Multi-time scale demand scheduling method and system considering user response fatigue

Abstract

The invention discloses a multi-time scale demand scheduling method and system considering user response fatigue. The method comprises the following steps: firstly, constructing a user portrait based on historical data, and calculating an accumulated fatigue state covering accumulation and recovery processes in real time by using a dynamic quantitative model; secondly, constructing a fatigue-capacity physical mapping mechanism, and converting a fatigue state into an effective schedulable capacity upper limit; a multi-objective optimization model considering the system operation cost and the user fatigue loss is established; and finally, performing fatigue state updating and scheduling target compensation based on the actual response power of the user by adopting a rolling time domain control strategy. According to the method, user fatigue characteristics in a continuous calling scene can be accurately described, and through a closed-loop feedback correction mechanism, the economy of scheduling and the sustainability of resources are effectively considered while the supply and demand balance of a power grid is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of demand side management and load regulation of power systems, and in particular to a multi-time scale demand scheduling method and system considering user response fatigue. BACKGROUND

[0002] As an important means to improve the flexibility of power systems, promote new energy consumption and reduce peak capacity investment, demand response has been widely used in peak load shifting, emergency load reduction, auxiliary services and other businesses. The existing demand response scheduling method usually estimates the user response potential based on the user's signed capacity, historical average response rate, load adjustable model or price elasticity model, and optimizes the allocation and instruction issuance on this basis.

[0003] However, in actual operation, demand response often presents the typical characteristics of continuous multiple calls, such as multi-period peak shaving under extreme weather, multi-round frequency modulation within a day, or continuous multi-day power supply protection scenarios. At this time, users will show obvious response fatigue under the influence of factors such as psychology, comfort, production disturbance, and equipment start-stop wear: as the number of calls increases, the duration lengthens or the intensity increases, the user's response willingness and commitment ability gradually decrease, leading to an enlarged response deviation in the later period, unsustainable response resources, and a significantly increased scheduling risk.

[0004] The existing technology has obvious deficiencies in handling such continuous multiple call scenarios. On the one hand, the existing scheduling method generally lacks a detailed description of the dynamic fatigue mechanism of users, and usually only sets static constraints such as maximum daily call times or minimum interval times to avoid over-calling, which cannot accurately quantify the physical consumption and recovery process of users during continuous response and their intermittent periods. On the other hand, traditional scheduling models mostly use local optimization strategies in a single time section, ignoring the overdraft effect of current high-intensity scheduling on user available capacity in future periods, which leads to the phenomena of enlarged response deviation in the later period of scheduling, exhausted resources too early, or a sharp increase in the risk of scheduling failure. In addition, the existing technology generally lacks a closed-loop correction mechanism based on actual execution deviation, and fails to feed back the actual commitment of users to the fatigue state evaluation link, resulting in a gradual distortion between the theoretical prediction value of the scheduling model and the actual physical state of users, making it difficult to guarantee the accuracy and reliability of scheduling in a long time scale.

[0005] In summary, it is urgent to propose a demand response scheduling method that can depict the dynamic evolution of user fatigue in multiple demand response scenarios and incorporate the impact of fatigue in the form of probability and effective capacity upper limit into multi-objective rolling optimization scheduling, achieving closed-loop correction and long-term sustainable operation. SUMMARY

[0006] This invention aims to address the technical problem of existing demand response scheduling methods in scenarios with multiple consecutive calls. Due to the lack of dynamic characterization and closed-loop correction mechanisms for user response fatigue, scheduling strategies cannot balance short-term economic efficiency and long-term fulfillment reliability, leading to premature exhaustion of response resources or increased scheduling deviations. This invention provides a multi-timescale demand scheduling method and system that considers user response fatigue. This method constructs a dynamic user fatigue quantification model, establishes a nonlinear mapping from fatigue to fulfillment probability, and introduces the upper limit of effective schedulable capacity as an optimization constraint into the multi-objective model solution. Simultaneously, it uses actual execution results to update fatigue status and key parameters online, achieving rolling closed-loop optimization to improve fulfillment reliability and resource sustainability in scenarios with multiple consecutive demand responses.

[0007] The present invention adopts the following technical solution.

[0008] Firstly, a multi-timescale demand scheduling method that considers user response fatigue includes: Step 1: Obtain basic power system data and historical behavior data of demand response users, and construct a set of basic user profile parameters that includes multi-dimensional sensitivity parameters; Step 2: Employ a user dynamic fatigue quantification model to calculate the user's cumulative fatigue state value at the current scheduling time using the user's basic profile parameter set; wherein, the user dynamic fatigue quantification model includes a fatigue accumulation term that monotonically increases with response power and a fatigue recovery term that decays exponentially with non-response time. Step 3: Employ a fatigue-capacity physical mapping mechanism to calculate the user's fulfillment probability in the current fatigue state based on the accumulated fatigue state value, and map the fulfillment probability to the user's effective schedulable capacity limit in the next time period; Step 4: Construct a multi-objective optimization scheduling model to minimize system operating costs and user fatigue costs; Step 5: Using the effective schedulable capacity upper limit obtained in Step 3 as a rigid inequality constraint, solve the multi-objective optimization scheduling model in Step 4 to generate and issue scheduling instruction sequences for each user; update the user's cumulative fatigue state value based on the actual execution power of the scheduling instruction sequence as the initial state for the next scheduling cycle's rolling optimization.

[0009] Preferably, in step 1, the basic power system data specifically includes: load forecast data, renewable energy output forecast data, electricity purchase price, marginal cost parameters, and dispatch targets; the historical behavior data of demand response users specifically includes: the user's contracted capacity, historical response power, historical dispatch instruction sequence, historical instruction power, historical start-stop count, and historical continuous call duration.

[0010] Preferably, in step 1, constructing a user basic profile parameter set containing multi-dimensional sensitivity parameters specifically includes: Define a set of basic user profile parameters to be identified, including: multi-dimensional sensitivity parameters, including response intensity sensitivity coefficient, response duration sensitivity coefficient, and start / stop count sensitivity coefficient, used to describe the user's sensitivity to response intensity, response duration, and start / stop count; recovery speed parameters, used to describe the rate at which the user's physical strength recovers during non-response periods; probability curve steepness coefficient and fatigue threshold parameters, used to describe the transformation characteristics of the user from a fatigue state to the probability of fulfillment; intensity power function convexity exponent and duration fatigue exponent growth coefficient, used to describe the nonlinear characteristics of fatigue accumulation; and conservative margin initial value, used to describe the user's capacity safety boundary. Based on the historical behavior data, the actual performance result of the user at each historical scheduling moment is calculated; the actual performance result is the ratio of historical response power to historical instruction power at the corresponding moment. The user dynamic fatigue quantification model and the fatigue-capacity physical mapping mechanism are invoked. The historical scheduling instruction sequence, historical continuous call duration and historical start-stop count are used as input variables, and the user basic profile parameter set to be identified is used as model coefficients. The theoretical performance probability is output as the theoretical prediction performance result. An optimization objective function is constructed to calculate the mean square error between the theoretically predicted performance result and the actual performance result. The least squares method is used to iteratively solve the optimization objective function to find a set of parameter values ​​that minimize the mean square error, and these values ​​are determined as the offline fitting values ​​of the user basic profile parameter set.

[0011] Preferably, in step 2, calculating the user's cumulative fatigue state value at the current scheduling moment specifically includes: A user dynamic fatigue quantification model based on discrete-time state space is adopted, and the user's cumulative fatigue state value at the current moment is defined as the sum of the fatigue recovery term and the fatigue accumulation term. The calculation method of the fatigue recovery item includes: obtaining the user's cumulative fatigue state value at the previous scheduling time, and using the recovery speed parameter in the user's basic profile parameter set, calculating the user's fatigue recovery value after one scheduling cycle through the exponential decay function. The fatigue accumulation term is obtained by weighted summation of the response intensity sub-term, the duration sub-term, and the start-stop switching sub-term. The response intensity sub-term is calculated by weighting the power amplitude of the pre-scheduled command using the convexity exponent of the intensity power function and then using the response intensity sensitivity coefficient. The duration sub-term is calculated by weighting the continuous execution time of the pre-scheduled command using the duration fatigue exponent growth coefficient and then using the response duration sensitivity coefficient. The start-stop switching sub-term is calculated by weighting the number of start-stop actions triggered by the pre-scheduled command using the start-stop number sensitivity coefficient.

[0012] Preferably, step 3 specifically includes: The nonlinear mapping function is an S-shaped function model, used to map the cumulative fatigue state value to the theoretical performance probability; The S-shaped function model uses the fatigue threshold parameter as the critical inflection point of the probability curve to define the boundary of the user's transition from the high performance willingness zone to the low performance willingness zone, and uses the steepness coefficient of the probability curve as a control variable to define the nonlinear decay rate of the theoretical performance probability as the cumulative fatigue state value increases. Based on the theoretical fulfillment probability, the user's contracted capacity is corrected and calculated to obtain the user's effective schedulable capacity limit at the current scheduling time. Specifically, the correction calculation includes multiplying the user's contracted capacity by the theoretical fulfillment probability at the current moment, and deducting the safety buffer amount determined based on the initial conservative margin value to obtain the upper limit of the effective schedulable capacity.

[0013] Preferably, step 4 specifically includes: A multi-objective optimization scheduling model is constructed, wherein the objective function of the multi-objective optimization scheduling model consists of the grid power purchase cost, the explicit compensation cost of demand response, the fatigue loss cost, and the target gap penalty term; The grid purchase cost is constructed as a quadratic function of the grid purchase power. The explicit compensation cost for demand response is constructed as the sum of the products of the compensation unit price and the response power of all participating response objects in each time period; The fatigue cost is constructed as a combined function containing a fatigue equilibrium penalty term and a total fatigue penalty term; the fatigue equilibrium penalty term is calculated based on the squared difference between the cumulative fatigue state value of each user and the current average fatigue value of the group. The target gap penalty term is constructed as a gap penalty variable characterizing the failure to meet the system target adjustment power; A power balance equation constraint is constructed to require that the sum of the grid purchase power, the predicted output power of renewable energy, and the response power of all users equal the predicted load demand power.

[0014] Preferably, step 5 specifically includes: The multi-objective optimization scheduling model is solved to obtain the optimal scheduling instruction sequence in the prediction time domain, and the instruction with the first time step in the optimal scheduling instruction sequence is selected as the actual execution instruction and issued to each user. Collect the actual response power of each user to the executed command; use the user dynamic fatigue quantification model to calculate the user's real fatigue increment at the current moment based on the actual response power; The user's cumulative fatigue state value is updated using the actual fatigue increment, and the updated cumulative fatigue state value is used as the initial state variable for the next scheduling cycle rolling optimization. The scheduling time window is advanced by one time step, and then the process returns to step 3.

[0015] Secondly, a multi-timescale demand scheduling system that considers user response fatigue, comprising running the aforementioned multi-timescale demand scheduling method that considers user response fatigue, including: The data acquisition and profiling module is used to acquire basic power system data and historical behavior data of demand response users, and to build a set of basic user profile parameters that includes multi-dimensional sensitivity parameters. The user dynamic fatigue quantification module is used to calculate the cumulative fatigue state value of a user at the current scheduling time by using a user dynamic fatigue quantification model and a user basic profile parameter set. The user dynamic fatigue quantification model includes a fatigue accumulation term that monotonically increases with response power and a fatigue recovery term that decays exponentially with non-response time. The fatigue-capacity physical mapping module adopts a fatigue-capacity physical mapping mechanism to calculate the user's fulfillment probability in the current fatigue state based on the accumulated fatigue state value using a nonlinear mapping function, and maps the fulfillment probability to the user's effective schedulable capacity limit in the next time period. The multi-objective optimization scheduling modeling module is used to build multi-objective optimization scheduling models to minimize system operating costs and user fatigue loss costs. The rolling optimization and feedback execution module is used to solve the multi-objective optimization scheduling model of the multi-objective optimization scheduling modeling module by using the effective schedulable capacity upper limit obtained by the fatigue-capacity physical mapping module as a rigid inequality constraint, so as to generate and issue the scheduling instruction sequence for each user; and update the user's cumulative fatigue state value according to the actual execution power of the scheduling instruction sequence as the initial state for the rolling optimization of the next scheduling cycle.

[0016] Thirdly, an electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when loaded onto the processor, implements the aforementioned multi-timescale demand scheduling method that considers user response fatigue.

[0017] Fourthly, a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned multi-timescale demand scheduling method considering user response fatigue.

[0018] The beneficial effects of this invention are compared with those of the prior art: 1. Existing technologies typically rely on static assessments based on contracted capacity or historical average response rates, neglecting the psychological and physical fatigue effects on users in scenarios with repeated calls. This can easily lead to an overestimation of adjustable potential. This invention innovatively constructs a dynamic user fatigue quantification model. Through this refined modeling, the system can capture in real time the decline in user fulfillment capacity due to frequent calls, thereby more accurately predicting the actual available resources at the current moment and avoiding scheduling failures caused by inflated potential assessments.

[0019] 2. To address the issues of decreased response probability and increased execution deviation caused by user fatigue, this invention does not rely on a single prediction value. Instead, it establishes a nonlinear mapping function to transform the accumulated fatigue state into a response correction probability and further calculates the upper limit of the effective schedulable capacity.

[0020] 3. This invention constructs a multi-objective optimization model that includes fatigue loss costs. It transforms the multi-objective problem into a single-objective solution using a weighted normalization method and combines it with a rolling time-domain control strategy to achieve real-time updating and closed-loop correction of scheduling instructions. This invention effectively solves the problems of lack of fatigue mechanisms, static potential, and unsustainable scheduling in existing demand response scheduling, providing systematic methodological support for achieving highly reliable, sustainable, and intelligent demand response scheduling. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating a multi-timescale demand scheduling method that considers user response fatigue, as provided by the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this invention. The embodiments described in this application are merely some embodiments of this invention, and not all embodiments. Based on the spirit of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of this invention.

[0023] Example 1: like Figure 1 As shown, this invention provides a multi-timescale demand scheduling method that considers user response fatigue, including: Step 1: Obtain basic power system data and historical behavior data of demand response users, and construct a set of basic user profile parameters that includes multi-dimensional sensitivity parameters.

[0024] Acquiring basic power system data and historical behavior data of demand response users includes: Power system basic data and user behavior data are obtained through SCADA systems, AMI measurement systems, or power trading platform interfaces. The power system basic data specifically includes: load forecast data, renewable energy output forecast data, electricity purchase price, marginal cost parameters, and dispatch targets. The historical behavior data of demand response users specifically includes: users' contracted capacity, historical response power, historical dispatch command sequences, historical command power, historical start / stop counts, and historical continuous dispatch duration, etc.

[0025] Construct a user profile parameter set that includes multi-dimensional sensitivity parameters, specifically including: For each user Construct a set of basic user profile parameters to be identified , represented as:

[0026] Among them, the multidimensional sensitivity parameters are used to describe the user's sensitivity to response strength, response duration, and number of start / stop cycles, including: response strength sensitivity coefficient. This characterizes the user's sensitivity to power amplitude; the larger the value, the faster the user accumulates fatigue when performing high-power tasks; response time sensitivity coefficient. This represents the user's sensitivity to continuous job duration; a higher value indicates that the user is more prone to fatigue during long periods of continuous operation. The start / stop frequency sensitivity coefficient... This value represents the user's sensitivity to state transitions. The larger the value, the more significant the additional fatigue caused by frequent start-stop operations.

[0027] Recovery speed parameters This describes the rate at which a user recovers physical or mental capacity during periods of non-response (when command power is 0).

[0028] The probability curve morphology parameter describes the transformation pattern of a user's state from fatigue to fulfillment probability, including: the probability curve steepness coefficient. The fatigue threshold parameter describes the drastic decline in the contract fulfillment rate. The inflection point of the probability curve represents the user's tolerance limit.

[0029] Convexity index of power function of intensity and the fatigue index growth coefficient of duration This is used to describe the nonlinear characteristics of fatigue accumulation; specifically, It describes the phenomenon of increasing marginal fatigue, that is, when the response power doubles, the resulting fatigue does not just double, but increases dramatically in the form of a power function. It describes the cumulative effect over time, reflecting that as continuous working time increases, the increase in fatigue per unit time is not constant, but rather grows exponentially.

[0030] Conservative margin initial value This parameter is used to describe the user's capacity safety boundary. When converting the probability of fulfillment into the upper limit of physical capacity, this parameter can be introduced as a safety buffer deduction item.

[0031] Based on historical behavioral data, the actual performance results of users at each historical scheduling moment are calculated. It should be noted that, in this embodiment, in order to quantify the user's performance behavior, the actual performance result is the ratio of historical response power to historical instruction power at the corresponding moment. This processing is based on the assumption that the user's historical execution level in a statistical sense can objectively reflect the strength of their willingness to perform in the current state.

[0032] The user dynamic fatigue quantification model described in step 2 and the fatigue-capacity physical mapping mechanism described in step 3 are invoked. The historical scheduling instruction sequence, historical continuous call duration and historical start-stop count are used as input variables, and the user basic profile parameter set to be identified is used as model coefficients. The theoretical performance probability is output as the theoretical prediction performance result. An optimization objective function is constructed to calculate the mean square error between the theoretically predicted performance result and the actual performance result. The least squares method (such as the Levenberg-Marquardt algorithm) is used to iteratively solve the optimization objective function to find a set of parameter values ​​that minimize the mean square error, and these values ​​are determined as the offline fitting values ​​of the user basic profile parameter set.

[0033] Step 2: Employ a user dynamic fatigue quantification model, utilizing the user's basic profile parameter set, to calculate the user's cumulative fatigue state value at the current scheduling moment. This model includes a fatigue accumulation term that monotonically increases with response power and a fatigue recovery term that decays exponentially with non-response time. This quantifies the fatigue caused by frequent calls into a recursive cumulative fatigue state value. This is used to predict the fulfillment probability and effective capacity boundary in subsequent operations, ensuring that the state can be continued and corrected during rolling optimization.

[0034] Step 2 involves calculating the user's cumulative fatigue state value at the current scheduling moment, specifically including: Response intensity sub-item The power amplitude of the pre-scheduling command, after being corrected by the convexity exponent of the intensity power function, is then weighted using the response intensity sensitivity coefficient and expressed as:

[0035] In the formula, For a moment For users The planned demand response power (kW) issued; For users The maximum adjustable capacity of the contract; This is the response intensity sensitivity coefficient.

[0036] Duration sub-item The continuous execution time of the pre-scheduled instructions, after being corrected by the fatigue exponential growth factor of duration, is weighted using the response duration sensitivity coefficient and expressed as follows:

[0037]

[0038] In the formula, This indicates whether the function is invoked; if not invoked, then... The duration is continuously incremented when the function is called. Duration of continuous calls; This refers to the fatigue index growth coefficient based on duration. This is the response time sensitivity coefficient.

[0039] Start / Stop Switching Sub-item The number of start / stop actions triggered by pre-scheduled instructions is calculated using a start / stop count sensitivity coefficient, and is expressed as:

[0040] In the formula, This indicates whether a start / stop state switch has occurred at time t relative to time t-1; This is the sensitivity coefficient for the number of start-stop cycles.

[0041] In summary, the fatigue accumulation term is obtained by weighted summation of the response intensity sub-term, duration sub-term, and start / stop switching sub-term, and is expressed as:

[0042] In the formula, These are preset weighting coefficients. Different users have varying sensitivities to response intensity fatigue, response time fatigue, and start-stop fatigue. For different user types, domain experts determine the weights to achieve differentiated weight configurations for each user, while also satisfying the following requirements: .

[0043] The calculation method for the fatigue recovery term includes: obtaining the user's cumulative fatigue state value at the previous scheduling time; using the recovery speed parameter in the user's basic profile parameter set, calculating the user's fatigue recovery value after one scheduling cycle through an exponential decay function; and employing a user dynamic fatigue quantification model based on discrete-time state space, defining the user's cumulative fatigue state value at the current time as the sum of the fatigue recovery term and the fatigue accumulation term. This is expressed as:

[0044] In the formula, To restore the speed parameters, For the corresponding time number index, The scheduling time interval is typically 1 hour. For users The cumulative fatigue state value at this scheduling moment, when the user is not invoked, i.e. Fatigue value according to Recovery occurs through exponential decay; when the user invokes it, i.e. The fatigue value increases cumulatively from the previous moment. Furthermore, in this embodiment, the continuous call duration and cumulative fatigue state value at the initial time of the scheduling cycle (t=0) are both set to 0.

[0045] Step 3: Employ a fatigue-capacity physical mapping mechanism to calculate the user's fulfillment probability in the current fatigue state based on the accumulated fatigue state value using a nonlinear mapping function, and map the fulfillment probability to the user's effective schedulable capacity limit in the next time period.

[0046] This step aims to accumulate fatigue state values. The probability of converting into the user's "feasibility in the next time period". And further converted into the effective capacity limit. The effects of fatigue are injected into the subsequent optimization model in the form of "dynamic boundaries".

[0047] (1) Theoretical performance probability prediction A nonlinear mapping function is used to characterize the response correction probability. In this embodiment, the nonlinear mapping function is an sigmoid function model, used to map the accumulated fatigue state value to the theoretical performance probability. It is expressed as:

[0048] In the formula, To predict the theoretical fulfillment probability (range 0~1) of user i in the next scheduling period t+1, the sigmoid function model uses the fatigue threshold parameter. The critical inflection point of the probability curve, such as the fatigue value at a probability of 0.5, is used to define the boundary where users transition from a high willingness to fulfill obligations to a low willingness to fulfill obligations; the steepness coefficient of the probability curve is used as a reference. As a control variable, it is used to define the nonlinear decay rate of the theoretical performance probability as the cumulative fatigue state value increases. The larger the value, the more sensitive the user is to fatigue, and the steeper the decrease in probability.

[0049] (2) Effective schedulable capacity correction Based on the theoretical fulfillment probability, the user's contracted capacity is recalculated to obtain the user's effective schedulable capacity limit at the current scheduling moment. Specifically, the correction calculation includes: multiplying the user's contracted capacity by the theoretical fulfillment probability at the current moment, and subtracting the safety buffer amount determined based on the initial conservative margin value to obtain the upper limit of the effective schedulable capacity. This is expressed as:

[0050] In the formula, The maximum adjustable capacity (kW) for user i's contract. This serves as the initial conservative margin in the user profile. It should be noted that the above formula employs the modeling concept of probabilistic expectation. Although physically, users... It is fixed, but considering the uncertainty brought about by fatigue, this embodiment uses the response probability as the confidence coefficient to map the uncertain response behavior to a deterministic upper bound of the equivalent available capacity. The purpose of this approach is to automatically reduce the scheduling weight of highly fatigued users during the scheduling decision-making phase, thereby mitigating the risk of power deficit during actual execution in the mathematical model.

[0051] Step 4: Construct a multi-objective optimization scheduling model to minimize system operating costs and user fatigue costs.

[0052] This step comprehensively considers grid operation costs, demand response compensation costs, and fatigue loss costs within the prediction domain, and introduces a gap risk term to obtain an economical, reliable, and sustainable dispatch strategy under continuous multiple response scenarios.

[0053] (1) Multi-objective weighted objective function A multi-objective optimization scheduling model is constructed. The objective function of the multi-objective optimization scheduling model consists of the grid power purchase cost, the explicit compensation cost of demand response, the fatigue loss cost, and the target gap penalty term. It is expressed as:

[0054] in, For the current scheduling time, To predict the time index within the domain, This is the length of the prediction domain. The cost of purchasing electricity from the power grid; Explicitly compensate for costs in response to demand; Cost of fatigue wear; Let k be the target gap variable at time k, i.e., the amount of unmet regulation power.

[0055] To eliminate the interference of different physical dimensions on weight allocation and ensure the mathematical equivalence and comparability of various optimization objectives, each sub-objective is normalized:

[0056] in, Number the target; and These are the lower and upper bounds of the objective, respectively, which can be obtained from historical statistics, engineering experience upper bounds, or pre-solved solutions. To prevent extremely small constants with a denominator of zero, interval normalization based on statistical extrema is used. This eliminates the interference of different physical dimensions on weight allocation and ensures the mathematical equivalence and comparability of various optimization objectives.

[0057] Finally, a single-objective weighted optimization function is established:

[0058] in, The scalarized objective function after multi-objective normalization; The weighting coefficients can be set by the scheduling strategy, transforming a multi-objective problem into a single-objective solution.

[0059] (2) Electricity purchase cost of the power grid The grid purchase cost is constructed as a quadratic function of the grid purchase power, expressed as:

[0060] in, For a moment Power purchased from the grid (kW) The coefficients for the electricity purchase and operation cost function are determined by the electricity pricing mechanism or historical settlement data.

[0061] (3) Explicit compensation costs for demand response The explicit compensation cost for demand response is constructed as the sum of the products of the compensation unit price and the response power of all participating entities in each time period, expressed as:

[0062] in, The number of objects participating in the demand response. For object During the period The compensation unit price, Let be the response power of user i.

[0063] (4) Fatigue loss cost The fatigue cost is constructed as a combined function containing a fatigue equilibrium penalty term and a total fatigue penalty term; the fatigue equilibrium penalty term is calculated based on the squared difference between the cumulative fatigue state value of each user and the current average fatigue value of the group, and is expressed as follows:

[0064] in, For fatigue balance penalty weights, The fatigue total penalty weight is used to adjust the trade-off between "rotation fairness" and "overall fatigue level". A fatigue mean square error penalty term is introduced into the objective function. This enables the scheduling system to have an automatic rotation mechanism. When a user's fatigue level is higher than the group average, calls to that user will be automatically reduced, and users with lower fatigue levels will be called instead. This effectively avoids over-consumption of high-quality resources and extends the overall service life of the demand response resource pool.

[0065] Average fatigue value Represented as:

[0066] (5) Target gap penalty The target gap penalty term is constructed as a gap penalty variable characterizing the failure to meet the system's target regulation power. Small power gaps are allowed (soft constraints).

[0067] in, For time period The target regulation power required by the system Gap variables to meet the objectives.

[0068] (6) System operation constraints A power balance equation constraint is constructed to require that the sum of the grid's purchased power, the predicted output power of renewable energy, and the response power of all users equal the predicted load demand power. The power balance is expressed as:

[0069] in, For time period The predicted renewable energy output power (kW). For time period The predicted power of the load demand.

[0070] Step 5: Using the effective schedulable capacity upper limit obtained in Step 3 as a rigid inequality constraint, solve the multi-objective optimization scheduling model in Step 4 to generate and issue scheduling instruction sequences for each user; update the user's cumulative fatigue state value based on the actual execution power of the scheduling instruction sequence as the initial state for the next scheduling cycle's rolling optimization.

[0071] This step employs a Receding Horizon Control (RHC) strategy, injecting the "effective capacity upper limit after fatigue correction" as a hard constraint into the optimization solution, making the solution result inherently executable due to fatigue constraints; output instructions are then sent to the terminal for execution.

[0072] (1) Construction of an optimization model with fatigue constraints When solving the multi-objective optimization problem in step 4, the following key constraints are added: Effective capacity limit constraint:

[0073] in, For the binary decision variable within the prediction domain, used to represent the time period. Does the object need to be called? ; This is the upper limit of the effective schedulable capacity calculated in step 3.

[0074] State-space continuity constraints:

[0075] Solving for the optimal sequence using the weighted sum method:

[0076] in, The scalarized objective function after multi-objective normalization; The weighting coefficients can be set by the scheduling strategy, transforming a multi-objective problem into a single-objective solution.

[0077] (2) Rolling time domain control Operating days are divided into There are 1 basic time period, and the time period set is: Intraday scrolling optimization uses a fixed scrolling step size. (e.g., one time period) To predict the time domain length, i.e., the number of future time periods anticipated in each optimization. At each rolling start point. The prediction time domain is optimized for a future segment, with a prediction time domain length of [length missing]. Each time period.

[0078]

[0079] No. After each roll, time advances forward by one step:

[0080] (3) Solving and issuing instructions The multi-objective optimization scheduling model is solved to obtain the optimal scheduling instruction sequence in the prediction time domain. The instruction with the first time step in the optimal scheduling instruction sequence is selected as the actual execution instruction and issued to each user to control them to perform the corresponding power adjustment action.

[0081] (4) Closed-loop feedback and error correction After executing the instructions for the current time period, the following double correction is performed: First, the actual response power of each user to the executed command is collected and expressed as follows:

[0082] In the formula, This represents the actual response power. This refers to the power command issued. When When the time is insufficient, it indicates that the actual response is inadequate.

[0083] Secondly, using the aforementioned user dynamic fatigue quantification model, the user's actual fatigue increment at the current moment is calculated based on the actual response power, expressed as:

[0084] in, This is based on the user's actual response power. Substitute the values ​​into the formula in step 2 to calculate the actual fatigue increment.

[0085] Next, the prediction for the next cycle is performed. Using the execution deviation of this cycle, the system regulation target for the next cycle is corrected to achieve integral correction for total power balance.

[0086] (5) Iteration The user's cumulative fatigue state value is updated using the actual fatigue increment. The updated actual fatigue state is assigned to the initial state variable of the next round of optimization. The corrected target is substituted into the constraints, and the process returns to step 3 until the whole day's scheduling is completed.

[0087] Example 2: A multi-timescale demand scheduling system that considers user response fatigue, running the multi-timescale demand scheduling method considering user response fatigue as described in Embodiment 1, includes: The data acquisition and profiling module is used to acquire basic power system data and historical behavior data of demand response users, and to build a set of basic user profile parameters that includes multi-dimensional sensitivity parameters. The user dynamic fatigue quantification module is used to calculate the cumulative fatigue state value of a user at the current scheduling time by using a user dynamic fatigue quantification model and a user basic profile parameter set. The user dynamic fatigue quantification model includes a fatigue accumulation term that monotonically increases with response power and a fatigue recovery term that decays exponentially with non-response time. The fatigue-capacity physical mapping module is used to calculate the user's fulfillment probability in the current fatigue state based on the accumulated fatigue state value using a nonlinear mapping function, and to map the fulfillment probability to the user's effective schedulable capacity limit in the next time period. The multi-objective optimization scheduling modeling module is used to build multi-objective optimization scheduling models to minimize system operating costs and user fatigue loss costs. The rolling optimization and feedback execution module is used to solve the multi-objective optimization scheduling model of the multi-objective optimization scheduling modeling module by using the effective schedulable capacity upper limit obtained by the fatigue-capacity physical mapping module as a rigid inequality constraint, so as to generate and issue the scheduling instruction sequence for each user; and update the user's cumulative fatigue state value according to the actual execution power of the scheduling instruction sequence as the initial state for the rolling optimization of the next scheduling cycle.

[0088] To address the potential issues that the data collection and profiling module may encounter in practical applications, such as new users or users with missing historical data, this embodiment designs a parameter initialization strategy based on industry-specific similarity matching.

[0089] In practical applications, for newly connected users who have not participated in demand response or have insufficient historical data accumulation (e.g., data length less than 3 scheduling days), it is impossible to directly identify their profile parameters through historical data. Therefore, the specific steps are as follows: (1) Static feature extraction: First, obtain the static tag attributes of new users, including their industry category (such as steel, chemical, commercial complex, etc., determined by SIC / GICS codes), contracted capacity level, and typical daily load curve characteristics (such as peak-valley difference rate and load factor).

[0090] (2) Similar group clustering matching: Using K-Means clustering or K-Nearest Neighbors (KNN) algorithms, the system searches existing mature user databases for similar user clusters that have the closest Euclidean distance to the new user's static features. For example, for a new commercial office building user, the system will automatically match profile data of other commercial users of similar size and type in the database.

[0091] (3) Prior parameter transfer: Calculate the statistical mean or median of the set of basic user profile parameters for all users in the similar user cluster to generate a template of typical fatigue characteristics in the industry. Assign the parameters from this template (including multidimensional sensitivity parameters, recovery speed parameters, probability curve steepness coefficients, etc.) directly to new users as their initial set of profile parameters.

[0092] (4) Online rolling correction: As new users begin to participate in actual scheduling, the system activates online correction mode. Whenever new scheduling data is accumulated, the system uses Bayesian updates or a sliding window-based nonlinear least squares method (i.e., periodically re-running the parameter optimization algorithm in Example 1 using the newly accumulated data) to fine-tune and update the parameters using the initial profile parameter set as prior knowledge, combined with the newly generated actual performance data. This gradually achieves a smooth transition from industry-wide common parameters to individual personalized parameters.

[0093] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when loaded onto the processor, implements the multi-timescale demand scheduling method that takes into account user response fatigue.

[0094] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the multi-timescale demand scheduling method that considers user response fatigue.

[0095] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0096] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0097] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0098] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A multi-timescale demand scheduling method considering user response fatigue, characterized in that, include: Step 1: Obtain basic power system data and historical behavior data of demand response users, and construct a set of basic user profile parameters including multi-dimensional sensitivity parameters; Step 2: Using the user dynamic fatigue quantification model, the cumulative fatigue state value of the user at the current scheduling time is calculated using the user basic profile parameter set; wherein, the user dynamic fatigue quantification model includes a fatigue accumulation term that monotonically increases with response power and a fatigue recovery term that decays exponentially with non-response time. Step 3: Using a fatigue-capacity physical mapping mechanism, based on the accumulated fatigue state value, calculate the user's fulfillment probability in the current fatigue state, and map the fulfillment probability to the user's effective schedulable capacity limit in the next time period; Step 4: Construct a multi-objective optimization scheduling model to minimize system operating costs and user fatigue costs; Step 5: Using the effective schedulable capacity upper limit obtained in Step 3 as a rigid inequality constraint, solve the multi-objective optimization scheduling model in Step 4 to generate and issue scheduling instruction sequences for each user; update the user's cumulative fatigue state value based on the actual execution power of the scheduling instruction sequence as the initial state for the next scheduling cycle's rolling optimization.

2. The multi-timescale demand scheduling method considering user response fatigue according to claim 1, characterized in that, In step 1, the basic data of the power system specifically includes: load forecast data, renewable energy output forecast data, electricity purchase price, marginal cost parameters and dispatch targets; the historical behavior data of demand response users specifically includes: the user's contracted capacity, historical response power, historical dispatch instruction sequence, historical instruction power, historical start and stop times and historical continuous call duration.

3. The multi-timescale demand scheduling method considering user response fatigue according to claim 2, characterized in that, In step 1, the user profile parameter set containing multi-dimensional sensitivity parameters is constructed, specifically including: Define a set of basic user profile parameters to be identified, including: multi-dimensional sensitivity parameters, including response intensity sensitivity coefficient, response duration sensitivity coefficient, and start / stop count sensitivity coefficient, used to describe the user's sensitivity to response intensity, response duration, and start / stop count; recovery speed parameters, used to describe the rate at which the user's physical strength recovers during non-response periods; probability curve steepness coefficient and fatigue threshold parameters, used to describe the transformation characteristics of the user from a fatigue state to the probability of fulfillment; intensity power function convexity exponent and duration fatigue exponent growth coefficient, used to describe the nonlinear characteristics of fatigue accumulation; and conservative margin initial value, used to describe the user's capacity safety boundary. Based on the historical behavior data, the actual performance result of the user at each historical scheduling moment is calculated; the actual performance result is the ratio of historical response power to historical instruction power at the corresponding moment. The user dynamic fatigue quantification model and the fatigue-capacity physical mapping mechanism are invoked. The historical scheduling instruction sequence, historical continuous call duration and historical start-stop count are used as input variables, and the user basic profile parameter set to be identified is used as model coefficients. The theoretical performance probability is output as the theoretical prediction performance result. An optimization objective function is constructed to calculate the mean square error between the theoretically predicted performance result and the actual performance result. The least squares method is used to iteratively solve the optimization objective function to find a set of parameter values ​​that minimize the mean square error, and these values ​​are determined as the offline fitting values ​​of the user basic profile parameter set.

4. The multi-timescale demand scheduling method considering user response fatigue according to claim 3, characterized in that, Step 2 involves calculating the user's cumulative fatigue state value at the current scheduling moment, specifically including: A user dynamic fatigue quantification model based on discrete-time state space is adopted, and the user's cumulative fatigue state value at the current moment is defined as the sum of the fatigue recovery term and the fatigue accumulation term. The calculation method of the fatigue recovery item includes: obtaining the user's cumulative fatigue state value at the previous scheduling time, and using the recovery speed parameter in the user's basic profile parameter set, calculating the user's fatigue recovery value after one scheduling cycle through the exponential decay function. The fatigue accumulation term is obtained by weighted summation of the response intensity sub-term, the duration sub-term, and the start-stop switching sub-term. The response intensity sub-term is calculated by weighting the power amplitude of the pre-scheduled command using the convexity exponent of the intensity power function and then using the response intensity sensitivity coefficient. The duration sub-term is calculated by weighting the continuous execution time of the pre-scheduled command using the duration fatigue exponent growth coefficient and then using the response duration sensitivity coefficient. The start-stop switching sub-term is calculated by weighting the number of start-stop actions triggered by the pre-scheduled command using the start-stop number sensitivity coefficient.

5. The multi-timescale demand scheduling method considering user response fatigue according to claim 4, characterized in that, Step 3 specifically includes: The nonlinear mapping function is an S-shaped function model, used to map the cumulative fatigue state value to the theoretical performance probability; The S-shaped function model uses the fatigue threshold parameter as the critical inflection point of the probability curve to define the boundary of the user's transition from the high performance willingness zone to the low performance willingness zone, and uses the steepness coefficient of the probability curve as a control variable to define the nonlinear decay rate of the theoretical performance probability as the cumulative fatigue state value increases. Based on the theoretical fulfillment probability, the user's contracted capacity is corrected and calculated to obtain the user's effective schedulable capacity limit at the current scheduling time. Specifically, the correction calculation includes multiplying the user's contracted capacity by the theoretical fulfillment probability at the current moment, and deducting the safety buffer amount determined based on the initial conservative margin value to obtain the upper limit of the effective schedulable capacity.

6. The multi-timescale demand scheduling method considering user response fatigue according to claim 5, characterized in that, Step 4 specifically includes: A multi-objective optimization scheduling model is constructed, wherein the objective function of the multi-objective optimization scheduling model consists of the grid power purchase cost, the explicit compensation cost of demand response, the fatigue loss cost, and the target gap penalty term; The grid purchase cost is constructed as a quadratic function of the grid purchase power. The explicit compensation cost for demand response is constructed as the sum of the products of the compensation unit price and the response power of all participating response objects in each time period; The fatigue cost is constructed as a combined function containing a fatigue equilibrium penalty term and a total fatigue penalty term; the fatigue equilibrium penalty term is calculated based on the squared difference between the cumulative fatigue state value of each user and the current average fatigue value of the group. The target gap penalty term is constructed as a gap penalty variable characterizing the failure to meet the system target adjustment power; A power balance equation constraint is constructed to require that the sum of the grid purchase power, the predicted output power of renewable energy, and the response power of all users equal the predicted load demand power.

7. The multi-timescale demand scheduling method considering user response fatigue according to claim 6, characterized in that, Step 5 specifically includes: The multi-objective optimization scheduling model is solved to obtain the optimal scheduling instruction sequence in the prediction time domain, and the instruction with the first time step in the optimal scheduling instruction sequence is selected as the actual execution instruction and issued to each user. Collect the actual response power of each user to the executed command; use the user dynamic fatigue quantification model to calculate the user's real fatigue increment at the current moment based on the actual response power; The user's cumulative fatigue state value is updated using the actual fatigue increment, and the updated cumulative fatigue state value is used as the initial state variable for the next scheduling cycle rolling optimization. The scheduling time window is advanced by one time step, and then the process returns to step 3.

8. A multi-timescale demand scheduling system considering user response fatigue, running the multi-timescale demand scheduling method considering user response fatigue as described in any one of claims 1-7, characterized in that, include: The data acquisition and profiling module is used to acquire basic power system data and historical behavior data of demand response users, and to build a set of basic user profile parameters that includes multi-dimensional sensitivity parameters. The user dynamic fatigue quantification module is used to calculate the cumulative fatigue state value of a user at the current scheduling time by using a user dynamic fatigue quantification model and a user basic profile parameter set. The user dynamic fatigue quantification model includes a fatigue accumulation term that monotonically increases with response power and a fatigue recovery term that decays exponentially with non-response time. The fatigue-capacity physical mapping module adopts a fatigue-capacity physical mapping mechanism. Based on the accumulated fatigue state value, it uses a nonlinear mapping function to calculate the user's fulfillment probability in the current fatigue state and maps the fulfillment probability to the user's effective schedulable capacity limit in the next time period. The multi-objective optimization scheduling modeling module is used to build multi-objective optimization scheduling models to minimize system operating costs and user fatigue loss costs. The rolling optimization and feedback execution module is used to solve the multi-objective optimization scheduling model of the multi-objective optimization scheduling modeling module by using the effective schedulable capacity upper limit obtained by the fatigue-capacity physical mapping module as a rigid inequality constraint, so as to generate and issue the scheduling instruction sequence for each user; and update the user's cumulative fatigue state value according to the actual execution power of the scheduling instruction sequence as the initial state for the rolling optimization of the next scheduling cycle.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is loaded into the processor, it implements the multi-timescale demand scheduling method that takes into account user response fatigue as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-timescale demand scheduling method that takes into account user response fatigue as described in any one of claims 1-7.

Images