A multi-test device energy efficiency collaborative scheduling method and system and a storage medium
By employing a multi-test equipment energy efficiency collaborative scheduling method, and utilizing data acquisition, stage analysis, and distributed dual coordination optimization, the problem of high-power load overlap of environmental test equipment in the same test chamber was solved, achieving optimized scheduling of load and cost, and reducing peak load and electricity costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 成都天奥技术发展有限公司
- Filing Date
- 2026-05-14
- Publication Date
- 2026-06-12
AI Technical Summary
When existing environmental testing equipment operates in parallel within the same laboratory, multiple high-power phases overlap, leading to increased peak loads on the power distribution side and higher contract demand costs. Furthermore, manual scheduling relies on experience or simple peak-shifting delay rules, which cannot effectively reduce the peak loads of multiple devices and the overall electricity cost.
The method of coordinated energy efficiency scheduling of multiple experimental devices, including data acquisition, stage analysis, power prediction, distributed dual coordination optimization and rolling time-domain scheduling, combined with the start-up impact power term model, optimizes the electricity cost and load distribution of multiple devices.
While meeting the process constraints and deadlines of the test procedures, the peak load, maximum demand, and overall electricity cost of multiple devices are reduced, and the workload of manual scheduling is reduced.
Smart Images
Figure CN122198567A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of environmental testing technology, and in particular to a method, system and storage medium for coordinated energy efficiency scheduling of multiple testing devices. Background Technology
[0002] Environmental testing (life testing, aging testing, reliability growth testing, etc.) is typically long-duration, energy-intensive, and has distinct phases: high power consumption is required during phases such as temperature / humidity changes, rapid temperature adjustments, defrosting, and compressor startup, while maintaining relatively stable power consumption during these phases. In the same test chamber, parallel operation of multiple test chambers often leads to overlapping of several high-power phases, causing increased peak loads on the power distribution side, increased contract demand costs, and even triggering power distribution protection and affecting temperature control stability. Existing scheduling methods largely rely on manual experience or simple peak-shifting delay rules. Therefore, a method and system for energy-efficient collaborative scheduling of multiple test devices is needed to reduce peak loads and overall electricity costs, and decrease the workload of manual scheduling. Summary of the Invention
[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, system and storage medium for coordinated energy efficiency scheduling of multiple test equipment. Under the premise of meeting the process constraints of the test procedure and the deadline of the task, it reduces the peak load, maximum demand and comprehensive electricity cost of multiple equipment, and reduces the workload of manual scheduling.
[0004] The objective of this invention is achieved through the following technical solution:
[0005] A method for coordinated energy efficiency scheduling of multiple experimental devices includes the following steps:
[0006] S1: Data Acquisition: Acquire the data set of the test task, which includes the equipment identifier, test program segment data, task deadline, and allowed start time window of the task;
[0007] S2: Stage Analysis: Based on energy consumption, the experimental task is decomposed into multiple stages arranged in chronological order. These stages are classified into high-energy-consumption stages and other stages. Each stage is assigned a schedulable attribute, which includes fixed continuous stages, stages with delayed start-up, and stages with adjustable slope.
[0008] S3: Based on the power prediction model of each environmental test equipment, predict each stage obtained from the analysis in step S2 to obtain the predicted power data of each stage. The power prediction model takes into account the starting impact power of the corresponding environmental test equipment.
[0009] S4: Based on the predicted power data, under the time-of-use pricing and maximum demand billing rules, with the goal of minimizing the overall electricity cost, the cooperative scheduling optimization model is solved using distributed dual coordination under time sequence constraints, time window constraints, deadline constraints, equipment occupancy constraints, schedulable attribute constraints, and total power limit constraints, to obtain the scheduling result;
[0010] S5: Issue execution instructions to each environmental testing equipment based on the scheduling results;
[0011] Steps S3 to S5 are executed repeatedly in a rolling time domain manner.
[0012] Furthermore, the power prediction model is as follows:
[0013]
[0014] In the formula: For device indexing, , The total number of devices. For time slice index, For the first The equipment in the first Predicted power for each time slice, For the first The regression function of the equipment. For the first The first piece of equipment Each time slice feature vector For the first The parameter vector of the power prediction model for the equipment. For the first The starting impact power of the device;
[0015] The starting impact power term is:
[0016]
[0017] In the formula: For the first The starting impact amplitude of the equipment. For time slice index, For the first The time slice index corresponding to the startup time of the device. The time slice length, For the first The startup impact decay time constant of the equipment. For unit step function, It is a natural exponential function.
[0018] Furthermore, the power prediction model performs online correction based on the measured power of each environmental test device.
[0019] The online correction operation uses a recursive update:
[0020]
[0021] In the formula: For the first The first piece of equipment The feature vectors described in each time slice For the first The parameter vector described in the power prediction model for a single device. To correct the learning rate online, For the first The equipment in the first Measured power for each time slice, This represents the assignment / update symbol.
[0022] Furthermore, the collaborative scheduling optimization model uses the comprehensive electricity cost as the objective function;
[0023] The objective function of the collaborative scheduling optimization model includes at least time-of-use electricity cost and demand over-limit penalty cost;
[0024] The objective function of the cooperative scheduling optimization model is:
[0025]
[0026] In the formula: To account for the overall cost of electricity, For the first Electricity price per time slot, For the first Total power per time slice , For the first The equipment in the first Predicted power for each time slice, The time slice length, The unit price is penalized based on demand. For contract demand, It is a non-negative truncation function. Maximum demand;
[0027] The maximum demand is obtained through the maximum demand model;
[0028] The maximum demand model is as follows:
[0029]
[0030]
[0031] in: The maximum demand within the scrolling window. For the demand statistics window start time slice index, For Average power of the demand statistics window starting from [starting point] The length of the demand statistics window. For the first Total power of each time slice.
[0032] Furthermore, the schedulable attribute constraints are completed based on the schedulable attributes of each stage;
[0033] The distributed dual coordination introduces dual variables to the total power upper limit constraint and updates them iteratively. Based on the dual variables, local solution operations are performed for each environmental test equipment.
[0034] The update of the dual variables in the distributed dual coordination is as follows:
[0035]
[0036] In the formula: For the first In the nth iteration The dual variables of each time slice, For the first In the nth iteration The dual variables of each time slice, For iteration counting index, The iteration step size, For the first In the nth iteration The predicted total power for each time slice. To preset the power limit, For operators projecting onto a nonnegative field;
[0037] The local solution operation minimizes the local weighted energy consumption under time constraints, device occupancy constraints, and schedulable attribute constraints.
[0038] in, For the first Electricity price per time slot, For the first The equipment in the first Predicted power for each time slice, This represents the time slice length.
[0039] Furthermore, the local solution operation is implemented using a dynamic programming method;
[0040] The dynamic programming method uses a state cost function. Completed, the state cost function Indicates completion up to the [number]th The first stage and the Each stage in time slice Minimum local cost at the end;
[0041] The local solution operation selects the conditional value corresponding to the minimum value of the state cost function as the local solution under the time constraints, equipment occupancy constraints and schedulable attribute constraints.
[0042] The state cost function The recursive formula is:
[0043]
[0044] In the formula: For the first Phase The value of the state cost function under time slice, For the first Phase The value of the state cost function under time slice, This is the index for the end of the previous stage's time slice. For the first Phase start time slice index, , For the first Phase duration number of pieces For the first Phase at the start time slice The energy-weighted cost under the following conditions for Time slice to The waiting cost of a time slice;
[0045] The formula for calculating the energy consumption weighted cost is as follows:
[0046]
[0047] In the formula: For electricity price data, For the dual variable sequence, For the first The relative power sequence of the stages, The time slice length, For the first Phase start time slice index, This is a relative time slice index within a stage;
[0048] The waiting cost term is calculated based on the schedulable attribute.
[0049] Furthermore, the rolling time-domain method involves: locking the scheduling decisions of the executed time slices, collecting the measured power and operating status of each environmental test equipment, and re-predicting the subsequent power data for the unfinished stages and recalculating the subsequent scheduling plan.
[0050] A multi-experimental equipment energy efficiency collaborative scheduling system, applied to a multi-experimental equipment energy efficiency collaborative scheduling method, includes:
[0051] The task management module is used to acquire and store the data sets of experimental tasks;
[0052] The phase analysis module is used to decompose the test task into multiple phases arranged in time sequence according to energy consumption, classify the multiple phases into high-energy-consumption phases and other phases, and assign each phase a schedulable attribute such as a fixed continuous phase, a phase that can be delayed in start, or a phase with an adjustable slope.
[0053] The power module is used to predict each stage obtained by the stage analysis module based on the power prediction model of each environmental test equipment, and to obtain the predicted power data of each stage. The power prediction model at least takes into account the start-up impact power of the corresponding environmental test equipment.
[0054] The optimization solution module is used to solve the collaborative scheduling optimization model based on the predicted power data, under the time-of-use pricing and maximum demand billing rules, with the goal of minimizing the comprehensive electricity cost, and under the constraints of time sequence constraints, time window constraints, deadline constraints, equipment occupancy constraints, schedulable attribute constraints and total power limit constraints, to obtain the scheduling result, and periodically recalculate the subsequent scheduling plan in a rolling time domain manner;
[0055] The instruction issuing module is used to issue execution instructions to each environmental test equipment based on the scheduling results.
[0056] A storage medium for coordinated energy efficiency scheduling of multiple experimental devices is provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of a coordinated energy efficiency scheduling method for multiple experimental devices.
[0057] A computing device includes a processor and a memory, wherein the memory stores a computer program that can be executed by the processor, and the processor executes the computer program to implement the steps of a multi-experimental equipment energy efficiency collaborative scheduling method.
[0058] The beneficial effects of this invention are:
[0059] By performing phased analysis of the test tasks, and combining a power prediction model that takes into account the starting impact power term with rolling optimization of distributed dual coordination, energy efficiency collaborative scheduling of multiple test equipment was achieved. Under the premise of meeting the process constraints of the test procedures and the task deadline, the peak load, maximum demand and comprehensive electricity cost of multiple equipment were reduced, and the workload of manual scheduling was reduced. Attached Figure Description
[0060] Figure 1 A flowchart of a method for coordinated energy efficiency scheduling of multiple experimental devices;
[0061] Figure 2 This is a schematic diagram of a multi-experimental equipment energy efficiency collaborative scheduling system. Detailed Implementation
[0062] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.
[0063] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.
[0064] Example 1:
[0065] like Figures 1 to 2 As shown, a method for coordinated energy efficiency scheduling of multiple experimental devices includes the following steps:
[0066] S1: Data Acquisition: Acquire the dataset for the experimental task;
[0067] The dataset for the test task includes at least: task identifier, device identifier, test program segment data, task deadline, and allowed start time window for the task.
[0068] The experimental procedure segment data is stored in a structured data format, which includes at least: step number, step type, target temperature, target humidity, upper limit of temperature change rate, upper limit of humidity change rate, minimum holding time, maximum allowable holding time, schedulable attribute identifier, and communication address; wherein the structured data format is one or more of JSON, XML, or CSV.
[0069] S2: Stage Analysis: The experimental task is decomposed into multiple stages arranged in chronological order based on energy consumption; the multiple stages are classified into high-energy-consumption stages and other stages, and each stage is assigned a schedulable attribute, which includes fixed continuous stages, delayed start stages, and adjustable slope stages.
[0070] Step S2 performs phased analysis on the dataset for each experimental task.
[0071] Step S2 decomposes the test program segment data in the dataset of the test task into multiple stages arranged in chronological order.
[0072] Other stages can be categorized into medium-energy-consumption stages and low-energy-consumption stages, or they can be left uncategorized.
[0073] When performing phase analysis, at least one of the following high-energy-consumption phases must be identified: heating phase, cooling phase, humidification phase, dehumidification phase, defrosting phase, and compressor / heater / humidifier start-up shock phase.
[0074] Fixed continuous phase: No waiting or pausing is allowed within or between phases, and it must be seamlessly connected to the previous phase;
[0075] Delayed start phase: Pausing is not allowed during the phase, but waiting is allowed before the start of the phase to delay the start time of the phase; the waiting is achieved by extending the duration of the previous holding phase, and is not less than the minimum holding time of the holding phase and does not exceed the maximum allowed holding time of the holding phase;
[0076] Adjustable slope stage: Within the allowable rate of change of the set values according to the test standards, the temperature rise / fall slope and humidity rise / fall slope of this stage can be adjusted. This allows for changes to the stage duration and power trajectory.
[0077] Among them, the valve switching action section and the impact cycle section of the thermal shock test chamber are set as fixed continuous stages.
[0078] S3: Based on the power prediction model of each environmental test equipment, predict each stage obtained from the analysis in step S2 to obtain the predicted power data of each stage. The power prediction model takes into account the start-up impact power of the corresponding environmental test equipment.
[0079] The power prediction model is built for each environmental test equipment based on historical operating data.
[0080] Step S3 generates the predicted power sequence (i.e., predicted power data) for each stage on discrete time slices based on the stage analysis results of step S2.
[0081] The power prediction model is as follows:
[0082]
[0083] In the formula: For device indexing, , The total number of devices. For time slice index, For the first The equipment in the first Predicted power for each time slice, For the first The regression function of the equipment. For the first The first piece of equipment Each time slice feature vector For the first The parameter vector of the power prediction model for the equipment. For the first The starting impact power of the equipment;
[0084] , This represents the total number of time slices within the scrolling window.
[0085] The starting impact power term is:
[0086]
[0087] In the formula: For the first The starting impact amplitude of the equipment. For time slice index, For the first The time slice index corresponding to the startup time of the device. The time slice length, For the first The startup impact decay time constant of the equipment. For unit step function, It is a natural exponential function.
[0088] The unit is hours. The unit is hours.
[0089] In a unit step function, the value is 1 if the independent variable is not less than 0, and 0 otherwise.
[0090] The power prediction model is a discrete time slice model.
[0091] The power prediction model performs online correction based on the measured power of each environmental test device.
[0092] The online correction operation uses a recursive update:
[0093]
[0094] In the formula: For the first The first piece of equipment The feature vectors described in each time slice For the first The parameter vector described in the power prediction model for a single device. To correct the learning rate online, For the first The equipment in the first Measured power for each time slice, This represents the assignment / update symbol.
[0095] S4: Based on the predicted power data, under the time-of-use pricing and maximum demand billing rules, with the goal of minimizing the overall electricity cost, the cooperative scheduling optimization model is solved using distributed dual coordination under time sequence constraints, time window constraints, deadline constraints, equipment occupancy constraints, schedulable attribute constraints, and total power limit constraints, to obtain the scheduling result;
[0096] The collaborative scheduling optimization model is established based on time-of-use pricing and demand billing rules. Time-of-use pricing is reflected in time-of-use electricity costs; the demand billing rules adopt the maximum demand model.
[0097] The collaborative scheduling optimization model takes the comprehensive electricity cost as the objective function.
[0098] The objective function of the collaborative scheduling optimization model includes at least time-of-use electricity cost and demand over-limit penalty cost;
[0099] The objective function of the cooperative scheduling optimization model is:
[0100]
[0101] In the formula: To account for the overall cost of electricity, For the first Electricity price per time slot, For the first Total power per time slice , For the first The equipment in the first Predicted power for each time slice, The unit price is penalized based on demand. For contract demand, It is a non-negative truncation function. Maximum demand;
[0102] This indicates the time-of-use electricity cost. This indicates the penalty cost for exceeding the agreed demand.
[0103] The maximum demand is obtained through the maximum demand model;
[0104] The maximum demand model is as follows:
[0105]
[0106]
[0107] in: The maximum demand within the scrolling window. For the demand statistics window start time slice index, For Average power of the demand statistics window starting from [starting point] The length of the demand statistics window. For the first Total power of each time slice.
[0108] The maximum demand model is displayed within a scrolling window with a length of [missing information]. The average power is calculated using the demand statistics window for each time slice. and with the maximum value This indicates the maximum demand.
[0109] Scheduling attribute constraints include uninterruptible phase continuity constraints, which means that the pre-wait period for fixed continuous phases is 0 and there is seamless connection between phases.
[0110] The schedulable attribute constraints are completed based on the schedulable attributes of each stage.
[0111] The total power limit constraint means that the total power of multiple devices does not exceed the preset power limit.
[0112] The deadline constraint means that the final stage must end no later than the deadline for the slice index.
[0113] Device occupancy constraint means that each device can only execute one phase at a time.
[0114] The distributed dual coordination introduces dual variables to the total power upper limit constraint and updates them iteratively. Based on the dual variables, local solution operations are performed for each environmental test equipment.
[0115] Each environmental test equipment solves the start time of the high-energy-consumption stage locally based on the dual variables, thereby achieving peak shifting while satisfying the requirements of the cutoff time and the continuity of the uninterrupted stage.
[0116] The update of the dual variables in the distributed dual coordination is as follows:
[0117]
[0118] In the formula: For the first In the nth iteration The dual variables of each time slice, For the first In the nth iteration The dual variables of each time slice, For iteration counting index, The iteration step size, For the first In the nth iteration The predicted total power for each time slice. To preset the power limit, For operators that project onto a nonnegative field.
[0119] Operators projected onto nonnegative fields .
[0120] The local solution operation minimizes the local weighted energy consumption under timing constraints, device occupancy constraints, and schedulable attribute constraints.
[0121] in, For the first Electricity price per time slot, For the first The equipment in the first Predicted power for each time slice, This represents the time slice length.
[0122] The local solution operation is implemented using dynamic programming.
[0123] The dynamic programming method uses a state cost function. Completed, the state cost function Indicates completion up to the [number]th The first stage and the Each stage in time slice Minimum local cost at the end.
[0124] For the first Dynamic programming state cost function for each device.
[0125] The local solution operation selects the condition value corresponding to the minimum value of the state cost function as the local solution under the time constraints, equipment occupancy constraints, and schedulable attribute constraints.
[0126] The conditions include the start time, the temperature rise / fall slope adjustment value, and the humidity rise / fall slope adjustment value.
[0127] For the Stage chain of the device by stage index ( , Establish a state cost function for the total number of stages within the rolling window of the device. (Indicates completion up to the 1st) The first stage and the Each stage in time slice Minimum local cost at the end).
[0128] The state cost function The recursive formula is:
[0129]
[0130] In the formula: For the first Phase The value of the state cost function under time slice, For the first Phase The value of the state cost function under time slice, This is the index for the end of the previous stage's time slice. For the first Phase start time slice index, , For the first Phase duration number of pieces For the first Phase at the start time slice The energy-weighted cost under the following conditions for Time slice to The waiting cost of a time slice;
[0131] The formula for calculating the energy consumption weighted cost is as follows:
[0132]
[0133] In the formula: For electricity price data, For the dual variable sequence, For the first The relative power sequence of the stages, The time slice length, For the first Phase start time slice index, This is a relative time slice index within a stage.
[0134] The waiting cost term is calculated based on the schedulable attribute.
[0135] In this embodiment, the minimum value is selected while satisfying the constraints of stage time window and schedulable attribute. The corresponding starting time series is used as the local solution.
[0136] Each device solves the problem locally using dynamic programming (in this embodiment) or the shortest path algorithm.
[0137] S5: Issue execution instructions to each environmental testing device based on the scheduling results.
[0138] Based on the start time obtained from the solution, execution instructions are issued to each environmental test equipment.
[0139] Steps S3 to S5 are executed repeatedly in a rolling time domain manner.
[0140] In the rolling time-domain mode, the scheduling decision for the executed time slice is locked and will not be changed.
[0141] The rolling time-domain method includes: locking the scheduling decision of the executed time slice, collecting the measured power and operating status of each environmental test equipment, and re-predicting the subsequent power data for the unfinished stage and recalculating the subsequent scheduling plan.
[0142] A multi-experimental equipment energy efficiency collaborative scheduling system includes:
[0143] The task management module is used to acquire and store the data sets of experimental tasks;
[0144] The phase analysis module is used to decompose the test task into multiple phases arranged in time sequence according to energy consumption, classify the multiple phases into high-energy-consumption phases and other phases, and assign each phase a schedulable attribute such as a fixed continuous phase, a phase that can be delayed in start, or a phase with an adjustable slope.
[0145] The phase analysis module decomposes the test program segment into multiple phases and marks the high-energy-consuming phases and their schedulable attributes;
[0146] The power module is used to predict each stage obtained by the stage analysis module based on the power prediction model of each environmental test equipment, and to obtain the predicted power data of each stage. The power prediction model at least takes into account the start-up impact power of the corresponding environmental test equipment.
[0147] The power module establishes a power prediction model and generates a phased power sequence.
[0148] The power module also performs online calibration based on the measured power.
[0149] The billing and constraint configuration module is used to configure time-of-use pricing, contract demand, and preset power limits;
[0150] The optimization modeling module is used to build a collaborative scheduling optimization model;
[0151] The optimization solution module is used to solve the collaborative scheduling optimization model based on the predicted power data, under the time-of-use pricing and maximum demand billing rules, with the goal of minimizing the comprehensive electricity cost, and under the constraints of time sequence constraints, time window constraints, deadline constraints, equipment occupancy constraints, schedulable attribute constraints and total power limit constraints, to obtain the scheduling result, and periodically recalculate the subsequent scheduling plan in a rolling time domain manner;
[0152] The optimization solution module is a distributed dual coordination solution module, used to iteratively update dual variables and coordinate the local solution results of each device;
[0153] The instruction issuing module is used to issue execution instructions to each environmental test equipment based on the scheduling results.
[0154] The data acquisition module is used to collect measured power and operating status.
[0155] The optimization solution module includes a rolling scheduling module, which is used to lock the executed time slices and periodically recalculate the subsequent scheduling plan;
[0156] Each module is executed by the processor and connected via a communication network.
[0157] A storage medium for coordinated energy efficiency scheduling of multiple experimental devices is provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of a coordinated energy efficiency scheduling method for multiple experimental devices.
[0158] A computing device includes a processor and a memory, wherein the memory stores a computer program that can be executed by the processor, and the processor executes the computer program to implement the steps of a multi-experimental equipment energy efficiency collaborative scheduling method.
[0159] A method, system, and storage medium for coordinated energy efficiency scheduling of multiple test equipment are applicable to at least two environmental test equipment, including one or more of temperature and humidity test chambers, constant temperature and humidity chambers, and thermal shock test chambers.
[0160] Under the premise of meeting the process constraints of the test procedure, the continuity of uninterrupted segments and the task deadline, an integrated modeling and online rolling algorithm framework for "stage-level power curves (including start-up impact)" and "time-of-use pricing + maximum demand billing" has been realized.
[0161] By performing phased analysis of the test tasks, and combining a power prediction model that takes into account the starting impact power term with rolling optimization of distributed dual coordination, energy efficiency collaborative scheduling of multiple test equipment was achieved. Under the premise of meeting the process constraints of the test procedures and the task deadline, the peak load, maximum demand and comprehensive electricity cost of multiple equipment were reduced, and the workload of manual scheduling was reduced.
[0162] Example 2:
[0163] like Figures 1 to 2 As shown, Embodiment 2 has all the features of Embodiment 1, except that:
[0164] (I) Site Layout and System Deployment
[0165] This embodiment pertains to an environmental testing laboratory, which includes... Taiwan environmental testing equipment (of which) (This refers to the total number of devices), for example: 10 temperature and humidity test chambers and 2 thermal shock test chambers ( Each device is equipped with a branch smart meter to collect active power, cumulative electricity consumption, and operating status. The system adopts a deployment of "central coordinator + device edge agent":
[0166] 1) Central coordinator: Industrial server or industrial control computer, responsible for task management, billing configuration, pair coordination and rolling recalculation;
[0167] 2) Device edge agent: This can be an embedded gateway or an industrial control acquisition module, responsible for communicating with the test chamber controller, reporting status, executing local solutions, and issuing commands;
[0168] 3) Communication interface: Modbus / TCP or OPC UA is preferred; when the test box only supports RS485 / Modbus RTU, it is converted to TCP through the gateway; power acquisition preferably uses standard meter protocols (such as DL / T645 or Modbus register).
[0169] 4) Billing method: Time-of-use pricing is adopted. (in For the first Billing is based on time-segment pricing and maximum demand; maximum demand is calculated in 15-minute statistical windows (corresponding to...). A time slice, (For the length of the demand statistics window) Take the maximum average power within the scrolling window. (Maximum demand). Contracted demand is... (Contract demand, unit kW), penalty price for exceeding the contract is... (Unit: Yuan / kW).
[0170] (II) Time Discretization and Basic Symbol Conventions
[0171] To achieve online scheduling, continuous time is discretized into time slices of equal length. Let the length of each time slice be... (Time slice length, in hours, for example) (hours, i.e., 1 minute); let the length of the scrolling window be... (Scroll window length, in hours, for example) (hours); then the total number of time slices in the scrolling window is (Total number of time slices, satisfying) ); Let the time slice index be ( ).
[0172] Let the device index be ( ,in (Total number of devices). For the first... The number of stages obtained by parsing the device within the scrolling window is [number]. (Total number of stages), stage index is ( ).
[0173] (III) Task Program Format and Data Structure
[0174] Experimental tasks are recorded using a structured format, and records can be exported from the experiment management system or entered manually. A task record must contain at least:
[0175] job_id: Task identifier;
[0176] device_id: Device identifier;
[0177] deadline: The deadline for the task;
[0178] start_window: Enables the start time window;
[0179] steps: An array of program steps, sorted by step number.
[0180] Examples of fields for each step (or stage):
[0181] step_no: Step number;
[0182] step_type: Step type (ramp / hold / shock_transfer / defrost, etc.);
[0183] T_target: Target temperature setpoint (corresponding to a discrete sequence) ,in For the first (temperature setpoint for each time slice);
[0184] H_target: Target humidity setpoint (corresponding to a discrete sequence) ,in For the first (humidity setting for each time slot);
[0185] rT_max: Upper limit of temperature change rate (used as a constraint in the adjustable slope phase);
[0186] rH_max: Upper limit of humidity change rate;
[0187] hold_min: Minimum holding time (used to ensure test standard requirements);
[0188] hold_max: Maximum allowed hold time (used to allow extended hold times to achieve peak shifting);
[0189] sched_attr: A schedulable attribute identifier;
[0190] must_continuous: The uninterruptible flag (1 indicates uninterruptible, 0 indicates interruptible);
[0191] comm_addr: Controller communication address (e.g., Modbus register address / OPC node).
[0192] The device edge agent reads back from the controller: current step number, current set value, current mode, compressor / heater / humidifier start / stop status, and measured power.
[0193] (iv) Phase analysis and schedulable attributes
[0194] 1) Stage Analysis: Mapping program steps to a stage chain. Calculating the rate of change for a discrete setpoint sequence:
[0195] Rate of temperature change:
[0196]
[0197] Humidity change rate
[0198]
[0199] For the first Temperature setpoint for each time slice For the first The humidity setting value for each time slice. For the first Temperature change rate over time slices For the first Humidity change rate over time slices. The units for temperature change rate and humidity change rate are ℃ / h and %RH / h, respectively.
[0200] Set threshold (Temperature change rate threshold) and (Humidity change rate threshold). When or When this occurs, the corresponding section is marked as a high-energy-consumption stage (variable temperature / humidity stage); the holding stage and the stationary stage are usually marked as low-energy-consumption stages. The defrosting and start-up impact stages are identified by the controller status or rules (such as the compressor turning from off to on, the defrosting relay activating).
[0201] Step S2 defines the stage where the rate of change exceeds the threshold as a high-energy-consumption stage.
[0202] 2) Scheduling attributes:
[0203] Fixed continuous phase: satisfies the uninterrupted continuity constraint. Let the pre-wait time for this type of phase be... (No. The first piece of equipment The number of time slices inserted before the phase (i.e., the number of time slices of pre-phase waiting) then determines the fixed consecutive phases. Furthermore, it forces a seamless transition between the start time slice of this phase and the end time slice of the previous phase. The "valve switching action segment," "high and low temperature shock cycle segment," and "transfer segment (shock_transfer)" of the thermal shock test chamber are all classified as fixed continuous phases to meet the requirements for the continuity of shock cycles.
[0204] Delayed startup phase: Allows insertion of pre-wait. It begins with a delayed high-energy consumption phase, but cannot be paused during the phase. This wait is achieved by extending the previous holding phase, requiring that the duration of the previous holding phase be no less than its minimum holding time and no more than its maximum allowable holding time.
[0205] Adjustable slope stage: Allows the temperature / humidity slope to be adjusted to... or Within the range (of which) , These are the lower and upper limits of the rate of temperature change, respectively. , These are the lower and upper limits of the humidity change rate, respectively, in exchange for lower peak power or lower energy consumption; when this optional implementation is adopted, the stage duration varies with the slope and is constrained by the cutoff time.
[0206] 3) Phase start time and time window:
[0207] Let the first The first piece of equipment The phase start time slice index is (Start time slice index), the number of slices for the stage duration is (Number of pieces for duration).
[0208] The timing constraints are:
[0209]
[0210] in: For the first Phase pre-wait time slices; fixed consecutive phase forced , For the first The first piece of equipment Phase start time slice index, For the first The first piece of equipment Phase start time slice index, For the first The number of segments for the duration of a phase.
[0211] The time window constraint is:
[0212]
[0213] in: For the first The first piece of equipment The earliest start time slice index of the phase. For the first The first piece of equipment Latest start time slice index of the stage.
[0214] The time window constraint setting phase allows the starting time window to be... .
[0215] The deadline constraint is:
[0216]
[0217] in: For the first The deadline slice index of the task corresponding to each device within the scrolling window. This is the final stage.
[0218] The task deadline is achieved by applying an upper bound to the end time of the final phase.
[0219] (v) Power prediction model (including startup impact) and online correction
[0220] 1) Discrete Power Prediction: The first The device in the time slice The predicted power is (Predicted power, unit kW), using:
[0221] .
[0222] in, These are the eigenvectors.
[0223] Example of a feature vector:
[0224] ,
[0225] in: For the first Ambient temperature, in °C. For the first Ambient humidity, in %RH. , , The first The first piece of equipment The indicator for cooling, heating, and humidification modes can be set to 0 or 1. This is a transpose.
[0226] Regression function Linear form is acceptable: .
[0227] 2) Start-up impact power item: When a start-up event is detected (compressor or heater goes from off to on), the start-up impact power item uses:
[0228] .
[0229] 3) Stage-level power sequence generation: For each stage Generate relative power sequence ( For the first The first piece of equipment The relative power sequence of the stage item, This is a relative time slice index within a stage. , (for the duration of a phase), and by using By start time slice The device power sequence is obtained by translation and superposition. ( For the first After the equipment was dispatched, on the first Predicted power of the plate (in kW). For the first The starting impact amplitude of the equipment.
[0230] 4) Online calibration: Acquire measured power (Measured power, unit kW) is then updated recursively:
[0231] ,in This is the learning rate. This update allows the model to adapt to seasonality, load, and device aging.
[0232] (vi) Time-of-use pricing and demand-based billing model
[0233] 1) Time-of-use pricing: Defining the electricity price sequence (No. The electricity price per kWh (unit: yuan / kWh) can be mapped from the electricity price table to the time slice.
[0234] 2) Maximum Demand: Define the demand statistics window length as 15 minutes. (If the time slice length...) If the time is 1 minute, then the length of the demand statistics window is: (Number of time slices). Within the scrolling window, for each statistical window starting point... ( Calculate the average power :
[0235] ,in For the first Total power of the unit (kW).
[0236] Maximum demand is .
[0237] 3) Contract demand: The contract demand is (Unit: kW), Overdue penalty price is: (Unit: Yuan / kW), Demand penalty term is: The above modeling can be directly applied to the power company's "maximum demand billing / over-demand penalty" rules, thus binding the technical effect (reducing peak demand and demand) with the application scenario in a closed loop.
[0238] (vii) Cooperative scheduling optimization model (including total power upper limit constraint)
[0239] Define the total power sequence (Unit: kW) is: Define the preset power limit as... (Power limit, in kW), which may be determined by transformer capacity, distribution protection threshold or contractual restrictions.
[0240] The total power limit constraint is:
[0241] For all Established.
[0242] The objective function (total electricity cost) is:
[0243]
[0244] (viii) Distributed dual coordination solution (shadow electricity price driven peak shifting)
[0245] Because the total power upper limit constraint couples all devices, a distributed dual coordination approach is used to achieve a scalable solution. (Regarding the total power upper limit constraint...) Introducing dual variables by time slice (Dual variable / shadow electricity price, non-negative) decomposes the global coupling into local problems of each device.
[0246] In the The next iteration ( Under the iterative counting index, the coordinator broadcasts the sequence of dual variables. Each device's edge agent solves the local problem and completes the local solution operation:
[0247] Minimize local weighted energy consumption And satisfy the timing constraints, time window constraints, and uninterrupted phase continuity constraints of this device, thereby obtaining the local power sequence. With the start time of the phase The coordinator aggregates the predicted total power. .
[0248] Then update according to the projected subgradient:
[0249]
[0250] When the convergence criterion is satisfied Stop iteration when, where For the allowable power constraint tolerance (in kW); or when the number of iterations reaches the upper limit. Stop and enter constraint repair when (maximum number of iterations) is reached: prioritize shifting the start time of the high-energy-consuming stage that has the smallest increment to the total objective until the constraint is met. .
[0251] (ix) Solving the local subproblem of the equipment
[0252] For the For each device, the stage chain is treated as a single-machine serial execution. Each stage can select its start time within a time window (a delayed start stage is allowed, but fixed continuous stages are not). Dynamic programming is used to solve the problem.
[0253] Define the state cost function (Completed up to the 1st) Phase and this phase in time slice Minimum local cost at the end). For each candidate end time slice Its corresponding start time slice is Execution cost of the definition phase for:
[0254] ,in Let the sequence of dual variables be the current iteration. This is a phase-relative power sequence.
[0255] Waiting cost item Used to characterize the effect of waiting on the hold segment:
[0256] When a delayed start phase allows for waiting to be achieved through extended hold, it is preferable to... ,in To maintain segment power estimation; forced for fixed continuous stages Therefore, there is no waiting.
[0257] The recursion is as follows:
[0258] and constrained by time window Find the minimum under the constraints of schedulable properties. Finally, from... Backtracking yields the starting time of each stage. .
[0259] This local algorithm can run on the device edge agent, and its computational complexity is linearly related to the width of the time window, meeting the requirements for online rolling recalculation.
[0260] (x) Rolling time-domain execution and online closed loop
[0261] Let the rolling recalculation period be... (Rolling cycle, in hours, for example) (Hours, or 10 minutes). Each time a rolling cycle boundary is reached:
[0262] 1) Lock the executed time slice The phased execution within this scope will not be changed.
[0263] 2) Collect the current status and measured power of each device. Update power model parameters ;
[0264] 3) Remap the incomplete stages to a new scrolling window. (in (Given the current rolling start time in hours), reconstruct the predicted power sequence;
[0265] 4) Perform distributed dual coordination iterative solution and generate the next window schedule;
[0266] 5) Issue control commands and execute them.
[0267] Through a closed loop of "prediction-scheduling-execution-correction", the algorithm ensures long-term stable reduction of peak demand and demand.
[0268] A method, system, and storage medium for coordinated energy efficiency scheduling of multiple experimental devices have the following characteristics:
[0269] 1) Parse the program segments of the experimental task into a chain of stages, and assign schedulable attributes to the stages, clarifying "which stages are allowed to insert wait / overall translation / adjustable slope, and which stages must be fixed and continuous";
[0270] 2) Establish a power prediction model for the equipment based on historical electricity meter data, and explicitly model the initiation of the impact power term; then combine it with online measured power for recursive correction to obtain a stage-level power sequence that is closer to the actual situation.
[0271] 3) Establish a collaborative scheduling optimization model under the time-of-use pricing and maximum demand billing rules, taking into account stage sequence constraints, equipment occupancy constraints, deadline constraints, uninterrupted continuity constraints, and total power limit constraints.
[0272] 4) Distributed dual coordination is adopted: the time-slice shadow pricing guides each device to adjust the start time of high-energy-consuming segments locally, so as to achieve scalable peak-shifting scheduling;
[0273] 5) Rolling time-domain execution: Lock the executed part and periodically recalculate the subsequent plan to adapt to task insertion, equipment status changes and prediction errors.
[0274] A method, system, and storage medium for coordinated energy efficiency scheduling of multiple test chambers in an environmental laboratory, based on power prediction and distributed dual coordination, has the following advantages:
[0275] 1) Without altering the continuity of uninterrupted process sections, significantly reduce peak load and maximum demand caused by concurrent startup of multiple containers;
[0276] 2) By combining time-of-use pricing with automatic shifting of electricity to avoid peak-hour high-energy-consuming periods, the overall electricity cost can be reduced;
[0277] 3) Online correction enables power prediction to closely match changes in field equipment efficiency over the long term, improving the stability of dispatching revenue;
[0278] 4) The scale of centralized solution is reduced through distributed dual coordination, which is suitable for online rolling scheduling of multiple devices and multiple tasks.
[0279] The embodiments described above are merely illustrative of specific implementations of the present invention, and while the descriptions are detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.
Claims
1. A method for coordinated energy efficiency scheduling of multiple experimental devices, characterized in that: Includes the following steps: S1: Data Acquisition: Acquire the data set of the test task, which includes the equipment identifier, test program segment data, task deadline, and allowed start time window of the task; S2: Stage Analysis: Based on energy consumption, the experimental task is decomposed into multiple stages arranged in chronological order. These stages are classified into high-energy-consumption stages and other stages. Each stage is assigned a schedulable attribute, which includes fixed continuous stages, stages with delayed start-up, and stages with adjustable slope. S3: Based on the power prediction model of each environmental test equipment, predict each stage obtained from the analysis in step S2 to obtain the predicted power data of each stage. The power prediction model takes into account the starting impact power of the corresponding environmental test equipment. S4: Based on the predicted power data, under the time-of-use pricing and maximum demand billing rules, with the goal of minimizing the overall electricity cost, the cooperative scheduling optimization model is solved using distributed dual coordination under time sequence constraints, time window constraints, deadline constraints, equipment occupancy constraints, schedulable attribute constraints, and total power limit constraints, to obtain the scheduling result; S5: Issue execution instructions to each environmental testing equipment based on the scheduling results; Steps S3 to S5 are executed repeatedly in a rolling time domain manner.
2. The method for coordinated energy efficiency scheduling of multiple experimental devices according to claim 1, characterized in that: The power prediction model is as follows: ； In the formula: For device indexing, , The total number of devices. For time slice index, For the first The equipment in the first Predicted power for each time slice, For the first The regression function of the equipment. For the first The first piece of equipment Each time slice feature vector For the first The parameter vector of the power prediction model for the equipment. For the first The starting impact power of the device; The starting impact power term is: ； In the formula: For the first The starting impact amplitude of the equipment. For time slice index, For the first The time slice index corresponding to the startup time of the device. The time slice length, For the first The startup impact decay time constant of the equipment. For unit step function, It is a natural exponential function.
3. The method for coordinated energy efficiency scheduling of multiple experimental devices according to claim 2, characterized in that: The power prediction model is corrected online based on the measured power of each environmental test device. The online correction operation uses a recursive update: ； In the formula: For the first The first piece of equipment The feature vectors described in each time slice For the first The parameter vector described in the power prediction model for a single device. To correct the learning rate online, For the first The equipment in the first Measured power for each time slice, This represents the assignment / update symbol.
4. The method for coordinated energy efficiency scheduling of multiple experimental devices according to claim 1, characterized in that: The collaborative scheduling optimization model takes the comprehensive electricity cost as the objective function. The objective function of the collaborative scheduling optimization model includes at least time-of-use electricity cost and demand over-limit penalty cost; The objective function of the cooperative scheduling optimization model is: ； In the formula: To account for the overall cost of electricity, For the first Electricity price per time slot, For the first Total power per time slice , For the first The equipment in the first Predicted power for each time slice, The time slice length, The unit price is penalized based on demand. For contract demand, It is a non-negative truncation function. Maximum demand; The maximum demand is obtained through the maximum demand model; The maximum demand model is as follows: ； ； in: The maximum demand within the scrolling window. For the demand statistics window start time slice index, For Average power of the demand statistics window starting from [starting point] The length of the demand statistics window. For the first Total power of each time slice.
5. The method for coordinated energy efficiency scheduling of multiple experimental devices according to claim 1, characterized in that: The schedulable attribute constraints are completed based on the schedulable attributes of each stage; The distributed dual coordination introduces dual variables to the total power upper limit constraint and updates them iteratively. Based on the dual variables, local solution operations are performed for each environmental test equipment. The update of the dual variables in the distributed dual coordination is as follows: ； In the formula: For the first In the nth iteration The dual variables of each time slice, For the first In the nth iteration The dual variables of each time slice, For iteration counting index, The iteration step size, For the first In the nth iteration The predicted total power for each time slice. To preset the power limit, For operators projecting onto a nonnegative field; The local solution operation minimizes the local weighted energy consumption under time constraints, device occupancy constraints, and schedulable attribute constraints. in, For the first Electricity price per time slot, For the first The equipment in the first Predicted power for each time slice, This represents the time slice length.
6. The method for coordinated energy efficiency scheduling of multiple experimental devices according to claim 5, characterized in that: The local solution operation is implemented using dynamic programming. The dynamic programming method uses a state cost function. Completed, the state cost function Indicates completion up to the [number]th The first stage and the Each stage in time slice Minimum local cost at the end; The local solution operation selects the conditional value corresponding to the minimum value of the state cost function as the local solution under the time constraints, equipment occupancy constraints and schedulable attribute constraints. The state cost function The recursive formula is: ； In the formula: For the first Phase The value of the state cost function under time slice, For the first Phase The value of the state cost function under time slice, This is the index for the end of the previous stage's time slice. For the first Phase start time slice index, , For the first Phase duration number of pieces For the first Phase at the start time slice The energy-weighted cost under the following conditions for Time slice to The waiting cost of a time slice; The formula for calculating the energy consumption weighted cost is as follows: ； In the formula: For electricity price data, For the dual variable sequence, For the first The relative power sequence of the stages, The time slice length, For the first Phase start time slice index, This is a relative time slice index within a stage; The waiting cost term is calculated based on the schedulable attribute.
7. The method for coordinated energy efficiency scheduling of multiple experimental devices according to claim 1, characterized in that: The rolling time-domain method involves locking the scheduling decisions of the executed time slices, collecting the measured power and operating status of each environmental test equipment, and re-predicting subsequent power data for unfinished stages and recalculating subsequent scheduling plans.
8. A multi-experimental equipment energy efficiency collaborative scheduling system, applied to the multi-experimental equipment energy efficiency collaborative scheduling method described in claim 1, characterized in that: include: The task management module is used to acquire and store the data sets of experimental tasks; The phase analysis module is used to decompose the test task into multiple phases arranged in time sequence according to energy consumption, classify the multiple phases into high-energy-consumption phases and other phases, and assign each phase a schedulable attribute such as a fixed continuous phase, a phase that can be delayed in start, or a phase with an adjustable slope. The power module is used to predict each stage obtained by the stage analysis module based on the power prediction model of each environmental test equipment, and to obtain the predicted power data of each stage. The power prediction model at least takes into account the start-up impact power of the corresponding environmental test equipment. The optimization solution module is used to solve the collaborative scheduling optimization model based on the predicted power data, under the time-of-use pricing and maximum demand billing rules, with the goal of minimizing the comprehensive electricity cost, and under the constraints of time sequence constraints, time window constraints, deadline constraints, equipment occupancy constraints, schedulable attribute constraints and total power limit constraints, to obtain the scheduling result, and periodically recalculate the subsequent scheduling plan in a rolling time domain manner; The instruction issuing module is used to issue execution instructions to each environmental test equipment based on the scheduling results.
9. A multi-experimental equipment energy efficiency collaborative scheduling storage medium, on which a computer program is stored, characterized in that: When the computer program is executed by the processor, it implements the steps of the energy efficiency collaborative scheduling method for multiple experimental devices as described in any one of claims 1 to 7.
10. A computing device, comprising a processor and a memory, wherein the memory stores a computer program executable by the processor, characterized in that: When the processor executes the computer program, it implements the steps of the energy efficiency coordinated scheduling method for multiple experimental devices as described in any one of claims 1 to 7.