Method for scheduling cleaning operations and system for scheduling cleaning operations

Abstract

This application relates to the field of cleaning equipment scheduling technology, and provides a method and system for scheduling cleaning operations, including: acquiring electricity price forecast information for multiple time periods within the forecast time domain; acquiring load forecast information for multiple time periods within the forecast time domain; acquiring cleaning operation demand information, including cleaning duration demand information and completion deadline information; generating an optimal start-stop scheduling scheme for the cleaning equipment within the forecast time domain based on the electricity price forecast information, load forecast information, and operation demand information; and controlling the cleaning equipment to start or stop operation according to the optimal start-stop scheduling scheme. This application generates a start-stop scheduling scheme for the cleaning equipment with the optimization objective of minimizing cleaning electricity costs, making the start-stop periods of the cleaning operation adaptable to electricity price fluctuations, load fluctuations, and cleaning operation demands, reducing cleaning electricity costs, avoiding the start-up of high-power cleaning equipment during high-load periods, and improving the economic efficiency of unit operation.

Description

Technical Field

[0001] This application relates to the field of cleaning equipment scheduling technology, and in particular to a scheduling method and system for cleaning operations. Background Technology

[0002] The condenser is the core equipment of the cold-end system of a gas-fired combined cycle power plant, and the cleanliness of its heat exchange tubes directly affects the unit's vacuum level and operating economy. To maintain the cleanliness of the heat exchange tubes, power plants typically configure high-pressure cleaning pump stations to perform cleaning operations periodically. These cleaning operations are lengthy and consume a large amount of electricity. In related technologies, the start and stop times of cleaning operations are often arranged according to fixed cleaning cycles or the experience of operators. The pump station starts at the designated time and runs continuously for the preset duration before stopping.

[0003] With the implementation of time-of-use pricing in the electricity market, electricity prices vary significantly across different time periods during the day. Cleaning sessions under a fixed schedule may fall during periods of higher electricity prices, leading to increased cleaning electricity costs. Unit load fluctuates throughout the day, and cleaning sessions may overlap with periods of high unit load. Starting high-power cleaning pumps consumes plant power load, weakening the unit's peak output capacity and impacting its operational economy. Furthermore, the cleaning schedule is not adjusted according to changes in the condition of the object being cleaned. If the cleaning session is selected too early, the object may not yet require cleaning, resulting in wasted energy. If the cleaning session is selected too late, dirt accumulates, increasing unit performance loss. This mismatch between the cleaning schedule and the unit's operating environment restricts the overall economic efficiency of the cleaning operation. Summary of the Invention

[0004] Therefore, it is necessary to address the technical problem that the selection of start and stop times for cleaning operations in the existing technology lacks a reasonable basis, resulting in high cleaning electricity costs and affecting the economic efficiency of unit operation. A scheduling method and system for cleaning operations should be proposed.

[0005] Firstly, a scheduling method for cleaning operations is provided, comprising: acquiring electricity price forecast information for multiple time periods within the forecast time domain; acquiring load forecast information for multiple time periods within the forecast time domain; acquiring operation demand information for the cleaning operation, wherein the cleaning operation is performed by cleaning equipment, and the operation demand information includes cleaning duration demand information and completion time limit information; generating an optimal start-stop scheduling scheme for the cleaning equipment within the forecast time domain based on the electricity price forecast information, load forecast information, and operation demand information; and controlling the cleaning equipment to start or stop operation according to the optimal start-stop scheduling scheme. The optimal start-stop scheduling scheme is obtained by solving a scheduling optimization model; the scheduling optimization model uses the start-stop state of the cleaning equipment in each time period as the decision variable, and minimizes the cleaning electricity cost as the optimization objective; the cleaning electricity cost is determined based on the electricity price forecast information and the start-stop state; the constraints of the scheduling optimization model are determined based on the cleaning duration demand information, completion time limit information, and load forecast information.

[0006] Secondly, a scheduling system for cleaning operations is provided. The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the aforementioned scheduling method for cleaning operations.

[0007] Thirdly, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the above-described scheduling method for cleaning operations.

[0008] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described scheduling method for cleaning operations.

[0009] The cleaning operation scheduling method provided in this application reduces cleaning electricity costs by acquiring electricity price forecast information and scheduling cleaning operations during periods of lower electricity prices. This solves the problem in related technologies where cleaning periods overlap with high-electricity-price periods under fixed scheduling, leading to increased cleaning electricity costs. By acquiring load forecast information, the method ensures that the start-up and shutdown periods of cleaning operations avoid periods of high unit load, preventing the start-up of high-power cleaning equipment from occupying plant power load. This also solves the problem in related technologies where overlapping cleaning periods with high-load periods negatively impacts unit operating economy. Furthermore, by acquiring operation demand information that includes cleaning duration requirements and completion deadlines, the method ensures that cleaning operations are completed before the completion deadline while meeting the cleaning duration requirements. This solves the problem in related technologies where mismatches between cleaning schedules and the state of the cleaning objects lead to wasted cleaning energy or increased performance loss. Electricity price forecast information, load forecast information, and operation demand information are all incorporated into the scheduling optimization model. The optimal start-up and shutdown scheduling scheme is solved with the goal of minimizing cleaning electricity costs, thereby improving the overall economic efficiency of cleaning operations. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] in: Figure 1 This is a flowchart illustrating the scheduling method for cleaning operations in an embodiment of this application; Figure 2 This is a schematic diagram of the module of the cleaning operation scheduling system in the embodiments of this application; Figure 3 This is a structural block diagram of the computer device in the embodiments of this application; Figure 4 This is a structural block diagram of the computer device in the embodiments of this application. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] Please see Figure 1 As shown, Figure 1 A flowchart illustrating the cleaning operation scheduling method provided in this application embodiment includes the following steps: Step 01: Obtain electricity price forecast information for multiple time periods within the forecast time domain.

[0014] The prediction time domain refers to the pre-defined time span for this scheduling decision; in this example, it is 48 hours. A time period refers to the interval between two adjacent time periods after dividing the prediction time domain according to a preset time period step. The time period step is the time interval between the boundary moments of two adjacent time periods; in this example, the time period step is 15 minutes, and the prediction time domain contains a total of 192 time periods. Electricity price prediction information refers to the predicted electricity price value corresponding to each time period within the prediction time domain, representing the cost required to consume a unit of electricity during the cleaning operation in that time period.

[0015] Electricity price forecast information is obtained through an electricity price forecasting model. The input to the model is historical electricity price information and calendar feature information, and the output is the predicted electricity price value for each time period within the forecast time domain. Historical electricity price information refers to the electricity price sequence within a preset time period before the current moment. In this embodiment, electricity price data collected at 15-minute intervals over the past 30 days is used, totaling 2880 sampling points. Calendar feature information includes weekday, rest day, and holiday markers. In this embodiment, the electricity price forecasting model uses a Long Short-Term Memory (LSTM) network. LSTM is a recurrent neural network capable of modeling long-term dependencies in time series data. The input is the normalized historical electricity price sequence and calendar feature information, and the output, after inverse normalization, is the predicted electricity price value for each time period within the forecast time domain.

[0016] Step 02: Obtain load forecast information for multiple time periods within the forecast time domain.

[0017] Load forecasting information refers to the predicted unit load values ​​for each time period within the forecasting time domain. The predicted unit load values ​​represent the expected electrical power generated by the units during the corresponding time period. Load forecasting information is obtained through a load forecasting model. The input to the load forecasting model is historical load information and meteorological information, and the output is the predicted load values ​​for each time period within the forecasting time domain. Historical load information refers to the unit load sequence within a preset time period prior to the current moment. In this embodiment, load data collected at 15-minute intervals over the past 14 days is used. Meteorological information includes environmental parameters such as temperature, humidity, and wind speed. In this embodiment, the load forecasting model uses an extreme gradient boosting (XGBoost) regression model. Extreme gradient boosting is an ensemble learning model based on a gradient boosting framework. Its features include the lag values ​​of historical loads, meteorological information, hour sequence number, week sequence number, and holiday markers. The output is the predicted load values ​​for each time period within the forecasting time domain.

[0018] Step 03: Obtain the cleaning operation requirements information.

[0019] The cleaning operation is performed by cleaning equipment, which refers to high-power equipment used to perform cleaning operations on the object to be cleaned. In this embodiment, the cleaning equipment is an online condenser cleaning pump station, the object to be cleaned is the condenser, and the online condenser cleaning pump station performs row-by-row fixed-point cleaning of the condenser heat exchange tubes with high-pressure water.

[0020] The job requirement information refers to the job constraints that this cleaning operation must meet, including cleaning duration requirements and completion deadlines. The cleaning duration requirement refers to the total cumulative cleaning time required for this operation, which is the total time the cleaning equipment is actually in operation; in this example, it is taken as 10 hours. The completion deadline refers to the latest time that this cleaning operation should be completed. If the completion time is later than the completion deadline, the performance loss suffered by the cleaned object due to continuous accumulation of dirt exceeds the allowable range. The completion deadline can be preset by the operators or directly issued by the cleaning management system.

[0021] Step 04: Based on electricity price forecast information, load forecast information, and work demand information, generate the optimal start-stop scheduling scheme for the cleaning equipment within the forecast time domain.

[0022] The scheduling optimization model uses the start-stop status of the cleaning equipment in each time period as the decision variable and minimizes the cleaning electricity cost as the optimization objective. The start-stop status refers to the operating state of the cleaning equipment during the corresponding time period. When the cleaning equipment is performing cleaning during the time period, the start-stop status is the first state; when the cleaning equipment stops working during the time period, the start-stop status is the second state. In the embodiment, the first state is represented by the value 1, and the second state is represented by the value 0.

[0023] Cleaning electricity costs refer to the electricity costs incurred by the cleaning equipment during cleaning operations within the predicted time period. These costs are determined based on electricity price forecasts and the equipment's start-up and shutdown status. The specific calculation process is as follows: For each time period within the predicted time period, multiply the operating power of the cleaning equipment, the time period step, and the predicted electricity price for that time period to obtain the corresponding cleaning electricity cost. Then, sum the cleaning electricity costs for each time period to obtain the total cleaning electricity cost for the predicted time period. The total cleaning electricity cost is calculated using the following formula: C_elec = Σ_{t=1}^{T} x_t × P × Δt × ρ_t Wherein, C_elec is the total cleaning electricity cost in the prediction time domain, in yuan; t is the time period number, ranging from 1 to T; T is the total number of time periods in the prediction time domain, in this example T is 192; x_t is the start / stop status of the t-th time period, x_t is 1 when cleaning is performed in the t-th time period, and x_t is 0 when cleaning is stopped in the t-th time period; P is the operating power of the cleaning equipment, in kW, in this example, the rated power of the condenser online cleaning pump station is 185kW, the average operating power is 140kW, and P is 140kW; Δt is the time period step, in hours, in this example Δt is 0.25 hours; ρ_t is the predicted electricity price for the t-th time period, in yuan / kWh.

[0024] The constraints of the scheduling optimization model are determined based on the cleaning duration requirement information, completion deadline information, and load forecast information. These constraints limit the feasible range of values ​​for start-up and shutdown states. In this embodiment, the constraints include cumulative cleaning duration constraints, completion deadline constraints, and load constraints.

[0025] The cumulative cleaning time constraint means that the cumulative time during which the cleaning equipment is in operation within the predicted time domain equals the cleaning time required by the cleaning time demand information. The cumulative cleaning time is calculated using the following formula: T_acc = Σ_{t=1}^{T} x_t × Δt Where T_acc is the cumulative cleaning time of the cleaning equipment in the prediction time domain, in hours; x_t is the start / stop status of the t-th time period; Δt is the time period step. The mathematical form of the cumulative cleaning time constraint is T_acc = T_need, where T_need is the cleaning time corresponding to the cleaning time requirement information. In this embodiment, T_need is taken as 10 hours.

[0026] The completion time constraint means that the cleaning operation must be completed no later than the completion time corresponding to the completion time limit information. The completion time of the cleaning operation refers to the moment when the cumulative cleaning time of the cleaning equipment within the predicted time domain reaches the cleaning time limit corresponding to the cleaning time requirement information, i.e., the end time corresponding to the last cleaning session. The mathematical form of the completion time constraint is t_end ≤ T_deadline, where t_end is the completion time of the cleaning operation, and T_deadline is the completion time corresponding to the completion time limit information.

[0027] Load constraint means that the predicted load value corresponding to the start-up period of the cleaning equipment is not greater than a preset load threshold. The start-up period refers to the time when the start-up state switches from the second state to the first state. When the predicted load value corresponding to the start-up period of the cleaning equipment is not higher than the preset load threshold, the impact of the cleaning equipment start-up on the peak output capacity of the unit is within the allowable range. The mathematical form of the load constraint is L_t ≤ L_thr, where L_t is the predicted load value of the t-th period in MW; L_thr is the preset load threshold in MW. The preset load threshold is determined based on the rated load of the unit. In this embodiment, the rated load of the unit is 470MW, and the preset load threshold is 80% of the rated load, i.e., 376MW.

[0028] In another embodiment, the constraints may further include start-stop frequency constraints and minimum continuous cleaning duration constraints. The start-stop frequency constraint limits the number of times the cleaning equipment can be started within the predicted time domain to no greater than a preset start-stop frequency threshold, which is set to 2 in this embodiment. The minimum continuous cleaning duration constraint limits the duration of continuous operation of the cleaning equipment after any start-up to no less than a preset minimum continuous duration, which is set to 2 hours in this embodiment. These two constraints can reduce the starting current surge caused by frequent start-stop of high-power motors and extend the service life of the cleaning equipment.

[0029] Solving the scheduling optimization model involves searching for combinations of start-stop states that satisfy the constraints within the prediction time domain, and selecting the combination that minimizes the total electricity cost for cleaning as the optimal start-stop scheduling scheme. The optimal start-stop scheduling scheme includes the start-stop states for each time period within the prediction time domain, indicating the start time, stop time, and operating time distribution of the cleaning equipment.

[0030] When solving the scheduling optimization model, in addition to exhaustive search, a mixed integer programming algorithm can be used. This algorithm treats start-up and shutdown states as integer variables and transforms constraints into linear inequalities, then searches for the optimal solution using the branch and bound method. Electricity price forecasting information can also be obtained by directly reading the time-of-use (TOU) price tables published by the electricity market, using the electricity prices for each time period in the TOU table as the corresponding predicted price values. Load forecasting information can also be obtained by directly using the day-ahead planned load of the generating units, which is the load arrangement for each time period of the following day predetermined by the scheduling plan.

[0031] Step 05: Control the cleaning equipment to start or stop working according to the optimal start-stop scheduling scheme.

[0032] In the optimal start-stop scheduling scheme, at the start of the time period when the start-stop state is in the first state, a start command is sent to the control system of the cleaning equipment, and the cleaning equipment enters the running state to perform cleaning. At the end of the time period when the start-stop state switches from the first state to the second state, a stop command is sent to the control system of the cleaning equipment, and the cleaning equipment exits the running state.

[0033] The cleaning operation scheduling method provided in this application incorporates electricity price forecast information into scheduling decisions, ensuring that the cleaning duration falls within periods of lower predicted electricity prices, thereby reducing the total electricity cost for cleaning. It also incorporates load forecast information into scheduling decisions, ensuring that cleaning equipment starts during periods of lower unit load, guaranteeing peak unit output capacity. Furthermore, it incorporates cleaning duration requirements and completion deadlines into constraints, guaranteeing timely completion of cleaning operations and avoiding insufficient or excessive cleaning. The scheduling optimization model balances electricity prices, load, and cleaning operation requirements, adapting the start and stop times of cleaning operations to fluctuations in electricity prices, load, and cleaning operation demands, thus improving the overall economic efficiency of cleaning operations and reducing unit operating costs.

[0034] In some embodiments, step 03 above includes: Obtain the dirt status information of the object to be cleaned corresponding to the cleaning operation. The dirt status information is determined based on the image information inside the object to be cleaned. Determine the level of dirtiness based on the dirtiness status information; The completion time is determined based on the level of dirtiness.

[0035] Specifically, the dirt status information refers to a quantitative description of the degree of dirt inside the object being cleaned. In this embodiment, the object being cleaned is a condenser, and the dirt status information is the percentage of dirt area at the condenser heat exchanger tube inlets. Image information refers to image data obtained by photographing the interior of the object being cleaned. This image information is acquired by a camera installed in the condenser water chamber. The camera is a waterproof and corrosion-resistant industrial camera, with its lens facing the heat exchanger tube array area. The camera is equipped with an underwater supplementary lighting device to ensure image clarity in the heat exchanger tube area under circulating water conditions. The camera captures images of the heat exchanger tube area at a preset sampling interval, which is 15 minutes in this embodiment. The acquired images are transmitted to an image processing server via Ethernet or fiber optic cable.

[0036] The degree of soiling refers to the result of dividing a continuous measure of soiling into a finite number of discrete levels. In this embodiment, the degree of soiling levels include light, moderate, and heavy. The degree of soiling is determined based on the percentage of soiled area, which is calculated using the following formula: r = S_dirt / S_total Where r represents the percentage of the dirty area; S_dirt represents the area of ​​the dirty region identified in the image; and S_total represents the total area of ​​the heat exchanger tube inlets. The dirty region area is extracted from the image using an image segmentation algorithm. This algorithm performs pixel-level classification, dividing pixels into dirty and non-dirty categories. The dirty region area is obtained by multiplying the number of dirty pixels by the actual area of ​​each pixel. In this embodiment, when r is less than 0.30, the dirtiness level is mild; when r is greater than or equal to 0.30 and less than 0.70, the dirtiness level is moderate; and when r is greater than or equal to 0.70, the dirtiness level is severe.

[0037] The completion time corresponding to the completion deadline is determined based on the degree of dirtiness. The higher the degree of dirtiness, the more severe the dirt accumulation, and the earlier the cleaning operation needs to be completed. In this example, when the degree of dirtiness is light, the completion time corresponding to the completion deadline is the current time plus 96 hours; when the degree of dirtiness is moderate, the completion time corresponding to the completion deadline is the current time plus 72 hours; and when the degree of dirtiness is heavy, the completion time corresponding to the completion deadline is the current time plus 24 hours.

[0038] Image information can also be used to identify dirty regions using a semantic segmentation network based on deep learning. The semantic segmentation network is a convolutional neural network with an encoder-decoder structure. It outputs the probability that each pixel belongs to the dirty category and determines the pixels with a probability greater than a preset probability threshold as dirty regions. The area of ​​dirty regions can also be extracted through threshold segmentation or clustering algorithms.

[0039] In some implementations, step 04 above includes: Obtain information on expected performance improvements for cleaning operations; Determine the expected benefits of the cleaning operation based on information about anticipated performance improvements; The optimal start-stop scheduling scheme is determined by maximizing the difference between expected revenue and cleaning electricity cost.

[0040] Specifically, the expected performance improvement information refers to the expected performance improvement of the cleaned object after the cleaning operation. In this example, the cleaned object is the condenser, and the expected performance improvement information is the expected increase in condenser vacuum, expressed in kPa. Condenser vacuum improvement refers to the improvement in the vacuum level corresponding to the condenser exhaust pressure after the cleaning operation. After the condenser vacuum level is improved, the unit's gas consumption for power supply decreases.

[0041] Expected revenue refers to the value after converting expected performance improvements into economic worth. Expected revenue is determined based on expected performance improvements, unit load information, gas price information, and the revenue assessment period. Expected revenue is calculated using the following formula: R_clean = ΔP_vac × k_g × C_gas × L_avg × (H_benefit - (t_end -t_now)) Where R_clean represents the expected revenue in yuan; ΔP_vac represents the expected increase in condenser vacuum corresponding to the expected performance improvement information in kPa; k_g is the gas consumption sensitivity coefficient, which refers to the reduction in unit power supply gas consumption corresponding to each unit increase in condenser vacuum. According to the "Analysis of Energy Consumption Indicators of Gas Generator Sets" (DL / T 2100-2020), for a 470 MW gas generator set, each 1 kPa increase in condenser vacuum reduces power supply gas consumption by 0.889 cubic meters per megawatt-hour. In this example, k_g is taken as 0.889, in cubic meters per megawatt-hour in kilopascals; C_gas represents the gas price in yuan per cubic meter, in this example, it is taken as 2.5 yuan per cubic meter; L_avg is the average unit load in the prediction time domain in MW. L_avg is the arithmetic mean of the predicted load values ​​for each period in the prediction time domain, calculated according to the following formula: L_avg = (1 / T) × Σ_{t=1}^{T} L_t Where L_t is the load forecast value for the t-th time period, and T is the total number of time periods; H_benefit is the benefit assessment period in hours. The benefit assessment period refers to the time span during which the condenser vacuum improvement continues to generate benefits after the cleaning operation is completed. In this embodiment, the benefit assessment period is the interval between the current cleaning operation and the planned next cleaning operation, i.e., the cleaning period, which is 96 hours; t_now is the current time; t_end is the completion time of the cleaning operation, and t_end-t_now is the delay time of the completion time relative to the current time in hours.

[0042] As shown by the above formula, the earlier the completion time, the shorter the delay time, and the greater the expected benefit. With the optimization objective of maximizing the difference between expected benefit and cleaning electricity cost, the scheduling optimization model balances reducing cleaning electricity cost with completing cleaning ahead of time to obtain benefits. The resulting optimal start-stop scheduling scheme completes the cleaning operation as early as possible while ensuring a low cleaning electricity cost, avoiding revenue loss due to delayed cleaning.

[0043] In this embodiment, the expected increase in condenser vacuum ΔP_vac is taken as 0.374 kPa, the gas consumption sensitivity coefficient k_g is taken as 0.889, the gas price C_gas is taken as 2.5 yuan per cubic meter, the average unit load L_avg is taken as 350 MW, the benefit assessment period H_benefit is taken as 96 hours, and the delay time between the completion time and the current time is taken as 14 hours. Therefore, the expected benefit R_clean = 0.374 × 0.889 × 2.5 × 350 × 82 ≈ 23862 yuan. The cleaning electricity cost, calculated at the flat rate, is approximately 868 yuan, and the difference between the expected benefit and the cleaning electricity cost is 22994 yuan. If the completion time is delayed to a delay of 20 hours, the expected benefit R_clean = 0.374 × 0.889 × 2.5 × 350 × 76 ≈ 22116 yuan. With the cleaning electricity cost remaining unchanged, the difference decreases to 21248 yuan. Therefore, completing the cleaning operation ahead of schedule can significantly increase the difference between the expected revenue and the cleaning electricity cost. Incorporating revenue factors into the dispatching decision can avoid revenue loss caused by delaying cleaning in pursuit of low electricity prices.

[0044] Expected benefits can also be expressed as gas savings, calculated as ΔP_vac × k_g × L_avg ×(H_benefit - (t_end - t_now)). The expected benefits are obtained by multiplying the gas savings by the gas price. The benefit assessment period can also be dynamically adjusted according to the level of pollution, with a smaller value used when the pollution level is higher.

[0045] In some implementations, obtaining the expected performance improvement information for the cleaning operation in step 04 above includes: Obtain the dirt status information of the objects to be cleaned in the corresponding cleaning operation; Obtain the circulating water temperature information of the object to be cleaned in the cleaning operation; Based on the dirt and grime status information and circulating water temperature information, the expected performance improvement information is determined.

[0046] Specifically, the circulating water temperature information refers to the temperature of the cooling medium of the object being cleaned. In this embodiment, the circulating water temperature information is the inlet temperature of the condenser circulating water, expressed in degrees Celsius. The circulating water temperature information is collected in real time by a temperature sensor installed in the condenser circulating water inlet pipe. The circulating water temperature information can also be obtained through a water temperature prediction model. This model predicts the temperature based on the coupling relationship between historical circulating water temperature and air temperature. The predicted circulating water temperature for the t-th time period within the prediction time domain is calculated using the following formula: T_cw,t = T_hist(t) + α × (T_air,t - T_air,hist(t)) Wherein, T_cw,t is the predicted circulating water temperature for the t-th time period in the prediction time domain, in degrees Celsius; T_hist(t) is the average circulating water temperature at the corresponding historical time for the t-th time period, in degrees Celsius. In this embodiment, T_hist(t) is the average circulating water temperature at the same time as the t-th time period in the 5 days prior to the current time; α is the air temperature coupling coefficient, used to characterize the degree of response of the circulating water temperature to changes in air temperature. In this embodiment, α is taken as 0.6; T_air,t is the predicted air temperature for the t-th time period, in degrees Celsius; T_air,hist(t) is the average air temperature at the corresponding historical time for the t-th time period, in degrees Celsius, which is the average air temperature at the same time as the t-th time period in the 5 days prior to the current time.

[0047] The expected performance improvement information is determined based on the fouling status information and circulating water temperature information. In this embodiment, the expected increase in condenser vacuum is calculated using the following formula: ΔP_vac = a_0 + a_1 × Δβ + a_2 × (T_cw,avg - T_ref) + a_3 × (L_avg - L_ref) Wherein, ΔP_vac is the expected increase in condenser vacuum, in kPa; Δβ is the expected increase in cleanliness coefficient for this cleaning, which is the ratio of the actual heat transfer coefficient of the condenser to the designed heat transfer coefficient. Δβ is determined according to the degree of fouling. In this embodiment, Δβ is 0.05 for light fouling, 0.10 for moderate fouling, and 0.15 for heavy fouling; T_cw,avg is the average circulating water temperature in the predicted time domain, in degrees Celsius, representing the circulating water temperature at each time period in the predicted time domain. The arithmetic mean of the predicted water temperature values; T_ref is the reference circulating water temperature, which is 25 degrees Celsius in this example; L_avg is the average unit load in the prediction time domain, in MW; L_ref is the reference load, which is 350MW in this example; a_0, a_1, a_2, and a_3 are revenue prediction parameters, which are calibrated from historical cleaning case data. In this example, a_0 is -0.05, a_1 is 4.0, a_2 is 0.008, and a_3 is 0.0002.

[0048] As shown in the formula above, the higher the circulating water temperature, the greater the expected increase in condenser vacuum. When the circulating water temperature is higher, the heat exchange temperature difference between the circulating water and the exhaust steam is smaller, making the condenser vacuum level more sensitive to changes in the cleanliness coefficient. Under the same cleanliness coefficient improvement conditions, the condenser vacuum improvement is more significant, and the marginal benefit of the cleaning operation is greater. Incorporating circulating water temperature information into the determination process of expected performance improvement information makes the expected performance improvement information more compatible with the water temperature environment of the cleaning target, resulting in more accurate calculations of expected benefits and improved rationality of scheduling decisions.

[0049] The circulating water temperature information can also be read directly from the unit's distributed control system. The condenser circulating water inlet temperature is a regular monitoring parameter of the distributed control system. The revenue prediction model can also use nonlinear models such as support vector regression or Gaussian process regression.

[0050] In some implementations, step 04 above includes: Using time period number and cumulative cleaning duration as states, a recursive relationship is established for the scheduling optimization model. The recursive relationship represents the correspondence between the cleaning electricity cost of the current state and the cleaning electricity cost of the previous state, the electricity price prediction information of the current time period, and the start / stop status of the current time period. Based on the recursive relationship, the optimal start-stop scheduling scheme is determined from the states that satisfy the constraints.

[0051] Specifically, dynamic programming (DP) is used to solve the scheduling optimization model. Dynamic programming is an algorithmic framework that decomposes a complex problem into interrelated subproblems and solves them layer by layer. The state is defined by the time period number and the cumulative cleaning duration. The state refers to a portion of the decision results recorded during the scheduling process. In this embodiment, the state is represented by a tuple (t, s), where t is the time period number and s is the cumulative number of cleaning periods executed up to the t-th time period.

[0052] Recurrence relations are used to describe the transition relationships between states. A recurrence relation is established using the following formula: dp[t][s]= min( dp[t-1][s], dp[t-1][s-1]+ P × Δt × ρ_t ) Wherein, dp[t][s] is the minimum cleaning electricity cost corresponding to the t-th time period and the cumulative number of cleaning time periods is s, in yuan; dp[t-1][s] is the minimum cleaning electricity cost corresponding to the (t-1)-th time period and the cumulative number of cleaning time periods is s; dp[t-1][s-1] is the minimum cleaning electricity cost corresponding to the (t-1)-th time period and the cumulative number of cleaning time periods is s-1; P is the operating power of the cleaning equipment; Δt is the time period step; ρ_t is the predicted electricity price for the t-th time period. The first term dp[t-1][s] in the recursive relation corresponds to the case where the work is stopped in the t-th time period, and the cumulative number of cleaning time periods remains unchanged; the second term dp[t-1][s-1] + P × Δt × ρ_t corresponds to the case where cleaning is performed in the t-th time period, the cumulative number of cleaning time periods increases from s-1 to s, and the cleaning electricity cost increases by the cleaning electricity cost of the t-th time period.

[0053] The boundary conditions for the recursive relation are: dp[0][0] = 0, indicating that when no time period has been traversed and the cumulative number of cleaning time periods is 0, the minimum cleaning electricity cost is 0; dp[0][s] = +∞, indicating that when s is greater than 0, the cumulative number of cleaning time periods cannot be greater than 0 when no time period has been traversed. When the load constraint is violated in the t-th time period, the t-th time period can only stop working, and only the first term of the recursive relation is retained.

[0054] After the recursion is completed, the state corresponding to the minimum cleaning electricity cost is selected from the endpoint states that meet the constraints. The endpoint state refers to the state where the cumulative number of cleaning periods is equal to the number of cleaning periods corresponding to the cleaning duration requirement information, and the completion time is not later than the completion time corresponding to the completion deadline information. By backtracking along the recorded transition path, the start and stop states of each period in the prediction time domain are obtained, which is the optimal start and stop scheduling scheme. In this embodiment, the prediction time domain contains 192 periods, the maximum number of cumulative cleaning periods is 40, the state scale is approximately 7680, and the computational complexity of the solution is on the order of milliseconds, which meets the requirements of rolling real-time solution.

[0055] When using dynamic programming to solve the problem, the state can be expanded to include a tuple containing the number of current continuous cleaning periods and the number of times it has been started, in order to handle the constraints of the shortest continuous cleaning duration and the number of start / stop times. Alternatively, a greedy algorithm can be used to solve the problem. The greedy algorithm allocates the cleaning duration to the periods that meet the constraints in ascending order of the electricity price forecast. If the allocation result does not meet the completion time constraint, the allocation of the cleaning period is adjusted backward from the completion time.

[0056] In some implementations, the training process of the scheduling optimization model includes: Acquire historical cleaning case data, which includes information on the state of dirt and the actual performance improvement after each cleaning operation; Based on the dirt status information and actual performance improvement information in historical cleaning case data, the revenue prediction parameters are determined. The revenue prediction parameters are used to determine the expected revenue based on the expected performance improvement information. After the cleaning operation is completed, obtain the actual execution data of the cleaning operation, which includes actual performance improvement information; Update the revenue prediction parameters based on the actual performance improvement information in the actual execution data.

[0057] Specifically, historical cleaning case data refers to the data collection of all previous cleaning operations of the unit. Each data entry in the historical cleaning case data corresponds to a complete cleaning event, and each data entry includes the dirt status information corresponding to a cleaning operation and the actual performance improvement information after the cleaning is completed. The dirt status information is the degree of dirtiness obtained through image recognition before the cleaning operation begins, and the actual performance improvement information is the actual increase in condenser vacuum obtained through performance testing after the cleaning operation is completed. In the example, a gas turbine unit of a power plant performed approximately 50 condenser cleaning operations in the past two years. For each cleaning operation, dirt status information, cleaning parameters, vacuum test data before and after cleaning, and cleaning duration were recorded, forming a total of 50 historical cleaning case data entries.

[0058] The revenue prediction parameters are the regression coefficients in the revenue prediction model. This model determines the expected revenue based on anticipated performance improvement information. The process for determining the revenue prediction parameters is as follows: A sample set is constructed from the dirt status information and actual performance improvement information from historical cleaning case data. The input for each sample is the cleaning coefficient improvement and circulating water temperature correction corresponding to the dirt status information, and the output is the actual performance improvement information. The values ​​of the revenue prediction parameters are determined using least squares estimation. Least squares estimation is a parameter estimation method that minimizes the sum of squared residuals, where the residuals are the difference between the sample output and the model output. The revenue prediction parameters are determined using the following formula: θ = (XᵀX)⁻¹Xᵀy Wherein, θ is the revenue prediction parameter vector, containing four parameters: a_0, a_1, a_2, and a_3; X is the sample design matrix, where each row of X corresponds to a historical cleaning case data point, and the columns of X correspond to constant terms, cleaning coefficient improvement, circulating water temperature correction, and load correction; Xᵀ is the transpose of X; y is the sample output vector, where each element corresponds to the actual performance improvement information in each historical cleaning case data point; (XᵀX)⁻¹ is the inverse matrix of XᵀX. In this embodiment, the sample size is 50, and the revenue prediction parameter vector θ is a 4-dimensional vector.

[0059] After the cleaning operation is completed, the actual execution data of this cleaning operation is added to the historical cleaning case data as a new sample, and the least squares estimation is re-executed to update the benefit prediction parameters. The actual performance improvement information in the actual execution data is the condenser vacuum increase measured after the cleaning operation. In this embodiment, the sample size of the historical cleaning case data is capped at 50 records. When the sample size reaches the limit, the oldest sample is discarded before adding a new sample. The benefit prediction parameters are continuously updated as the cleaning operation is performed, and the fitting accuracy of the benefit prediction model to the current operating characteristics of the unit gradually improves, resulting in more accurate calculations of the expected benefits.

[0060] The revenue forecast parameters can also be solved iteratively using the gradient descent method, which is an optimization method that iteratively updates parameters along the negative gradient direction of the loss function. The actual execution data can also be used to update the revenue forecast parameters using recursive least squares estimation. When a new sample arrives, the parameter matrix is ​​updated recursively without needing to recalculate the matrix inverse. The prediction errors of electricity price forecast information and load forecast information are exponentially weighted smoothed. When the absolute value of the smoothed error continuously exceeds a preset error threshold, the corresponding prediction model is triggered to retrain. In this example, the preset error threshold is set to 0.10.

[0061] In some implementations, the method for scheduling cleaning operations further includes: In each scheduling cycle, update the electricity price forecast information and load forecast information; Based on the updated electricity price forecast information and the updated load forecast information, the scheduling optimization model is re-solved; Based on the first updated optimal start-stop scheduling scheme obtained by resolving the problem, control the cleaning equipment to start or stop working in the next scheduling cycle; During the period when the cleaning equipment is started, obtain the actual load information; When the deviation between the actual load information and the load forecast information is greater than the preset load deviation threshold, the start-up of the cleaning equipment is delayed. The scheduling optimization model is solved again to determine the second updated optimal start-stop scheduling scheme.

[0062] Specifically, the scheduling cycle refers to the time interval for rolling scheduling decisions. At the end of the scheduling cycle, electricity price forecast information and load forecast information are regenerated using the latest historical data, and the scheduling optimization model is resolved based on the updated forecast information to obtain the first updated optimal start-stop scheduling scheme. In this embodiment, the scheduling cycle is 15 minutes. Electricity price forecasting, load forecasting, and scheduling optimization model solving are performed once in each scheduling cycle, and start-stop control is only performed within the next scheduling cycle.

[0063] The preset load deviation threshold refers to the maximum allowable deviation between the actual load information and the load forecast information. When the deviation between the actual load information and the load forecast information is greater than the preset load deviation threshold, it indicates that the actual change in the unit load exceeds the forecast range. Starting the cleaning equipment according to the original dispatch plan may cause the actual load corresponding to the start-up period of the cleaning equipment to be higher than the preset load threshold. In this embodiment, the preset load deviation threshold is taken as 5% of the unit's rated load. When the unit's rated load is 470MW, the preset load deviation threshold is taken as 23.5MW.

[0064] During the period when the cleaning equipment is controlled to start, the actual load information is obtained in real time from the unit's distributed control system (DCS). The actual load information is compared with the load forecast information for the corresponding period. If the deviation is greater than a preset load deviation threshold, the start-up of the cleaning equipment is delayed. The scheduling optimization model is then re-solved using the current time as a new starting point to determine the second updated optimal start-up and shutdown scheduling scheme. The second updated optimal start-up and shutdown scheduling scheme is used to indicate the start-up and shutdown status for the remaining period. Delaying the start-up can prevent the cleaning equipment from starting when the load exceeds expectations, thus ensuring the unit's peak output capacity.

[0065] When executing scheduling decisions on a rolling basis, the prediction time domain and the control time domain are separated, with the control time domain being shorter than the prediction time domain. In this embodiment, the control time domain is set to a single time period, and only the start-stop control for the next time period is executed. The start-stop states corresponding to the remaining time periods within the prediction time domain are used as backup references. The cleaning equipment is started using a ramp-up method, with the speed gradually increasing from zero to the rated speed within a preset start-up duration. In this embodiment, the preset start-up duration is 60 seconds to reduce the impact of the high-power motor's starting current on the plant's power system. The start-up period can also be used to verify the plant's power load by superimposing the starting power of the cleaning equipment with the power consumption of other equipment in the plant. If the superimposed result exceeds a preset plant power threshold, the start-up of the cleaning equipment is delayed.

[0066] In some implementations, after controlling the cleaning equipment to start or stop working according to the optimal start-stop scheduling scheme, the scheduling method for cleaning operations further includes: Obtain performance recovery information during the cleaning operation. The performance recovery information is determined based on the performance parameters of the object being cleaned. Determine the performance recovery rate based on the performance recovery information; When the performance recovery rate is less than the preset recovery rate threshold, the cleaning equipment will be controlled to stop working in advance.

[0067] Specifically, performance recovery information refers to data on how the performance of the object being cleaned changes over time during the cleaning operation. In this embodiment, performance recovery information refers to the recovery process data of the condenser cleanliness coefficient. The cleanliness coefficient is the ratio of the actual heat transfer coefficient of the condenser to the designed heat transfer coefficient. During the cleaning operation, the dirt on the condenser heat exchange tubes is gradually removed, and the cleanliness coefficient gradually recovers. Performance parameters are operating parameters used to calculate the cleanliness coefficient, including the condenser terminal temperature difference and vacuum level. During the cleaning operation, the condenser terminal temperature difference and vacuum level are collected at preset sampling intervals, and the cleanliness coefficient is calculated based on the collected data. In this embodiment, the sampling interval is 15 minutes.

[0068] The process of restoring the cleanliness coefficient is described by the following formula: β(t) = β_clean - (β_clean - β_dirty) × e^(-t / τ) Wherein, β(t) is the cleaning coefficient at the t-th sampling time after the start of cleaning; β_dirty is the cleaning coefficient before the start of cleaning; β_clean is the designed cleaning coefficient, which is 1.0 in this embodiment; τ is the recovery time constant, which characterizes the speed at which the cleaning coefficient recovers. The smaller the value of τ, the faster the cleaning coefficient recovers. In this embodiment, τ is 4 hours.

[0069] Performance recovery rate refers to the rate of change of the cleaning coefficient relative to the cleaning time. The performance recovery rate is calculated using the following formula: dβ / dt = ((β_clean - β_dirty) / τ) × e^(-t / τ) Where dβ / dt is the performance recovery rate at the t-th sampling time, in units of hour; the remaining parameters have the same meaning as those in the formula for the cleanliness coefficient recovery process. As can be seen from the above formula, the performance recovery rate is relatively high at the beginning of the cleaning process, and the cleanliness coefficient rises rapidly. As the cleaning duration increases, the performance recovery rate gradually decreases and approaches zero, and the cleanliness coefficient enters a plateau period. At this point, continuing the cleaning process has limited contribution to improving the cleanliness coefficient.

[0070] The preset recovery rate threshold refers to the lower limit of the allowable performance recovery rate. When the performance recovery rate is less than the preset recovery rate threshold, the cleaning process is determined to have entered a plateau phase, and the cleaning equipment is controlled to stop working early. In this embodiment, the preset recovery rate threshold is set to 0.005 per hour. When the performance recovery rate is less than the preset recovery rate threshold for four consecutive sampling times, the cleaning equipment is controlled to stop working early. Early shutdown of the cleaning equipment saves cleaning energy during the plateau phase and releases cleaning equipment resources for use by other cleaning tasks.

[0071] The recovery time constant can also be determined through online fitting. During the cleaning process, the cleaning coefficient is collected at a preset sampling interval, and the relationship between the cleaning coefficient and the cleaning time is regression-fitted. The recovery time constant is determined by the slope of the fitted curve. The recovery time constant is dynamically updated with the cleaning process. The early stop judgment can also be replaced by the performance recovery increment judgment. When the performance recovery increment at multiple consecutive sampling times is less than the preset increment threshold, the cleaning equipment is controlled to stop working early.

[0072] In some implementations, the cleaning operation includes a first cleaning task and a second cleaning task, generating an optimal start-stop scheduling scheme for the cleaning equipment in the predicted time domain, including: Obtain the first operational requirements information for the first cleaning task and the second operational requirements information for the second cleaning task; The optimal start-stop scheduling scheme is determined by minimizing the sum of the cleaning electricity costs corresponding to the first cleaning task and the second cleaning task. The constraints of the scheduling optimization model include shared equipment constraints, which stipulate that the number of cleaning tasks executed by the cleaning equipment in the same period is no more than one.

[0073] Specifically, the first cleaning task and the second cleaning task are two cleaning tasks that need to be executed sequentially by the same cleaning equipment. In this embodiment, the first cleaning task corresponds to the condenser cleaning operation of the first unit, and the second cleaning task corresponds to the condenser cleaning operation of the second unit. The condensers of the two units share the same cleaning pump station. The first operation requirement information is the operation requirement information for the first cleaning task, including the first cleaning duration requirement information and the first completion deadline information; the second operation requirement information is the operation requirement information for the second cleaning task, including the second cleaning duration requirement information and the second completion deadline information.

[0074] The shared equipment constraint means that the number of cleaning tasks performed by a cleaning device in the same time period is no more than one; that is, the cleaning device can perform at most one of the first and second cleaning tasks in the same time period. The mathematical form of the shared equipment constraint is x_{1,t} + x_{2,t} ≤ 1, where x_{1,t} represents the start / stop status of the first cleaning task in the t-th time period; x_{1,t} being 1 indicates that the first cleaning task is performed in the t-th time period, and x_{1,t} being 0 indicates that the first cleaning task is not performed in the t-th time period. Similarly, x_{2,t} represents the start / stop status of the second cleaning task in the t-th time period; x_{2,t} being 1 indicates that the second cleaning task is performed in the t-th time period, and x_{2,t} being 0 indicates that the second cleaning task is not performed in the t-th time period. The shared equipment constraint prevents the first and second cleaning tasks from overlapping in time periods, ensuring that the cleaning device provides cleaning water to only one cleaning task at any given time.

[0075] The sum of the cleaning electricity costs for the first cleaning task and the second cleaning task is calculated using the following formula: C_total = Σ_{t=1}^{T} x_{1,t} × P × Δt × ρ_t + Σ_{t=1}^{T} x_{2,t} × P × Δt × ρ_t Where C_total represents the total electricity cost for the first and second cleaning tasks, in yuan; x_{1,t} represents the start / stop status of the first cleaning task in time period t; x_{2,t} represents the start / stop status of the second cleaning task in time period t; P represents the operating power of the cleaning equipment; Δt represents the time period step; and ρ_t represents the predicted electricity price for time period t. The optimization objective is to minimize the total electricity cost for cleaning. This is achieved under the constraints of shared equipment and the operational requirements of each cleaning task, ensuring that the two cleaning tasks are staggered in the predicted time domain, with their respective cleaning durations falling within periods with lower predicted electricity prices, thus reducing the overall electricity cost for both tasks.

[0076] When solving the scheduling optimization model that includes the first cleaning task and the second cleaning task, the state can be expanded to a tuple containing the cumulative number of cleaning periods for the two cleaning tasks. The cumulative number of cleaning periods for at most one cleaning task increases in the same time period. In the example, the cumulative number of cleaning periods for the first cleaning task is 40 and the cumulative number of cleaning periods for the second cleaning task is 24, with a state size of approximately 184,000. The computational complexity of the solution is still within the range of real-time solvability. The first and second cleaning tasks can also be prioritized, with the cleaning task with the earlier completion time corresponding to the completion time information being scheduled first.

[0077] Please see Figure 2 As shown, Figure 2 This is a schematic diagram of the scheduling system for cleaning operations in this embodiment. The acquisition module acquires electricity price forecasts for multiple time periods within the forecast time domain, load forecasts for multiple time periods within the forecast time domain, and operation demand information for the cleaning operation. The cleaning operation is performed by cleaning equipment, and the operation demand information includes cleaning duration requirements and completion deadlines. The strategy generation module generates an optimal start-stop scheduling scheme for the cleaning equipment within the forecast time domain based on the electricity price forecasts, load forecasts, and operation demand information. The control module controls the cleaning equipment to start or stop operation according to the optimal start-stop scheduling scheme.

[0078] The cleaning operation scheduling system provided in this application includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of the cleaning operation scheduling method described above.

[0079] In one embodiment, a computer device is provided, which can be a server, and its internal structure diagram can be as follows. Figure 3As shown, it includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs. The database stores historical electricity price information, historical load information, and historical cleaning case data. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a cleaning operation scheduling method on the server side.

[0080] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows. Figure 4 As shown, the system includes a processor, memory, network interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs. The network interface is used to communicate with an external server via a network connection. The display screen is an LCD screen or an e-ink screen. The input device is a touch layer covering the display screen. When the computer program is executed by the processor, it implements the functions or steps of a cleaning operation scheduling method on the client side.

[0081] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described cleaning operation scheduling method.

[0082] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0083] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory and flash memory, etc., and volatile memory can include random access memory, etc.

[0084] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0085] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit the application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for scheduling cleaning operations, characterized in that, include: Obtain electricity price forecast information for multiple time periods within the forecast time domain; Obtain load forecast information for multiple time periods within the forecast time domain; Obtain the operation requirement information for the cleaning operation, which is performed by cleaning equipment. The operation requirement information includes cleaning duration requirement information and completion time limit information. Based on the electricity price forecast information, the load forecast information, and the work demand information, an optimal start-stop scheduling scheme for the cleaning equipment is generated within the forecast time domain. According to the optimal start-stop scheduling scheme, the cleaning equipment is controlled to start or stop working; The optimal start-stop scheduling scheme is obtained by solving the scheduling optimization model; The scheduling optimization model uses the start-up and shutdown status of the cleaning equipment in each time period as the decision variable and minimizes the cleaning electricity cost as the optimization objective. The cleaning electricity cost is determined based on the electricity price forecast information and the start-up and shutdown status. The constraints of the scheduling optimization model are determined based on the cleaning duration requirement information, the completion time limit information, and the load forecast information.

2. The method for scheduling cleaning operations according to claim 1, characterized in that, The process of generating the optimal start-stop scheduling scheme for the cleaning equipment within the predicted time domain includes: Obtain the expected performance improvement information for the cleaning operation; The expected benefits of the cleaning operation are determined based on the expected performance improvement information. The optimal start-stop scheduling scheme is determined by maximizing the difference between the expected revenue and the cleaning electricity cost.

3. The method for scheduling cleaning operations according to claim 2, characterized in that, The step of obtaining the expected performance improvement information for the cleaning operation includes: Obtain the dirt status information of the object to be cleaned corresponding to the cleaning operation; Obtain the circulating water temperature information of the object to be cleaned corresponding to the cleaning operation; Based on the dirt and grime status information and the circulating water temperature information, the expected performance improvement information is determined.

4. The method for scheduling cleaning operations according to claim 1, characterized in that, The step of obtaining the operation requirement information for the cleaning operation includes: Obtain the dirt status information of the object to be cleaned corresponding to the cleaning operation, and the dirt status information is determined based on the image information inside the object to be cleaned; The level of dirtiness is determined based on the dirtiness status information; The completion time corresponding to the completion time limit information is determined based on the level of dirtiness.

5. The method for scheduling cleaning operations according to claim 1, characterized in that, The process of generating the optimal start-stop scheduling scheme for the cleaning equipment within the predicted time domain includes: Using time period number and cumulative cleaning duration as states, a recursive relationship is established for the scheduling optimization model. The recursive relationship represents the correspondence between the cleaning electricity cost of the current state and the cleaning electricity cost of the previous state, the electricity price prediction information of the current time period, and the start / stop status of the current time period. Based on the recursive relationship, the optimal start-stop scheduling scheme is determined from the states that satisfy the constraints.

6. The method for scheduling cleaning operations according to claim 2, characterized in that, The training process of the scheduling optimization model includes: Acquire historical cleaning case data, which includes dirt status information corresponding to each cleaning operation and actual performance improvement information after cleaning; Based on the dirt status information and actual performance improvement information in the historical cleaning case data, a revenue prediction parameter is determined. The revenue prediction parameter is used to determine the expected revenue based on the expected performance improvement information. After the cleaning operation is completed, the actual execution data of the cleaning operation is obtained, and the actual execution data includes actual performance improvement information; The revenue prediction parameters are updated based on the actual performance improvement information in the actual execution data.

7. The method for scheduling cleaning operations according to claim 1, characterized in that, The method further includes: In each scheduling cycle, the electricity price forecast information and the load forecast information are updated; Based on the updated electricity price forecast information and the updated load forecast information, the scheduling optimization model is re-solved; Based on the first updated optimal start-stop scheduling scheme obtained by resolving the problem, the cleaning equipment is controlled to start or stop working in the next scheduling cycle; During the period when the cleaning equipment is started, the actual load information is obtained; When the deviation between the actual load information and the predicted load information is greater than a preset load deviation threshold, the start-up of the cleaning equipment is delayed. The scheduling optimization model is solved again to determine the second updated optimal start-stop scheduling scheme.

8. The method for scheduling cleaning operations according to claim 1, characterized in that, After controlling the cleaning equipment to start or stop working according to the optimal start-stop scheduling scheme, the method further includes: Obtain performance recovery information during the execution of the cleaning operation, wherein the performance recovery information is determined based on the performance parameters of the cleaning object corresponding to the cleaning operation; Determine the performance recovery rate based on the performance recovery information; When the performance recovery rate is less than a preset recovery rate threshold, the cleaning equipment is controlled to stop working in advance.

9. The method for scheduling cleaning operations according to claim 1, characterized in that, The cleaning operation includes a first cleaning task and a second cleaning task. Generating the optimal start-stop scheduling scheme for the cleaning equipment within the predicted time domain includes: Obtain the first job requirement information for the first cleaning task and the second job requirement information for the second cleaning task; The optimal start-stop scheduling scheme is determined with the goal of minimizing the sum of the cleaning electricity costs corresponding to the first cleaning task and the second cleaning task.

10. A scheduling system for cleaning operations, characterized in that, It includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the scheduling method for cleaning operations as described in any one of claims 1 to 9.

Images