy.

[0033] The model parameter identification step first operates each device in the stamping production line in sequence through manual or automatic control instructions Stable operation in all discrete operating states, including running state, idle standby state and shutdown state.

[0034] In each state, the power measurement device in the perception layer continuously collects real-time power data of the device at a high sampling frequency (e.g. 100 ms period) and records for a long enough time to eliminate transient fluctuations, thereby obtaining a steady-state power sample set in the state.

[0035] Subsequently, the optimization decision layer calls a system identification algorithm to process the collected power data; the system identification algorithm uses least squares method or subspace identification method to fit the reference power parameters of the device in each state.

[0036] Specifically, the following key parameters in the hybrid logic dynamic model are determined by identification: running state reference power , idle standby state reference power , shutdown state reference power , and additional energy consumption parameters in state switching process , representing the one-time startup energy consumption when switching from shutdown or standby to running state.

[0037] In addition, this step also identifies the time constraint parameters of state switching, including minimum running time and minimum shutdown time , which are obtained by recording the shortest duration allowed for safe operation of the device in state switching experiments, and are used to construct the logical constraints in the hybrid logic dynamic model.

[0038] All identified parameters are stored in the MLD model library and parameter library of the optimization decision layer, as fixed inputs for the mixed integer programming problem in subsequent system-level rolling optimization, ensuring that the optimization results not only conform to physical laws but also have engineering implementability.

[0039] This model parameter identification step is the prerequisite for accurate energy consumption modeling, and its output directly determines the approximation accuracy of the hybrid logic dynamic model to the actual system energy consumption behavior, thereby affecting the economy and reliability of rolling optimization.

[0040] S3: Based on the hybrid logic dynamic model, perform system-level rolling optimization through the upper coordinator to minimize the total system electricity cost including time-of-use electricity charges and demand electricity charges, and solve the optimal working state sequence of each device in the future prediction time domain, where the demand electricity charge cost is associated with the predicted system total power peak.

[0041] ​The trigger condition for the system-level rolling optimization includes event triggering and periodic triggering, and the event triggering is based on a received production plan change instruction or a device fault signal.

[0042] Specifically, the target function of the system total electricity cost is represented as: The sum of the product of the system total power at each time in the prediction time domain and the corresponding time-of-use electricity price, plus a demand penalty term.

[0043] The demand penalty term is calculated as the product of the predicted system total power peak value and a dynamically adjusted demand penalty coefficient.

[0044] Specifically, the adjustment mechanism of the demand penalty coefficient includes: Obtaining the actual maximum demand that has occurred in the current billing period.

[0045] Calculating the ratio of the actual maximum demand to the preset demand contract value.

[0046] When the ratio exceeds a first preset threshold, increasing the demand penalty coefficient according to a preset rule.

[0047] When the ratio is lower than a second preset threshold, decreasing the demand penalty coefficient according to a preset rule.

[0048] The upper coordinator is deployed on an edge computing device in the optimization decision layer, and its core function is to perform system-level rolling optimization once at each optimization period or event triggering time based on the established hybrid logic dynamic model.

[0049] The specific implementation of the system-level rolling optimization is to jointly formalize the hybrid logic dynamic model and a target function with the goal of minimizing the system total electricity cost into a mixed integer programming problem, and to call a mixed integer programming solver to solve it by mathematical optimization method to obtain the optimal working state sequence.

[0050] The system total electricity cost is composed of two parts: the first part is the time-of-use electricity cost, and the second part is the demand electricity cost; both of which constitute the target function The mathematical expression is: ; Wherein, N is the number of steps in the prediction time domain, which is a positive integer; All integers greater than or equal to zero), represents the absolute discrete time step index, such as the current is the Step; N represents the relative prediction step; the corresponding physical time (continuous) is: .

[0051] Variable denotes the prediction time The total power of all equipment in the stamping production line, defined as: ; where, denotes the total number of equipment in the stamping production line, which is a fixed positive integer; is the equipment index, taking values in the range ; is the power consumption of equipment at the prediction time .

[0052] Parameter denotes the time-of-use electricity price (unit: yuan / kWh) corresponding to the grid at time , which is obtained in real time through an external API interface with the energy management system or the grid, and is used to reflect the price difference of electricity energy in different time periods.

[0053] Parameter denotes the actual physical time length corresponding to each optimization step (unit: hour), for example, when the control period is 1 minute, , which is used to convert instantaneous power integration into electricity consumption.

[0054] Variable is a newly added continuous optimization variable, representing the predicted peak value of the total system power within the current prediction time domain , whose value is introduced through the following linear constraint: ; This modeling method converts the non-smooth maximum operation into a set of linear inequality constraints, so that the original problem can be handled by a mixed integer programming solver.

[0055] Parameter is the dynamically adjusted demand penalty coefficient (unit: yuan / kW), whose value is not fixed but is adaptively adjusted according to the actual operation in the current billing period, to simulate the billing mechanism of industrial user demand electricity, i.e., the maximum value of 15-minute average power in the monthly period determines the basic electricity fee.

[0056] Specifically, the adjustment mechanism of the demand penalty coefficient includes the following steps: first, obtain the actual maximum demand (unit: kW) that has occurred in the current billing period from the smart meter or energy management system; then, read the preset demand contract value (unit: kW), which is the upper limit of the contract capacity signed by the user and the power supply enterprise.

[0057] Next, calculate the proportion . If the proportion exceeds a first preset threshold , for example , it is determined that the current demand risk is high, and at this time , for example, the formula is used ; is the benchmark demand penalty coefficient, which is converted to the equivalent value of the current optimization period by the unit demand price, is a penalty gain coefficient, for example , used to control the growth slope of the penalty intensity.

[0058] On the contrary, if the ratio is lower than a second preset threshold , for example , and , it is determined that the demand margin is sufficient, and at this time , , for example, is reset to or further reduced to avoid excessive suppression of power and sacrifice production efficiency.

[0059] The decision variables of the mixed integer programming problem include: for each device and each time in the prediction time domain, where , three mutually exclusive binary variables , , are introduced, respectively representing the running state, the idle standby state and the shutdown state of the device at the prediction time .

[0060] At the same time, continuous variables are introduced to represent the total system power peak in the prediction time domain, and auxiliary continuous variables are introduced to represent the power consumption of the device at time .

[0061] The constraint conditions are composed of three types: the first type is the device ontology constraint in the mixed logical dynamic model, including state mutual exclusion constraints: .

[0062] The constraint ensures that any device is in only one discrete working state at any time; Also includes power calculation relationships: ; wherein , , are the device Reference power parameters in each state. Function represents the variable power component related to the current process load, modeled with linear or quadratic relationship.

[0063] Also includes state switching time constraints, minimum running time constraints: ; where, is the minimum running time of the device (unit: step), which ensures that the device maintains the running state for at least the specified time period after starting, preventing frequent start-stop. The second type is the material flow sequence constraint between devices, which is for any device pair

[0064] with process dependency, by introducing auxiliary binary variable and large constant , construct linear inequalities: ; where, is the process completion flag of the device (integer variable), which is used to record the number of completed processes of the device , and the process completion flag of the device is incremented by 1 every time it completes the processing of a workpiece; this flag is triggered by the local controller of the device to update according to the encoder position or photoelectric switch signal, and is reported to the upper coordinator through the OPC UA protocol; is the delay step number required for the material to be transmitted from device to device ; is the set of all device pairs with upstream and downstream process dependency relationships; is a large positive real number, whose value is greater than any possible process difference, to ensure logical correctness. This constraint ensures that the downstream device can only start after the upstream device has completed the corresponding process and has passed the transmission delay, thereby ensuring the timing correctness of the production process.

[0065] The third type is the system-level power constraint, including: ; and: ; where, is the allowed demand overrun margin (unit: kW), which is used to prevent hard violation of the demand contract value due to model errors or measurement noise, and to improve system robustness.

[0066] ​​After the above modeling is completed, the optimization decision layer calls a commercial mixed integer programming solver to solve the problem; the solver is Version, supporting mixed integer linear / quadratic programming.

[0067] The maximum solving time is set during the solving process seconds and the optimality gap , so as to obtain a high-quality feasible solution within a limited time.

[0068] Finally, the solver outputs the future control time domain step, wherein , and is the optimal working state sequence of each device within the actual control steps: ; The sequence is sent to the programmable logic controller of the execution layer as a control instruction, driving each device to switch the working state according to the optimization result, so as to complete a system-level rolling optimization closed loop.

[0069] The trigger conditions of the system-level rolling optimization include two types: periodic triggering and event triggering; periodic triggering means that under the condition of no external disturbance, the optimization is automatically started once every fixed time interval , for example, 5 minutes, to realize continuous rolling update.

[0070] Event triggering means that when the optimization decision layer receives one of the two types of specific signals, the current period is interrupted and emergency optimization is started: the first type is the production plan change instruction issued by the manufacturing execution system, such as inserting an emergency order or switching a mold. The second type is the device fault signal reported by the perception layer, such as an emergency stop of a press or a communication interruption of a robot.

[0071] Through the above mechanism, the upper coordinator dynamically balances the time-of-use electricity charge and the demand charge under the premise of guaranteeing the material flow sequence constraint and the production beat, generates an optimal working state sequence of the device that is both economical and feasible, so as to realize the joint optimization of the energy consumption and the demand power of the stamping production line at the system level.

[0072] S4: issuing the optimal working state sequence to the corresponding device local controller.

[0073] Specifically, the optimal working state sequence is issued to the corresponding device local controller through a publish-subscribe message protocol, and each device local controller subscribes to its exclusive command topic.

[0074] Issuing the optimal working state sequence to the corresponding device local controller means that after the upper coordinator completes the system-level optimization, it rolls and issues the obtained future a discrete state instruction set for each control cycle, reliably delivered from the optimization decision layer to the local controller of the specific physical device in the execution layer.

[0075] The process is implemented through a publish-subscribe message protocol, the core of which is to realize the direction, isolation and conflict-free delivery of instructions through the topic mechanism.

[0076] In the system deployment phase, a global unique command topic identifier is assigned to each device in the stamping production line , , , denoted as . The identifier is generated using a structured naming rule, for example: ; wherein represents the station number to which the device belongs, represents the device type identifier (such as press, robot, conveyor belt, etc.), so as to ensure that for any two different devices and , i.e. , the corresponding command topic strictly satisfies . This uniqueness is a basic design constraint of the system communication architecture, which is used to eliminate instruction routing ambiguity.

[0077] Each device local controller registers a persistent subscription to the message middleware (such as MQTT Broker) located in the workshop network during its startup initialization process, declaring that it only receives data published to its exclusive command topic ; once the subscription relationship is established, it remains effective during the controller operation and is completely orthogonal to the subscriptions of other devices.

[0078] When the upper coordinator completes a rolling optimization, for each device , the optimal work state sequence within the next steps is extracted, i.e.: ; wherein the state triple corresponding to each time step satisfies the mutual exclusion constraint .

[0079] Subsequently, the optimization decision layer organizes into a structured instruction and publishes it to the message middleware through a standard network interface, explicitly specifying the target topic as .

[0080] Because the message middleware internally maintains a subscription mapping table for all local controllers, it will only forward the instruction to those already subscribed, based on a topic matching mechanism. The only local controller for that specific device; any other device that is not subscribed to that topic will not receive this data at all. It is precisely because of this... For any Only through its establishment can the certainty and security of this one-to-one delivery be guaranteed.

[0081] This communication mechanism does not rely on IP addresses, port numbers, or connection states, and features loose coupling, high scalability, and strong fault tolerance: even if a device is temporarily offline, it can still receive subsequent commands normally after it comes back online without interfering with the control flow of other devices.

[0082] Through the above design, the optimal working state sequence is accurately transmitted to the corresponding local controller of the equipment in a highly reliable, low-latency, and strongly isolated manner, providing an accurate reference for the subsequent generation of smooth and safe equipment control commands, while avoiding mis-sending, omission, or cross-interference of commands, and ensuring the stability and safety of the entire stamping production line in coordinated operation.

[0083] S5: The local controller of the device generates specific device control instructions and executes them based on the received optimal working state sequence.

[0084] Specifically, the device local controller generates specific device control instructions including: Based on the device's own dynamic model, the received discrete working state sequence is converted into continuous and smooth motion trajectory setpoints or power system control commands.

[0085] The local controller for the equipment refers to the controller that is connected to each physical device in the stamping production line. Each programmable logic controller (PLC) is configured with a one-to-one correspondence, and its internal components have been loaded with the optimized instruction execution function blocks designed in this invention.

[0086] After receiving the optimal working state sequence from the upper-layer coordinator via the publish-subscribe messaging protocol, the device's local controller first extracts the future-oriented information from the message payload. The discrete state command for each control cycle is denoted as: ; Wherein, for any prediction step Satisfying mutual exclusion constraints: ; Subsequently, the device's local controller accesses the device's internal storage. The dynamic model is a continuous-time state-space equation, in the following form: ; ; in, For equipment In continuous time The state vector (such as motor angular velocity, hydraulic cylinder displacement, pressure, etc.). To control input vectors (such as voltage, current commands, valve opening). For measurable output, , , This is to identify the model parameter matrix through the system or provided by the manufacturer, and only related to the device. Binding.

[0087] Based on this dynamic model, the device's local controller performs trajectory planning: for each discrete state... ,in This is mapped to a set of predefined underlying behavior patterns and generates a corresponding continuous smooth setpoint sequence.

[0088] when At that time, the controller determines the process subroutine (such as the third-order stamping or robot part picking) based on the current production task (provided synchronously by MES), and combines the kinematic constraints in the dynamic model (such as maximum acceleration) Maximum speed The position setpoint curve is generated using either fifth-order polynomial interpolation or an S-shaped velocity planning algorithm. Or speed setpoint curve Defined in a continuous time interval Above, ensure the trajectory is continuously differentiable, at least Continuity (in engineering, mathematics, and control theory, it usually refers to the order of continuity of a function or trajectory, where the function value and its first derivative are continuous, i.e., the curve is smooth, without sharp corners, and the velocity is continuous) is used to avoid mechanical shock.

[0089] when At this time, the controller keeps some power system components enabled, such as keeping the servo drive enabled but setting its speed to zero, shutting down the main pump in the hydraulic system but maintaining pressure in the accumulator, and outputting low-power control commands. Its power level corresponds to the identified reference power. .

[0090] when When the controller executes the safety shutdown procedure, it first returns the actuator to a safe position (e.g., the press slide stops at the top dead center), and after a delay to confirm no movement, it disconnects the main contactor or sends an OFF command to the frequency converter, outputting a shutdown command. At this point, the device power consumption approaches [value missing]. .

[0091] To improve tracking accuracy, the device's local controller has a built-in disturbance observer that estimates the total disturbance in real time. This includes load variations, frictional nonlinearity, and workpiece quality deviations, which are then fed forward and compensated into the control law. Simultaneously, local sensors (such as encoders, torque sensors, and pressure transmitters) are used to collect the actual output. , and reference trajectory By comparison, the feedback correction is calculated using a discretized PID controller: ; in , For equipment Dedicated PID gain parameters.

[0092] The final control command is the sum of the feedforward and feedback commands: This instruction is output to the servo drive, hydraulic valve group, or contactor through the local I / O module, driving the device to perform the corresponding action.

[0093] At the same time, the device's local controller will report the actual execution status, such as whether it has entered the running state, the current speed is 98%, or the standby timeout triggers shutdown, to the upper-level coordinator in real time via the OPC UA protocol, forming a closed-loop monitoring system.

[0094] Through the above mechanism, the local controller of the equipment accurately transforms the high-level discrete optimization results into low-level continuous, smooth, and safe physical control behavior, ensuring the reliable execution of energy consumption optimization strategies while guaranteeing production cycle time and equipment lifespan.

[0095] Based on the aforementioned method for optimizing energy consumption in a multi-machine stamping line coupling, this invention further proposes a multi-machine stamping line coupling energy consumption optimization system, comprising: The architecture consists of three layers: perception layer, optimization and decision-making layer, and execution layer. Each layer has a clearly defined function and operates collaboratively.

[0096] The perception layer includes status sensors and power measurement devices deployed on various equipment in the stamping production line. These devices are used to synchronously collect real-time operating status data and real-time power data of each piece of equipment and upload the data to the optimization decision layer via an industrial communication network.

[0097] The optimization decision layer includes an edge computing device configured to: establish and maintain a hybrid logic dynamic model of the stamping production line based on data collected by the perception layer; perform system-level rolling optimization under event-triggered or period-triggered conditions to solve for the optimal working state sequence of each device in the future prediction time domain; and send the optimal working state sequence to the corresponding device local controller through a communication network.

[0098] The execution layer includes local controllers associated with each device, each configured to subscribe to its exclusive command topic to receive the optimal work state sequence, and based on the device's own dynamic model, convert the received discrete work state sequence into continuous smooth motion trajectory set values or power system control instructions, and drive the device to execute safely and smoothly.

[0099] In addition, the edge computing device is also configured to periodically correct the parameters of the hybrid logic dynamic model based on historical operation data: by comparing the deviation between the model predicted power and the actual measured power, when the average relative error exceeds the preset threshold (such as 5%), the parameter re-identification process is automatically started, and the latest operation data is used to update the baseline power parameters 、 、 and state switching parameters, so as to realize online self-learning and continuous optimization of the model.

[0100] In a specific embodiment, a multi-machine stamping production line composed of 3 presses (numbered P1, P2, P3), 2 transfer robots (numbered R1, R2), and 1 conveyor belt (numbered C1) is a specific application scenario.

[0101] Each device is equipped with a state sensor (such as a press slide position sensor, a robot joint encoder) and a power measurement device (a three-phase intelligent power meter) for real-time acquisition of the device's operating state (running, idle standby, shutdown) and power data.

[0102] The collected data is aggregated through the PROFINET network of the workshop, and is subscribed and obtained by the optimization decision layer software deployed on the edge computing server (such as Dell Edge Gateway 5000) through the OPC UA protocol.

[0103] In the initial stage, the system performs a model parameter identification step: sequentially operating each device in different operating states, collecting power data, and identifying the baseline power parameters in each state using the least squares method, such as the P1 press running power , idle power and the minimum time constraint of state switching.

[0104] Based on the identified parameters and real-time data, the optimization decision layer establishes a hybrid logic dynamic model (MLD) for the entire production line, which uses binary variables, such as to represent the discrete state of each device, describes the power consumption with linear relationships, and encodes the material flow sequence constraints between devices, such as P1 completing stamping, R1 can start picking up.

[0105] The upper coordinator in the decision layer triggers system-level rolling optimization based on events (e.g., receiving an emergency order issued by MES) or periods (e.g., every 5 minutes).

[0106] Each time the optimization is performed, the coordinator constructs an MLD model with a power cost objective function as a mixed integer programming problem for a prediction horizon of 15 minutes in the future; the objective function considers both time-of-use electricity charges (based on real-time electricity prices) and demand charges (based on predicted power peak and the product of dynamic coefficients ).

[0107] Here, the demand penalty coefficient is dynamically adjusted according to the ratio of the actual maximum demand that has occurred in the current billing period to the contract value , and if it approaches the upper limit, the penalty strength is automatically increased.

[0108] This mixed integer programming problem calls the Gurobi solver for calculation, aiming to minimize the total electricity cost, and solves the optimal working state sequence of each device in the future control period, for example: instruct P1 to keep running in the next minute, and R1 to switch to idle standby after 30 seconds.

[0109] The optimal working state sequence obtained by solving is issued through the publish-subscribe message protocol (MQTT); the edge server publishes the command for P1 to the topic pressline / station1 / device_press / command, and the local controller of P1 has subscribed to this topic, so it can accurately and exclusively receive its own command.

[0110] After the device local controller (such as the S7-1500 PLC of P1) receives the discrete state sequence, it converts the discrete command into continuous and smooth motion control commands based on its built-in device dynamics model and current production tasks. For devices with complex motion trajectories (such as robots and servo presses), a complete dynamics model is needed for trajectory planning. For conveyor belts and other simple start-stop devices, a preset speed curve or direct state switching control can be used, without the need to solve differential equations online.

[0111] For example, when the command requires P1 to enter the running state, the PLC will plan a smooth slider motion curve and combine the feedback of local sensors to drive the servo motor to perform the stamping action through the PID control algorithm. If the command requires entering idle standby, the servo driver is kept enabled but outputs a zero-speed command to reduce energy consumption.

[0112] The local controller of each device feeds back the instruction execution state to the upper coordinator in real time, thereby forming a closed-loop control. Meanwhile, the edge computing server corrects the parameters of the MLD model using historical operation data periodically, so that the model can continuously approximate the actual characteristics of the production line.

[0113] Through the above specific implementation, the present application realizes system-level energy consumption optimization on the stamping line. Under the premise of ensuring the production rhythm and process sequence, the total power curve of the system is effectively smoothed by coordinating the working states of each device, thereby reducing the costs of time-of-use electricity and demand electricity.

[0114] Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can modify the technical solutions recorded in the foregoing embodiments or make equivalent replacements to some of the technical features, and any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for optimizing energy consumption in a multi-machine stamping line coupling, characterized in that, Includes the following steps: Collect real-time operating status data and real-time power data of each piece of equipment in the stamping production line; Based on the collected data, a hybrid logic dynamic model is established for the stamping production line. The hybrid logic dynamic model describes the relationship between the discrete working state of each device and the continuous energy consumption dynamics, as well as the material flow sequence constraints between devices. Based on the hybrid logic dynamic model, system-level rolling optimization is performed through the upper-level coordinator to minimize the total system electricity cost, which includes time-of-use electricity charges and demand charges. The optimal operating state sequence of each device in the future prediction time domain is solved, where the demand charge cost is related to the predicted total system power peak. The optimal operating state sequence is sent to the corresponding device local controller; The device local controller generates and executes specific device control commands based on the received optimal operating state sequence. The triggering conditions for system-level rolling optimization include event triggering and periodic triggering. The event triggering is based on the received production plan change instruction or equipment fault signal.

2. The method for optimizing energy consumption in a multi-machine stamping line coupling according to claim 1, characterized in that, In the hybrid logic dynamic model, the discrete operating states of each device include at least the running state, the idle standby state, and the power-off state. The hybrid logic dynamic model uses binary variables to represent the discrete states of the device and uses linear or quadratic relationships to represent the power consumption and state switching constraints of the device in each state.

3. The method for optimizing energy consumption in a multi-machine stamping line coupling according to claim 1, characterized in that, The objective function for the total electricity cost of the system is expressed as: The sum of the products of the total system power at each time point in the prediction time domain and the corresponding time-of-use electricity price, plus a demand penalty term; The demand penalty term is calculated as the product of the predicted total system power peak and the dynamically adjusted demand penalty coefficient.

4. The method for optimizing energy consumption in a multi-machine stamping line coupling according to claim 3, characterized in that, The adjustment mechanism for the demand penalty coefficient includes: Get the actual maximum demand that has occurred within the current billing cycle; Calculate the ratio of the actual maximum demand to the preset demand contract value; When the ratio exceeds the first preset threshold, the demand penalty coefficient is increased according to preset rules; When the ratio is lower than the second preset threshold, the demand penalty coefficient is reduced according to the preset rules.

5. The method for optimizing energy consumption in a multi-machine stamping line coupling according to claim 1, characterized in that, The device local controller generates specific device control commands, including: Based on the device's own dynamic model, the received discrete working state sequence is converted into continuous and smooth motion trajectory setpoints or power system control commands.

6. The method for optimizing energy consumption in a multi-machine stamping line coupling according to claim 1, characterized in that, Before establishing the hybrid logic dynamic model, a model parameter identification step is also included: By operating each device in different working states, its power data is collected, and the power parameters and state switching parameters in the hybrid logic dynamic model are determined using the system identification algorithm.

7. The method for optimizing energy consumption in a multi-machine stamping line coupling according to claim 1, characterized in that, The specific aspects of performing system-level rolling optimization include: The hybrid logic dynamic model and the objective function are constructed as a hybrid integer programming problem, and a hybrid integer programming solver is called to solve it to obtain the optimal working state sequence.

8. The method for optimizing energy consumption in a multi-machine stamping line coupling according to claim 1, characterized in that, The optimal working state sequence is sent to the corresponding device local controller through a publish-subscribe messaging protocol, whereby each device local controller subscribes to its own command topic.

9. A multi-machine stamping line coupled energy consumption optimization system, characterized in that, include: The sensing layer includes status sensors and power measurement devices deployed on various equipment in the stamping production line, used to collect real-time operating status data and real-time power data; The decision-making layer is optimized, including edge computing devices configured as follows: Based on the data collected by the perception layer, a hybrid logic dynamic model of the stamping production line is established and maintained. Based on event-triggered or periodic-triggered processes, perform system-level rolling optimization to find the optimal working state sequence. The optimal working state sequence is transmitted via the communication network; The execution layer includes local controllers associated with each device, the local controllers being configured to: Receive the optimal working state sequence; Based on the optimal working state sequence, device control commands are generated and the device is driven to execute them.

10. The multi-machine stamping line coupling energy consumption optimization system according to claim 9, characterized in that, The edge computing device is also configured to: The parameters of the hybrid logic dynamic model are periodically corrected based on historical operating data in order to update the hybrid logic dynamic model.

Images