Building and smart grid collaborative optimization control method and system

Abstract

The application discloses a building and smart grid cooperative optimization control method and system, performs peak-valley matching analysis on building operation state data and electricity price information to generate an energy consumption state baseline; extracts a peak-valley electricity price period boundary after generating an electricity price response strategy, identifies a trend turning point to establish a peak-valley differentiated scheduling path; starts a thermal recovery system based on a device cooperative parameter to reserve a frequency modulation response margin to generate a peak period scheduling execution record; according to a peak period actual pressure reduction amount, back calculates a valley period filling amount, starts ice making and energy storage filling, and then performs peak-valley energy dynamic allocation to construct a cooperative optimization configuration; monitors a power grid load to generate a load distribution curve, determines an optimal scheduling time through multi-objective optimization of electricity cost and load stability, executes building and power grid cooperative control to output a cooperative optimization control instruction, and realizes release of peak-valley arbitrage potential of building energy storage resources and cooperative suppression of power grid load.

Description

Technical Field

[0001] This invention relates to the field of building energy management technology, and in particular to a method and system for collaborative optimization control of buildings and smart grids. Background Technology

[0002] Buildings serve as a crucial flexible load resource for urban power grids. The peak electricity consumption of their HVAC systems highly overlaps with peak electricity market prices, leading to persistently high building operating costs and exacerbating the pressure of concentrated regional loads on the power grid. Existing building energy management methods primarily rely on fixed-time-period strategies to control equipment start-up and shutdown, lacking the ability to dynamically perceive real-time electricity price fluctuations. The operating plans of chillers and energy storage devices cannot adaptively adjust to changes in electricity prices. Furthermore, the charging and discharging plans for various energy storage media within buildings, such as ice making and thermal storage, lack joint optimization. The independent operation of each energy storage system results in both wasted resources during off-peak periods and insufficient available energy storage during peak periods. Regarding the collaborative interaction between buildings and the power grid, current mainstream practices rely on manually preset demand response rules, leading to crude judgments of response timing. Building HVAC equipment often misses the optimal intervention window, resulting in low overall efficiency of collaborative interaction between buildings and the power grid.

[0003] Therefore, a method is urgently needed to solve at least one of the above problems. Summary of the Invention

[0004] This invention discloses a building-smart grid collaborative optimization control method and system. Addressing issues such as peak-valley mismatch in building heating and power supply, insufficient joint optimization of energy storage resources, and delayed equipment response, the method constructs differentiated scheduling paths by matching peak-valley energy consumption status and dynamically hierarchically layering electricity price ranges. It constrains the orderly intervention of the energy storage system using equipment collaborative parameters and uses the actual peak-period load reduction to infer the off-peak load plan. After determining the optimal scheduling time through multi-objective optimization, it outputs building-grid collaborative control commands, achieving efficient utilization of building energy storage resources and coordinated load mitigation of the grid.

[0005] The first aspect of this invention proposes a collaborative optimization control method for buildings and smart grids, comprising the following steps:

[0006] Collect building HVAC system operation status data and electricity market price information, and perform peak-valley energy consumption status matching analysis on the operation status data and the electricity price information to generate an energy consumption status baseline;

[0007] Based on the energy consumption status baseline, peak-valley electricity price optimization analysis is performed to generate an electricity price response strategy. The electricity price response strategy is then implemented to extract peak-valley electricity price time boundaries in a hierarchical manner. Based on the peak-valley electricity price time boundaries, trend turning points are identified in advance to establish peak-valley differentiated scheduling paths.

[0008] The peak-valley differentiated scheduling path is analyzed for collaborative response to generate equipment collaborative parameters. Based on the equipment collaborative parameters, the energy storage and heat recovery system is started with frequency regulation response margin to generate peak scheduling execution records.

[0009] Based on the actual reduction in the peak period scheduling execution record, the valley period charging amount is calculated to generate a valley period charging time sequence scheme. The valley period charging time sequence scheme is used to start ice making and energy storage charging to generate valley period energy storage parameters. The valley period energy storage parameters are used to perform dynamic peak and valley energy allocation and compensation to build a collaborative optimization configuration.

[0010] Based on the aforementioned collaborative optimization configuration, a load distribution curve is generated by monitoring the power grid load. The load distribution curve is then subjected to multi-objective optimization of electricity cost and load stability to determine the optimal scheduling time. At the optimal scheduling time, the building and power grid collaborative control is executed to output collaborative optimization control commands.

[0011] A second aspect of this invention provides a building and smart grid collaborative optimization control system, comprising:

[0012] The data acquisition module is used to collect the operating status data of the building's HVAC system and the electricity market price information, and to perform peak-valley energy consumption status matching analysis on the operating status data and the electricity price information to generate an energy consumption status baseline;

[0013] The strategy generation module is used to perform peak-valley electricity price optimization analysis based on the energy consumption status baseline to generate an electricity price response strategy, implement electricity price interval layering to extract peak-valley electricity price time boundaries for the electricity price response strategy, and identify trend turning points in advance based on the peak-valley electricity price time boundaries to establish peak-valley differentiated scheduling paths.

[0014] The peak scheduling module is used to perform collaborative response analysis on the peak-valley differentiated scheduling path to generate equipment collaborative parameters, and based on the equipment collaborative parameters, retain the frequency regulation response margin to start the energy storage and heat recovery system to generate peak scheduling execution records.

[0015] The off-peak energy storage module is used to generate an off-peak charging time sequence scheme by back-calculating the off-peak charging amount based on the actual reduction amount in the peak scheduling execution record, and to start ice making and energy storage charging using the off-peak charging time sequence scheme to generate off-peak energy storage parameters. The off-peak energy storage parameters are used to perform dynamic peak-valley energy allocation and compensation to build a collaborative optimization configuration.

[0016] The instruction output module is used to generate a load distribution curve by monitoring the power grid load based on the collaborative optimization configuration, perform multi-objective optimization of the load distribution curve on electricity cost and load stability to determine the optimal scheduling time, and execute the building and power grid collaborative control to output collaborative optimization control instructions at the optimal scheduling time.

[0017] The beneficial effects of this invention are reflected in the following points: First, by classifying the operating status data of building HVAC systems according to peak and valley periods, energy consumption characteristics are extracted and dynamically paired with real-time electricity price intervals to identify periods of overconsumption and underconsumption, establishing a quantitative correspondence between building electricity consumption behavior and electricity price signals. Based on this, peak and valley electricity price period boundaries are extracted through dynamic electricity price stratification thresholds, and continuous unidirectional inflection points are used to identify electricity price trend turning points and mark the advance timing of scheduling switching. This enables the building's peak and valley scheduling plan to respond in advance at the beginning of the electricity price change trend, solving the problem of delayed equipment action timing under fixed rule strategies. Second, by calculating the linkage coefficient of each equipment action node in the peak and valley differentiated scheduling path, strong collaborative constraint pairs are identified and master-slave equipment power balance constraint equations are established. Frequency regulation response margin is reserved when the energy storage and heat recovery systems are started in stages, achieving orderly collaborative output of building chillers, ice storage tanks, and heat recovery devices during peak-period reduction, avoiding hydraulic imbalance and over-response problems caused by independent adjustment of each device. Finally, the actual consumption of each energy storage medium during peak periods is used to calculate the charging target during off-peak periods. The energy gap distribution is identified through peak-valley energy balance verification, and the charging amount is calibrated in segments according to compensation priority. Dynamic allocation and compensation are carried out in conjunction with the measured state of charge in the energy storage parameters during off-peak periods. After determining the optimal scheduling time, the time with the best comprehensive score of electricity cost and load stability is used as the trigger point for building and grid collaborative control. This realizes closed-loop collaborative optimization of energy storage system charging and discharging plan and grid load scheduling. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a collaborative optimization control method for buildings and smart grids according to the present invention.

[0019] Figure 2 This is a structural block diagram of a building and smart grid collaborative optimization control system according to the present invention. Detailed Implementation

[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0021] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0022] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] The technical solutions of the embodiments of this application will be described below.

[0024] like Figure 1 As shown, this embodiment of the invention provides a building and smart grid collaborative optimization control method, including the following steps S110-S150:

[0025] Step S110: Collect the operating status data of the building's HVAC system and the electricity market price information, and perform peak-valley energy consumption status matching analysis on the operating status data and electricity price information to generate an energy consumption status baseline.

[0026] Specifically, the system collects operational status data of the building's HVAC system and electricity market price information. Operational status data is collected in real-time through a sensor network deployed at key nodes of the HVAC system. The data includes real-time power, inlet and outlet water temperatures, and operating frequencies of chillers, cooling towers, air conditioning units, and fresh air handling units. The sampling period is set to 1 minute, and the clocks of each sensor are synchronized with a standard time server via the NTP protocol, with a time alignment error not exceeding 50 milliseconds. The raw operational status data undergoes outlier removal and missing value imputation preprocessing. Points with power values ​​exceeding 120% of the rated power are considered outliers and replaced with the average of adjacent points. If more than 5 consecutive sampling periods are missing data, an equipment status check is triggered, and data collection on the corresponding channel is suspended. The operational status data is organized and stored using a dual index of equipment number and timestamp, covering three subsystems: chillers, distribution systems, and terminal equipment. The daily data volume is approximately 1440 rows multiplied by the number of sensor channels. Electricity price information is retrieved in real time from the power market dispatch platform via API interface. The retrieval frequency is synchronized with the sampling period of the operational status data. The electricity price information includes two categories: real-time electricity price and day-ahead forecast electricity price. The real-time electricity price reflects the settlement price at the current moment, while the day-ahead forecast electricity price is used to plan dispatch strategies in advance. Electricity price information is recorded with two fields: timestamp and price type. The typical intraday variation range of real-time electricity price is 0.35 to 1.20 yuan per kilowatt-hour. When the peak-valley price difference exceeds 0.5 yuan per kilowatt-hour, the peak-valley arbitrage space on the building side is significant. Both operational status data and electricity price information are collected under the same time base, and the timestamp alignment error between the two data sources does not exceed a single sampling period.

[0027] In some embodiments, the step of performing peak-valley energy consumption status matching analysis on the operating status data and the electricity price information to generate an energy consumption status baseline includes: extracting energy consumption features from the operating status data according to peak and valley time periods to generate an energy consumption feature distribution; dividing the electricity price information into peak and valley intervals to generate peak and valley electricity price interval labels; comparing and matching the energy consumption feature distribution with the peak and valley electricity price interval labels to generate an energy consumption deviation mark; and generating an energy consumption status baseline based on the energy consumption deviation mark.

[0028] Energy consumption characteristics are extracted from operational status data according to peak and valley periods to generate energy consumption characteristic distributions. Simply extracting the daily average will mask the differences in energy consumption structure between peak and valley periods; extraction must be categorized by time period to accurately reflect the energy consumption behavior of each period. The initial division of peak and valley periods is based on the local electricity market standard time period definition. Peak periods are typically from 8:00 to 11:00 and 18:00 to 21:00, while valley periods are from 23:00 to 7:00 the next day. Each sampling point in the operational status data is assigned to its corresponding time period category according to its timestamp. The equipment power time series within each time period in the operational status data is extracted using three statistical measures: mean, peak, and standard deviation. The mean reflects the average power level of that period, the peak reflects the highest load point, and the standard deviation reflects the degree of load fluctuation. For example, the standard deviation of chiller power in a building during a valley period is 12 kW, while the standard deviation during a peak period is 45 kW. The significantly greater load fluctuation during the peak period indicates unplanned interference during peak hours. The three statistical measures for each time period are arranged in a two-dimensional structure according to equipment type and time period category to form an energy consumption feature distribution. The item density of the energy consumption feature distribution is positively correlated with the number of equipment in the operation status data. The more types of equipment covered, the more complete the energy consumption feature distribution is in characterizing the energy consumption behavior of each subsystem. If a device in the operation status data has more than 30% invalid sampling points in a specific time period, the feature extraction result of that device in that time period is marked with a low confidence label. The low confidence label reduces the weight of the item during the deviation matching stage. When the proportion of low confidence items in the energy consumption feature distribution exceeds 15%, a sensor inspection suggestion is triggered.

[0029] Electricity price information is divided into peak and valley periods to generate peak and valley price range labels. Fixed peak and valley range division based on clock times introduces price assignment errors near the boundaries. Therefore, the range boundaries must be dynamically identified based on the actual values ​​of the electricity price information. Boundary identification uses periods where the average price in a 15-minute sliding window continuously increases by more than 0.05 yuan per kilowatt-hour every 15 minutes as the peak entry boundary, and periods where the average price continuously decreases by more than a threshold as the peak exit boundary. The real-time electricity price at each sampling point in the electricity price information is compared with the identified range boundaries. Points falling into the peak range and with a price higher than 1.3 times the daily median are marked with peak-hour labels; points falling into the valley range and with a price lower than 0.7 times the daily median are marked with valley-hour labels; and the remaining points are marked with normal-hour labels. Among the three types of labels, the peak-hour label has the strongest constraint on building scheduling. If a brief fluctuation in electricity price occurs within the normal-hour label period, label reclassification is not triggered to ensure the stability of the label sequence. Peak-valley electricity price range labels record the label type for each sampling point throughout the day, indexed by timestamps. The sum of the time spans for the three types of labels—peak hour, valley hour, and normal hour—equals the entire 24 hours of the day. When there is a classification conflict between the predicted electricity price and the real-time electricity price in the electricity price information, the peak-valley electricity price range label of the real-time electricity price shall prevail. If an anomaly occurs in the daily electricity price information where there are consecutive peak-entry boundaries without peak-exit boundaries, the corresponding time period in the peak-valley electricity price range label is marked as pending verification, and an electricity price data quality alarm is triggered. During the alarm period, the peak-valley electricity price range labels already generated are temporarily suspended from being sent to the matching stage pending data correction and confirmation.

[0030] Energy consumption characteristic distributions are matched with peak-valley electricity price interval labels to generate energy consumption deviation markers. Both excessively high energy consumption during high-price periods and excessively low energy consumption during low-price periods are optimizable deviation states. The energy consumption characteristic distribution and peak-valley electricity price interval labels describe the same time period from two dimensions: energy consumption behavior and electricity price status, respectively. Pairing them allows for the identification of the spatial distribution and magnitude of the two types of deviations. The average power of each time period in the energy consumption characteristic distribution is paired with the corresponding time period label in the peak-valley electricity price interval label period by period. A peak-hour label with an average power exceeding 1.1 times the daily average power is identified as a peak-hour overconsumption deviation, while a valley-hour label with an average power below 0.7 times the daily average power is identified as a valley-hour underconsumption deviation. For example, if the peak-hour label with an average power of 1.35 times the daily average power for a building indicates that the peak-hour energy consumption concentration is higher than reasonable. Pairing equipment-level statistics with peak-valley electricity price range labels in the energy consumption characteristic distribution can further identify the source of deviation for specific equipment. A higher average power output during peak hours and a lower average power output during off-peak hours for chiller units suggests that the potential for cooling load transfer is concentrated in chiller units. This equipment-level deviation location information is added to the extended field of the energy consumption deviation label for subsequent scheduling strategy reference. The pairing results for each time period are supplemented with deviation type and deviation magnitude to form an energy consumption deviation label. These labels are arranged in time period order, and the energy consumption deviation label items corresponding to low-confidence entries in the energy consumption characteristic distribution are simultaneously marked with a low-confidence flag.

[0031] An energy consumption status baseline is generated based on the energy consumption deviation markers. The deviation amplitude of each time period in the energy consumption deviation markers must be normalized to unify the dimensions. The normalization formula is φ=(Q_t-Q_avg) / Q_avg, where Q_t is the average power (kW) for that time period, Q_avg is the average power (kW) for the whole day, and φ is the dimensionless normalized deviation value. A positive value indicates overconsumption deviation, and a negative value indicates underconsumption deviation. The normalization operation eliminates the influence of differences in building size on the absolute value of the deviation. In the energy consumption deviation markers, the normalized value of peak-hour overconsumption deviation is positive, and the underconsumption deviation during off-peak hours is negative. The sum of the absolute values ​​of the two types of deviations reflects the total degree of peak-valley deviation in the current scheduling mode. For example, the sum of the normalized mean of peak-hour overconsumption (0.28) and the normalized mean of off-peak underconsumption (0.22) (0.50) indicates that there is a moderate degree of peak-valley mismatch in the daily energy consumption distribution. This total value is recorded as a global quality indicator of the energy consumption status baseline in the baseline header field. In the energy consumption deviation markers, segments with the same type of deviation across multiple consecutive time periods are identified as persistent deviation zones. Persistent overconsumption zones correspond to key reduction intervals for scheduling optimization, while persistent underconsumption zones correspond to priority scheduling intervals for replenishment compensation. The start and end times and average deviation magnitude of persistent deviation zones serve as key annotation items for the energy consumption status baseline. The energy consumption status baseline uses a time-period deviation normalized index sequence as its core data body, with values ​​for each time period arranged chronologically to cover the entire day. Low-confidence marker entries in the energy consumption deviation markers are assigned confidence weights to the corresponding time periods in the energy consumption status baseline. These weights adjust the reference strength of the baseline values ​​for that time period in subsequent peak-valley optimization analysis.

[0032] Step S120: Based on the energy consumption status baseline, perform peak-valley electricity price optimization analysis to generate an electricity price response strategy. Implement a price interval layering method to extract peak-valley electricity price time boundaries for the electricity price response strategy. Based on the peak-valley electricity price time boundaries, identify trend turning points in advance and establish peak-valley differentiated scheduling paths.

[0033] Specifically, peak-valley electricity pricing optimization analysis is conducted based on the energy consumption status baseline to generate electricity price response strategies. During typical summer workdays, the normalized deviation of the continuous over-consumption zone typically reaches 0.28-0.42, corresponding to a sustained actual operating power of chiller units exceeding the reasonable scheduling target during peak hours. This concentrated distribution of deviation directly triggers a focused review of this period in the peak-valley electricity pricing optimization analysis. The peak-valley electricity pricing optimization analysis uses the normalized deviation index sequence of the energy consumption status baseline as input. It pairs the deviation values ​​of each period in the sequence with the real-time electricity price sequence for each period. The pairing results reveal the degree of temporal overlap between the building's energy consumption peak and the electricity price peak. The proportion of the overlapping period to the total peak hour duration reflects the degree of peak-valley overlap in the current scheduling mode; a higher overlap ratio indicates a worse temporal match between the building's energy consumption peak and the electricity price peak. In the peak-valley electricity price optimization analysis, the target power reduction for the time period corresponding to the continuous over-consumption zone in the energy consumption baseline is calculated. The power reduction is converted into an absolute amount of kilowatts by multiplying the normalized deviation value of that time period by the daily average power, which is equal to the difference between the average power of that time period and the daily average power. The target charging power for the time period corresponding to the continuous under-charging zone is calculated. After the two types of target power are calculated, they are arranged in the order of time periods to form a daily power adjustment instruction sequence. The electricity price response strategy is based on the power adjustment instruction sequence. An electricity price sensitivity weight is added to the over-consumption deviation period. The weight of the over-consumption period when the real-time electricity price is higher than 1.2 times the daily average is increased to 1.5, ensuring that the electricity price response strategy has a more aggressive reduction target during high electricity price periods and a relatively conservative charging target during low electricity price under-charging periods. The power adjustment instructions for low-confidence periods in the energy consumption status baseline are given a confidence coefficient of 0.7 in the electricity price response strategy. This coefficient is converted into a reduction in the execution intensity of the scheduling instructions for the corresponding period at the strategy execution layer, preventing baseline errors caused by sensor anomalies from being transmitted to scheduling actions and causing unplanned large-scale adjustments to building HVAC equipment.

[0034] In some embodiments, the step of extracting peak-valley electricity price time boundaries by implementing electricity price interval stratification for the electricity price response strategy includes: performing amplitude deviation analysis on the electricity price response strategy to determine the electricity price fluctuation range; extracting electricity price change rate characteristics based on the electricity price fluctuation range to generate a dynamic electricity price stratification threshold; dividing the electricity price response strategy into stratified categories using the dynamic electricity price stratification threshold to form an electricity price level classification; and performing time period boundary calibration processing based on the electricity price level classification to extract peak-valley electricity price time boundaries.

[0035] Amplitude deviation analysis was performed on the electricity price response strategy to determine the magnitude of electricity price fluctuations. In a typical summer weekday at 6 PM, the target power adjustment range of the electricity price response strategy for a certain building abruptly changed from -80 kW to -220 kW. This nearly three-fold increase in the adjustment range per unit time during this period represents the segment with the most concentrated amplitude deviation in the entire electricity price response strategy and is a key extreme value location to capture in the electricity price fluctuation analysis. Amplitude deviation analysis calculated the difference in target power adjustment range between adjacent time periods within the electricity price response strategy segment by segment. The absolute value distribution of the difference sequence reflected the drastic changes in the amplitude of the electricity price response strategy over the entire daily time axis. Periods where the absolute value of the difference exceeded twice the standard deviation of the daily average were identified as high-fluctuation periods. Electricity price fluctuation is defined as the ratio of the maximum absolute value of the difference during high-fluctuation periods to the range of the daily target power adjustment range of the electricity price response strategy. The range is the difference between the maximum and minimum adjustment ranges throughout the day, and the fluctuation ratio ranges from 0 to 1. Under alternating winter and summer operating conditions, this ratio is typically 0.35-0.55, while under pure summer cooling conditions, due to concentrated load, the ratio is lower, approximately 0.25-0.40. An uncertainty coefficient of 0.7 is added to the amplitude deviation calculation results for low-confidence marked periods in the electricity price response strategy. This uncertainty coefficient reduces the contribution of the difference to the corresponding period when the electricity price fluctuation is finally summarized, preventing false amplitude jumps caused by abnormal sensor data collection from inflating the electricity price fluctuation assessment results. The electricity price fluctuation is calculated individually for each high-fluctuation period throughout the day, and a weighted average is taken to form a single quantitative indicator. The weight is based on the absolute value of the target power of the electricity price response strategy for each high-fluctuation period; periods with larger absolute power values ​​have a more significant impact on building operating costs and therefore have higher weights.

[0036] Dynamic electricity price stratification thresholds are generated by extracting electricity price change rate features based on electricity price fluctuation amplitude. After quantifying the global amplitude characteristics of the electricity price response strategy using electricity price fluctuation amplitude, this is input as a benchmark parameter into the rate feature extraction module. The rate feature is calculated based on the absolute value of the first-order difference of the target power adjustment amplitude sequence in the electricity price response strategy, with the difference window set to 15 minutes corresponding to 4 sampling points. The electricity price change rate feature sequence generally increases under conditions of large electricity price fluctuation amplitude. The mean rate feature of the building's summer peak electricity price response strategy is typically 1.6-2.2 times that of the winter heating condition. Using the same fixed stratification threshold for both conditions can lead to frequent misjudgments of boundaries in summer and missed judgments in winter. The introduction of dynamic electricity price stratification thresholds is precisely to correct this adaptability deficiency. The dynamic electricity price stratification threshold uses the electricity price fluctuation range as an adjustment factor. The base threshold is set as the 75th percentile value of the electricity price change rate characteristic sequence. A linear correction term for the electricity price fluctuation range is then added to this base threshold. The correction formula is Threshold_dynamic = P75 × (1 + α × V_fluctuation), where P75 is the 75th percentile of the rate characteristic sequence, V_fluctuation is the quantified value of the electricity price fluctuation range, and α is an adjustment coefficient with a value of 0.4. The dimensions of Threshold_dynamic are consistent with the electricity price change rate characteristic sequence, and the unit is kilowatts per minute. P75 is calculated using the accumulated electricity price change rate characteristic sequence up to the current calculation time as the data source. A 60-minute sliding time window (corresponding to 12 5-minute sampling points) is used for rolling updates, and the 75th percentile value of the sequence within the window is recalculated every 5 minutes. When the day's operation has run for less than 60 minutes, the actual accumulated data points are used for calculation; in this case, the window length is equal to the actual running time. α=0.4 is the engineering calibration coefficient, which is determined as follows: Collect the peak and valley electricity price sequence of the building for 30 natural days in history, and calculate the false trigger rate (the probability of misjudging non-boundary periods as boundaries) and the false omission rate (the probability of misjudging true boundary periods as non-boundaries) of the layer boundary with a step size of 0.05 in the interval α∈[0.1, 1.0]. Take the α value corresponding to the minimum sum of the two as the calibration result. For typical commercial buildings, the comprehensive false omission rate corresponding to α=0.4 is less than 5%, which can be directly adopted. For buildings with special load characteristics or electricity price structure, the α value can be re-determined based on the building's own historical data according to the above calibration process. After calculation, the dynamic electricity price tiering thresholds are output in three levels: high, medium, and low. The high threshold is Threshold_dynamic×1.5, corresponding to extreme electricity price fluctuation scenarios; the medium threshold is Threshold_dynamic, corresponding to normal fluctuation scenarios; and the low threshold is Threshold_dynamic×0.6, corresponding to minor electricity price adjustments. The three thresholds together constitute the decision boundary set for subsequent electricity price tier classification, covering the tiered needs of buildings under different electricity price fluctuation intensities throughout the year.

[0037] Electricity price response strategies are stratified using dynamic electricity price tiering thresholds to form electricity price level classifications. If the rate of change of electricity price during a certain period is 18 kW / min and the high-end dynamic electricity price tiering threshold is 15 kW / min, then that period exceeds the high-end threshold and is classified as Level 1 response. This corresponds to a scenario where building chillers and ice storage tanks need to initiate maximum power adjustments during that period. The electricity price level classification compares the rate of change of electricity price for each period in the electricity price response strategy with the three dynamic electricity price tiering thresholds one by one. Periods with a rate of change exceeding the high-end threshold are classified as Level 1; those between the medium and high-end thresholds are classified as Level 2; those between the low and medium-end thresholds are classified as Level 3; and those below the low-end threshold are classified as Level 4. This four-level classification covers all scenarios of electricity price response strategies, from extreme fluctuations to stable adjustments. The dynamic electricity price tiering threshold performs a lag processing on electricity price level classification within a buffer of ±5% near the boundary value. The time period level classification within the buffer period is based on the previous time period level. Level switching is triggered only when two consecutive sampling points meet the new level conditions. The lag processing eliminates level oscillations caused by frequent level jumps near the threshold boundary in the electricity price response strategy. After the electricity price level classification is completed for the entire day, the time period proportion of each level is calculated. Level 1 time periods typically account for 8%-15% of the entire day, concentrated in the rapid price increase periods before and after the morning and evening peak hours in buildings. Level 4 time periods typically account for 40%-55% of the entire day, mainly corresponding to the stable low electricity price periods during the late night off-peak hours. The distribution of the proportion of each level is a direct verification indicator to check whether the dynamic electricity price tiering threshold parameter settings are reasonable.

[0038] Based on electricity price level classification, time period boundary calibration is performed to extract peak-valley electricity price time period boundaries. The location of the jump between adjacent time period levels is the initial candidate point for time period boundary calibration. Jumps from level 1 to level 2 or above trigger strong boundary candidates, while jumps from level 2 to level 3 trigger weak boundary candidates. In building electricity price response strategies, strong boundary candidates are usually concentrated around 8:00 AM and 5:00 PM, which highly coincide with the peak start time of building HVAC load. The time period boundary calibration process performs window stability verification on both strong and weak candidate points. The verification window consists of 3 sampling points before and after the candidate point. The continuity coefficient of the same jump direction in the electricity price level classification sequence within the window is defined as the ratio of the number of sampling points within the window that satisfy the same jump direction to the total number of sampling points in the window. Candidate points with a continuity coefficient higher than 0.7 are confirmed as valid boundaries, while candidate points with a continuity coefficient lower than 0.7 are judged as false boundaries caused by transient disturbances and removed from the candidate set. After confirming the effective boundaries, peak-valley electricity price time periods are categorized into two types: peak-entry boundaries and peak-exit boundaries. Peak-entry boundaries correspond to the moment when the electricity price level changes from low to high, while peak-exit boundaries correspond to the moment when the price level changes from high to low. The timing accuracy of these two types of boundaries is based on a 5-minute sampling interval for the electricity price response strategy. Precise boundary times are calculated between adjacent sampling points using linear interpolation, with an interpolation accuracy better than 1 minute. After the peak-valley electricity price time period boundaries are calibrated throughout the day, an alternating sequence of peak-entry and peak-exit boundaries is formed. The time period enclosed by adjacent peak-entry and peak-exit boundaries in the sequence constitutes the core operating range for building peak-hour scheduling. Within the peak-hour range, the coordinated action commands for chiller units and ice storage tanks are triggered with the peak-valley electricity price time period boundaries as the timing anchor points.

[0039] In some embodiments, the step of identifying trend inflection points in advance based on the peak-valley electricity price time period boundary and establishing a peak-valley differentiated scheduling path includes: extracting electricity price change rate features based on the peak-valley electricity price time period boundary to generate a trend change sequence; using the trend change sequence to identify consecutive inflection points in the same direction to form a trend confirmation signal and a scheduling switch advance time; sorting the switching priority based on the scheduling switch advance time and the trend confirmation signal to generate a switching trigger condition; and integrating the switching trigger condition into a path to establish a peak-valley differentiated scheduling path.

[0040] The trend change sequence is generated by extracting the rate of change characteristics of electricity prices based on the peak-valley electricity price time-bound boundaries. The peak-valley electricity price time-bound boundaries divide the entire day into several alternating peak and valley intervals. The rate of change of electricity prices in the 30 minutes before the peak-entry boundary and the 20 minutes before the peak-exit boundary exhibit a monotonically accelerating characteristic distinct from other time periods in historical building data. The average rate of change before the peak is typically 2.1-3.4 times that of the normal period. This acceleration characteristic is the core basis for the trend change sequence to capture scheduling turning points. The real-time electricity price sequence within the preceding window corresponding to the peak-valley electricity price time-bound boundaries is calculated using a 5-minute step for first-order difference. The sign change of the difference sequence records the direction of price increase or decrease, and the absolute value of the amplitude records the strength of the rate of change of electricity prices. Together, these two aspects constitute the dual-attribute description of each element in the trend change sequence. The time coverage of the trend change sequence is set to 60 minutes before and 15 minutes after the peak-valley electricity price period boundary. The first 60-minute window provides sufficient observation length for identifying trend turning points, while the last 15-minute window is used to verify the trend continuity after the boundary is crossed. The full 75-minute coverage ensures that the trend change sequence has sufficient temporal description capability for the dynamic changes before and after the electricity price inflection point. The trend change sequence is generated independently at each peak-valley electricity price period boundary. For a typical workday with 3-5 peak-valley electricity price period boundaries throughout the day, 3-5 trend change sequences are generated. Each sequence is stored independently with the corresponding boundary time as the identifier. The differences in the characteristics of the trend change sequences at different boundaries reflect the dynamic differences in electricity prices during the three scheduling scenarios of morning peak, evening peak, and valley periods in the building.

[0041] For example, the step of using the trend change sequence to identify consecutive inflection points in the same direction to form a trend confirmation signal and an advance scheduling handover time includes: extracting adjacent inflection points with consecutive inflection points in the same direction based on the trend change sequence to generate inflection point candidate pairs; verifying the inflection point candidate pairs through directional consistency to form a trend confirmation signal; extracting the corresponding trigger time based on the trend confirmation signal to generate an initial handover time; and performing advance compensation processing on the initial handover time to determine the advance scheduling handover time.

[0042] Based on trend change sequences, adjacent inflection points with continuous same-direction changes are extracted to generate inflection point candidate pairs. In a building's difference sequence before the evening peak, four positive acceleration elements appeared consecutively between 17:05 and 17:35. Among them, the absolute values ​​of the two difference elements at 17:15 and 17:25 were 22 kW / min and 31 kW / min, respectively, with the latter increasing by 41% compared to the former. These two constitute a typical pair of continuous same-direction acceleration inflection points, which is a representative trigger scenario for inflection point candidate pair identification. The identification of continuous inflection points in the same direction is based on the first-order difference symbol sequence of the trend change sequence. The specific steps are as follows: Calculate the difference value d(i) = x(i+1)−x(i) for every two adjacent elements in the trend change sequence, and record the positive or negative sign of the difference value sign(d(i)) for each element, forming a symbol sequence of the same length as the trend change sequence; scan the symbol sequence bit by bit, and record the position index where each symbol flips (i.e., the time when sign(d(i)) and sign(d(i−1)) have opposite signs) as candidate inflection point positions; perform a consistency check on all difference symbols between two adjacent candidate inflection point positions. When all difference symbols in the interval have the same direction (all positive indicates a continuous upward accelerating trend, and all negative indicates a continuous downward accelerating trend), the two candidate positions are determined to constitute a valid pair of continuous inflection point candidates in the same direction, and proceed to the subsequent direction consistency verification and monotonicity evaluation; if the symbols alternate in the interval, the candidate pair is invalid, skipped, and the search continues for the next set of candidate positions. The extraction of inflection point candidate pairs begins with the flip position of the difference sign in the trend change sequence as the initial marker. For each flip position, the system slides backward to identify the next flip position in the same direction. If the difference element signs between two flip positions are all identical, the two flip positions are considered to form a candidate inflection point pair; otherwise, the system skips this step and continues searching. The spacing between candidate inflection point pairs in the trend change sequence is measured by the number of difference elements. Inflection point candidate pairs with a spacing of 1-2 elements correspond to scenarios of rapid electricity price increases, while those with a spacing of more than 5 elements correspond to scenarios of slow electricity price increases. Peak-valley electricity price switching in buildings typically corresponds to candidate inflection point candidate pairs with a spacing of 2-4 elements. Candidate pairs within this range are prioritized in subsequent directional consistency verification. Inflection point candidate pairs are generated segment by segment in the daily trend change sequence. Each trend change sequence generates 2-6 candidate inflection point pairs. The number of candidate inflection point pairs is positively correlated with the fluctuation frequency of the trend change sequence. In real-time electricity price scenarios with frequent price fluctuations, the number of candidate inflection point pairs is excessive and requires strict screening through directional consistency verification to control the false trigger rate.

[0043] Trend confirmation signals are generated by verifying the directional consistency of candidate inflection points. The pass condition for directional consistency verification is that the encoded symbols at both flip positions are either both positive or both negative. Candidate inflection points with opposite directions are identified as oscillating pseudo-inflection points. Oscillation pseudo-inflection points in building electricity price response strategies typically appear near the boundary between a trough and a flattening point. During this period, slight fluctuations in electricity prices cause the difference symbols to flip back and forth. Directional consistency verification excludes these invalid flips from the candidate inflection point pairs. Candidate inflection points that pass directional consistency verification are further examined for the monotonicity of the trend change sequence between the two flip positions. Monotonicity is defined as the ratio of the mean of the absolute values ​​of the difference elements within the interval to the mean of the entire trend change sequence. Candidate inflection point pairs with a ratio higher than 1.3 are identified as strong trend confirmation signals, those between 0.9 and 1.3 are identified as medium trend confirmation signals, and those below 0.9 are identified as weak trend confirmation signals. The type of trend confirmation signal is determined based on the direction encoding of the inflection point candidate pair. Positive inflection point candidate pairs generate peak-entry trend confirmation signals, corresponding to the start of peak-period reduction preparations for the building's HVAC system. Negative inflection point candidate pairs generate peak-exit trend confirmation signals, corresponding to the start of trough-period filling preparations. Weak trend confirmation signals do not trigger scheduling actions when they appear alone. They need to be combined with medium or strong trend confirmation signals in adjacent time periods to activate the switching process.

[0044] The initial switching time is generated by extracting the corresponding trigger time based on the trend confirmation signal. The trigger time of the trend confirmation signal is based on the timestamp of the second flip position in the candidate inflection point pair that has passed the direction consistency verification. This timestamp corresponds to the earliest credible time when the continuous unidirectional accelerating trend in the trend change sequence has been fully established. The trigger time of the peak trend confirmation signal corresponding to the morning peak of the building usually occurs between 7:40 and 8:05, and the trigger time corresponding to the evening peak usually occurs between 17:20 and 17:45. The initial switching time is directly assigned the trigger time of the trend confirmation signal. The initial switching time of the strong trend confirmation signal is directly written into the scheduling instruction queue to wait for advance compensation calculation. The initial switching time of the medium trend confirmation signal is appended with a stability check before writing. The check is based on whether the differential direction of the trend change sequence of the two sampling points after the trigger time continues in the same direction. If it continues in the same direction, the writing is confirmed; otherwise, the initial switching time is postponed to the earliest time when the stability check passes. When multiple trend confirmation signals are triggered within the same building scheduling cycle, the initial switching times of each signal are arranged in the order of triggering. Adjacent initial switching times with a time interval of less than 10 minutes are merged into one, and the earlier time is taken to ensure sufficient scheduling preparation time. The trend confirmation signal type corresponding to the merged initial switching time is determined by the higher intensity of the two signals.

[0045] Advance compensation is applied to the initial switching time to determine the advance switching time. It takes 8-12 minutes for a building chiller unit to output its rated cooling capacity after receiving the start-up command. If the initial switching time is less than 8 minutes from the peak threshold, the chiller unit cannot complete pre-cooling preparation before peak hours. Advance compensation is used to fill this gap between the equipment response delay and the switching sequence. The advance compensation is calculated based on the response delay of the corresponding equipment. The advance compensation for chiller units is taken as the upper limit of the response delay (12 minutes) plus a safety margin of 2 minutes, totaling 14 minutes. The advance compensation for ice storage tank melting and cooling is taken as 5 minutes plus a 1-minute margin, totaling 6 minutes. The advance compensation for terminal equipment adjustment is taken as 2 minutes. The difference in advance compensation for these three types of equipment forms a staggered start-up sequence in the advance switching time sequence. The advance switching time is calculated by subtracting the advance compensation amount of the corresponding device from the initial switching time. If the calculated result is earlier than the current scheduling decision time, it indicates that the device response delay exceeds the available lead time of the trend confirmation signal. In this case, the advance switching time is corrected to the current scheduling decision time, and a response delay alarm is set for the corresponding device, indicating that the device may not be able to complete the state switch before the peak-valley electricity price period boundary in the current scheduling. After the advance compensation processing is completed, the advance switching time is output in groups by device type, with the time accuracy of each group being 1 minute, consistent with the time resolution of the peak-valley differentiated scheduling path, ensuring that the advance switching time can be directly embedded into the device action orchestration framework of the peak-valley differentiated scheduling path for execution.

[0046] Switching trigger conditions are generated by prioritizing the switching lead time and trend confirmation signal. Multiple trigger candidates may exist simultaneously within the same building scheduling cycle. The construction of switching trigger conditions first involves conflict identification of these candidates. For example, the chiller pre-cooling lead time triggered by the peak trend confirmation signal overlaps with the ice storage tank filling lead time triggered by the off-peak trend confirmation signal on the timeline. Such overlapping scenarios account for approximately 25%-35% of the total daily scheduling switches during the building's summer dual-peak operating conditions. Switching trigger conditions need to resolve these overlapping conflicts through priority ranking. The switching priority ranking is based on a combined score of the strength of the trend confirmation signal and the remaining time between the scheduling switchover advance time and the peak-valley electricity price period boundary. The scoring formula is E_priority = S_signal × (1 + T_response / T_margin), where S_signal is the trend confirmation signal strength (0-1 normalized), T_margin is the remaining time (minutes), T_response is the response delay of the corresponding equipment (minutes), and E_priority is a dimensionless priority index. The scheduling switchover advance times in the switching trigger conditions are arranged in descending order of E_priority. Building chiller units, due to their longest response delay, typically receive the highest priority in the switching trigger conditions, followed by ice storage tanks. Terminal equipment adjustment commands have the lowest priority due to their rapid response, but they can still effectively intervene within the peak-valley electricity price period boundary after the first two types of equipment have completed their scheduling. After the priority sorting is completed, the switching trigger condition performs timing staggering processing on multiple scheduling switching advance moments within the overlapping time window. The staggering interval is set to 1.2 times the difference in response delay between adjacent devices to ensure that the scheduling actions of each device in the switching trigger condition do not interfere with each other in timing.

[0047] A peak-valley differentiated scheduling path is established by integrating the switching trigger conditions. The peak-valley differentiated scheduling path is a daily equipment action arrangement table with a time axis as its framework and the time-segmented operating parameters of each device as its content. Path integration expands the scheduling switching advance time and priority ranking results of each device in the switching trigger conditions along the time axis, aligning them with the peak-valley electricity price period boundary markers to form a complete sequence of equipment action nodes before and after each boundary. For example, the peak-valley differentiated scheduling path marks the start of chiller unit output reduction at 17:06, the start of ice storage tank melting and cooling at 17:12, and the increase of terminal equipment supply air temperature at 17:20. These three types of equipment action nodes are arranged sequentially according to the peak-shaving sequence in the switching trigger conditions, together constituting the path content for the evening peak reduction phase of that day. After path integration, the peak-valley differentiated scheduling path is divided into two sub-paths: peak-period reduction path and valley-period charging path. The peak-period reduction path starts with the peak-advancing trend confirmation signal and ends with the peak-receding boundary. The upper limit of building chiller output, ice melting and cooling rate of ice storage tanks, and adjustment range of terminal equipment are marked for each time period within the path. The valley-period charging path starts with the peak-receding trend confirmation signal and ends with the peak-advancing boundary of the next day. The charging rate of ice makers and the energy storage charging target are marked for each time period within the path. The total energy of the two sub-paths in the peak-valley differentiated scheduling path is balanced during the path integration phase. The difference between the cumulative reduction in the peak-period reduction path and the cumulative charging in the valley-period charging path does not exceed 10% of the rated capacity of the building energy storage device. If it exceeds this, the charging rate in the last period of the charging path is proportionally reduced until the balance condition is met, ensuring that the peak-valley differentiated scheduling path is actually feasible under the constraint of energy conservation.

[0048] Step S130: Perform collaborative response analysis on the peak-valley differentiated scheduling path to generate equipment collaborative parameters. Based on the equipment collaborative parameters, retain the frequency regulation response margin and start the energy storage and heat recovery system to generate peak scheduling execution records.

[0049] Specifically, collaborative response analysis is performed on peak-valley differentiated scheduling paths to generate equipment collaborative parameters. In the peak-period pressure reduction sub-path of the peak-valley differentiated scheduling path, the action nodes of the chiller units and ice storage tanks are closely intertwined in time. If the power adjustment commands of these two types of equipment are executed independently, it is highly likely to cause a brief overshoot in the chilled water supply temperature. Collaborative response analysis is precisely to identify the power linkage constraints between the equipment nodes in the peak-valley differentiated scheduling path. Collaborative response analysis calculates the pairwise correlation of the power adjustment commands of adjacent equipment action nodes in the peak-valley differentiated scheduling path. The linkage coefficient between chiller unit output reduction and chilled water pump speed reduction is typically between 0.65 and 0.82, and the linkage coefficient between chiller unit output adjustment and the air supply temperature response of the terminal air conditioning units is approximately 0.40 to 0.55. The equipment collaborative parameters use this linkage coefficient matrix as the core data body. Equipment pairs with linkage coefficients higher than 0.6 in the matrix are marked as strong collaborative constraint pairs in the equipment collaborative parameters. After identifying the strong collaborative constraint pairs in the peak-valley differentiated scheduling path, the equipment collaborative parameters establish constraint equations for the power balance conditions of each strong collaborative constraint pair. These constraint equations use the measured output of the master equipment as the independent variable and the upper limit of the available output of the slave equipment as the dependent variable. At the execution layer, the peak-valley differentiated scheduling path synchronously adjusts the output of the strong collaborative constraint pairs based on the constraint equations in the equipment collaborative parameters, ensuring that the hydraulic balance of the building's HVAC system is not disrupted by local equipment adjustments throughout the peak scheduling period. In the event of a single critical piece of equipment failing and exiting the building, the equipment collaborative parameters synchronously generate a backup collaborative scheme. This backup scheme is based on the recalculated linkage coefficient matrix after removing the faulty equipment. The power commands of the corresponding nodes in the peak-valley differentiated scheduling path are redistributed to the remaining online equipment according to the backup scheme.

[0050] In some embodiments, the step of generating peak-period scheduling execution records by reserving frequency regulation response margin based on the device coordination parameters includes: generating the available output distribution of each device based on the device coordination parameters; performing frequency regulation margin reservation calculations on the available output distribution of each device to obtain a margin reservation amount; performing graded startup of the energy storage and heat recovery system according to the margin reservation amount to determine the effective output configuration; and generating peak-period scheduling execution records by recording the execution status based on the effective output configuration.

[0051] The available output distribution of each device is generated based on the device coordination parameters. The linkage coefficient between the building's ice storage tank and the chiller unit is 0.78 in the device coordination parameters, while the linkage coefficient with the terminal fan coil units is only 0.32. This difference in linkage strength means that the upper limit of the ice storage tank's available output cannot be independently determined from the current operating status of the chiller unit; it must be calculated jointly with the real-time output of the chiller unit as a constraint. The generation of the available output distribution prioritizes strong coordination constraint pairs in the device coordination parameters. The current measured output of the main device in a strong coordination constraint pair is read in real-time from the building automation system. The upper limit of the available output of the device is determined by multiplying the measured output of the main device by the corresponding linkage coefficient in the device coordination parameters. For weakly coordinated devices, the upper limit of the available output is directly taken as a conservative estimate of 85% of the rated output. The available output distribution of each device is indexed by its device number, recording three values ​​for each device: the maximum available output, the current measured output, and the remaining adjustable margin. Under typical summer peak operating conditions in a building, the remaining adjustable margin for chiller units is approximately 12%-22% of their rated power; the adjustable margin for ice storage tanks is approximately 30%-50% of their rated ice melting rate; and the adjustable margin for heat recovery devices is approximately 25%-40% of their rated heat exchange power. The sum of the adjustable margins of these three types of equipment constitutes the maximum reducible power reserve for this peak-period scheduling. When multiple similar devices are operating in parallel in a building, the adjustable margins of similar devices are recorded independently in the distribution rather than being combined in the statistics. This avoids masking the problem of insufficient margins for individual devices by combining similar devices in the calculation.

[0052] The frequency regulation margin is calculated for the available output distribution of each device to obtain the margin reserve. After the remaining adjustable margin of the chiller unit is transferred from the available output distribution of each device to the frequency regulation margin reserve module, the module calculates the frequency regulation response share that must be reserved in the adjustable margin of each device, based on the lower limit of the response power specified in the frequency regulation service agreement signed between the building and the power grid. The reserve margin R_reserve = P_available × r_freq × k_response, where P_available is the remaining adjustable capacity (kW) of the equipment in the available output distribution, and r_freq is the frequency regulation margin ratio coefficient (valued between 0.15 and 0.30, set according to the building and grid agreement terms). The specific method for determining r_freq is as follows: Read the minimum response power commitment value P_agreement (kW) from the frequency regulation service agreement signed between the building and the grid. Combine this with the sum of the current adjustable capacity of each participating frequency regulation equipment to obtain P_available_total (kW). Calculate r_freq using the formula r_freq = (P_agreement / P_available_total) × k_safety, where k_safety is a safety margin coefficient ranging from 1.1 to 1.2, used to compensate for insufficient response that may be caused by equipment output measurement errors and command transmission delays. For example, if the agreement specifies a minimum response power of 50 kW and the current total adjustable capacity of the equipment is 300 kW, then r_freq = (50 / 300)×1.15 ≈0.19, falling within the range of 0.15 to 0.30. When P_available_total changes due to equipment commissioning / decommissioning, seasonal operating conditions, or equipment maintenance, r_freq is dynamically recalculated at the start of each scheduling cycle to ensure that the frequency regulation commitment can be met under any operating condition; k_response is the equipment response delay correction coefficient (the faster the response rate, the smaller k_response; k_response for chiller units is approximately 1.2, and for ice storage tanks it is approximately 0.85), and R_reserve is in kilowatts, with the same dimensions as P_available. When the remaining adjustable margin of the building chiller units in the distribution of available output of each equipment is 180 kilowatts, r_freq is 0.20, and k_response is 1.2, the margin reserve is 43.2 kilowatts. This value is subtracted from the adjustable margin of the chiller units to obtain the upper limit of the net output that the equipment can actually use for peak period reduction. The margin reserve is calculated for each piece of equipment in the building and then summarized to form an equipment-level margin reserve list. In the list, the ice storage tank has a lower margin reserve than the chiller unit under the same adjustable margin conditions because of its fast response rate and small k_response. This reflects the efficiency advantage of fast response equipment in frequency regulation margin reservation.After all equipment calculations are completed, the margin reserve list is checked for total quantity. The sum of the margin reserves of each equipment must not be less than the minimum response power commitment value specified in the building frequency regulation service agreement. If it is insufficient, the margin ratio coefficient r_freq of each equipment is increased proportionally until the total quantity meets the requirements.

[0053] The effective output configuration is determined by tiered startup of the energy storage and heat recovery systems based on the margin reserve. When the margin reserve of the ice storage tank is 28 kW and the net available output is 65% of the rated ice melting rate, and the margin reserve of the heat recovery unit is 15 kW and the net available output is 70% of the rated heat exchange power, the difference between the sum of the net available output of the two types of equipment and the target reduction amount of the peak-hour reduction sub-path for the current period determines whether further reduction of the chiller unit output is required. If the difference is positive, the energy storage and heat recovery systems can respond independently; if the difference is negative, the chiller unit will be triggered to intervene in a coordinated manner. The tiered startup strategy fixes the startup order of the energy storage and heat recovery systems as follows: heat recovery unit first, ice storage tank second, and chiller unit output reduction last. The reason for prioritizing the startup of the heat recovery unit is that its startup response time is the shortest, approximately 2-3 minutes, and it does not consume ice storage capacity, making it the most economical option within the peak-hour pre-peak window. The margin reserve imposes constraints on the upper limit of output at each stage of the energy storage and heat recovery system. After each stage of equipment is started, the sum of the cumulative loaded output of the equipment and the margin reserve is calculated in real time to see if it reaches the rated output limit of the equipment. If it does, further loading is prohibited until the grid frequency regulation response request is released, at which point the margin constraint can be released and the full usable output can be restored. The effective output configuration is determined after each stage of equipment has been started and the margin constraint is satisfied. The configuration content is a list of the actual loaded output of the currently started equipment. The output values ​​of each equipment in the list are based on the measured feedback values ​​from the building automation system, not the dispatch command values. Equipment whose measured output deviates from the command by more than 5% of the rated output is marked as execution deviation equipment in the effective output configuration. The measured output of execution deviation equipment replaces the command value in the subsequent peak period dispatch execution record summation of actual reduction.

[0054] Peak-period scheduling execution records are generated based on the effective output configuration and execution status recording. The measured output of each device continuously changes with the operating status of the building's HVAC system. The peak-period scheduling execution record performs snapshot-style sampling of the effective output configuration at a 1-minute time granularity. Each minute snapshot records three items: the measured output of each device, the consumption status of the margin reserve, and the device's operating status flags. The snapshot sequence for the complete peak period constitutes the time-series data body of the peak-period scheduling execution record. The measured output of devices with execution deviations in the effective output configuration is separately calculated in the peak-period scheduling execution record for actual pressure reduction contribution. The contribution is calculated by accumulating the difference between the measured output of the device with execution deviation and the scheduling command value. When the accumulated deviation exceeds 10% of the device's rated power, an abnormal alarm is triggered, and the alarm information is written to the abnormal event field of the peak-period scheduling execution record. The peak-period scheduling execution record calculates the actual total energy reduction for each time period after the peak-period energy reduction sub-path ends. The actual total energy reduction is calculated by multiplying the algebraic sum of the differences between the measured output of each device in the effective output configuration and its pre-peak baseline output by the duration of the statistical period (in hours). The baseline output is the average measured output of the most recent 10 minutes before the peak period starts. The unit of the actual total energy reduction is kilowatt-hours. The time-by-time comparison between the actual total energy reduction in the peak-period scheduling execution record and the target energy reduction of the peak-valley differentiated scheduling path is appended to the end of the record. A positive difference indicates that the actual energy reduction exceeds the target, and the building has an over-response risk. A negative difference indicates that the actual energy reduction is insufficient, and the output loading ratio of the energy storage and heat recovery systems needs to be increased in the next scheduling cycle. The system has a three-level feedback response mechanism for execution deviations. When the actual reduction in energy consumption deviates from the planned reduction by more than 10% during a certain period, the following response is triggered: Level 1 (10%-20%): During the remaining period of the peak period, the output commands of each device are increased proportionally according to the deviation, and the effective output configuration is corrected in real time; Level 2 (20%-35%): Based on the Level 1 correction, a correction signal is simultaneously sent to the valley period charging sequence scheme, and the corresponding energy storage medium charging target is reduced according to the deviation; Level 3 (more than 35%): The backup coordination scheme in the equipment coordination parameters is activated, and a cross-cycle correction signal is sent to the coordination optimization configuration, reducing the target reduction for the same period the next day to 90% of the actual reduction in this period, realizing rolling learning and correction.

[0055] Step S140: Based on the actual reduction in the peak period scheduling execution record, the valley period charging amount is deduced to generate a valley period charging time sequence scheme. The valley period charging time sequence scheme is used to start ice making and energy storage charging to generate valley period energy storage parameters. The valley period energy storage parameters are used to perform dynamic peak-valley energy allocation and compensation to build a collaborative optimization configuration.

[0056] In some embodiments, the step of generating a valley-period charging sequence scheme by reverse-engineering the valley-period charging amount based on the actual reduction amount in the peak-period scheduling execution record includes: extracting the actual reduction amount of each device from the peak-period scheduling execution record to generate a reduction amount summary; determining the total valley-period charging amount by reverse-engineering the valley-period charging target based on the reduction amount summary; allocating the total valley-period charging amount according to the charging amount of each time period in the valley period to generate a time period charging allocation table; and generating a valley-period charging sequence scheme by performing time period priority sorting based on the time period charging allocation table.

[0057] The actual reduction in power consumption for each device was extracted from the peak-period scheduling execution records to generate a summary of the reduction. The peak-period scheduling execution records show that the actual output of the building's ice storage tank during the three-hour peak period from 6 PM to 9 PM was 95 kWh, while the actual heat exchange contribution from the heat recovery unit was 42 kWh. Together, these two accounted for 82% of the total actual reduction in power consumption during this peak period. This high proportion reflects that the energy storage equipment undertook the main reduction task during this peak period, and the demand for charging during the off-peak period was concentrated in ice storage and hot water storage media. The actual reduction in power consumption for each device was extracted by integrating the difference between the measured output and the baseline output of each device in the time-series data of the peak-period scheduling execution records at the minute level. The integration result is in kWh. The start and end times of the peak period were read from the peak-to-valley differentiated scheduling path's entry and exit points, and the integration interval was strictly aligned with the actual duration of the peak period. The pressure reduction summary groups and accumulates the actual pressure reduction of each device by device type. Ice storage tanks are summarized as refrigeration-side pressure reduction, heat recovery devices as heat recovery-side pressure reduction, and chiller unit output reduction as main unit-side pressure reduction. The sum of the pressure reductions in these three groups is then checked for consistency with the peak-period actual pressure reduction total in the header field of the peak-period scheduling execution record. If the difference exceeds 1% of the total, it indicates that there is a misrecording of device metering, triggering a data integrity review process for each device and each time period. After the pressure reduction summary passes the verification, a peak-period execution quality score is added. The score is defined as the weighted average of the ratio of the actual pressure reduction of each device to the target pressure reduction of the corresponding device in the peak-valley differentiated scheduling path. The weight is based on the rated power of each device. A score higher than 0.9 is considered excellent execution, 0.7-0.9 is considered normal execution, and a score lower than 0.7 is marked as low-quality execution in the pressure reduction summary. The low-quality execution mark is used in the collaborative optimization configuration phase to lower the target pressure reduction of the corresponding device for the next day.

[0058] For example, the step of back-calculating the valley period filling target based on the sum of the reduction amounts to determine the total filling amount during the valley period includes: performing peak-valley energy balance verification on the sum of the reduction amounts to generate an energy gap distribution; identifying key compensation periods based on the energy gap distribution to generate a compensation priority sequence; using the compensation priority sequence to perform segmented filling target calibration on the energy gap distribution to generate segmented filling amounts; and summing the segmented filling amounts to determine the total filling amount during the valley period.

[0059] The energy deficit distribution was obtained by performing peak-valley energy balance verification on the sum of the energy reduction. In the sum of the energy reduction, the peak consumption of the building ice storage tank was 95 kWh on the cooling side and the heat recovery consumption of the hot water energy storage device was 42 kWh. However, the measured remaining energy of the two types of energy storage devices at the beginning of the valley period was 28% and 35% of their rated capacity, respectively. The gap from the preset target water level of 75% determines that the energy deficit to be supplemented during this valley period is much larger than the peak consumption itself. Peak-valley energy balance verification uses the energy corresponding to the preset target water level of each energy storage device minus the currently measured remaining energy as the basic gap. This gap is then supplemented by the peak consumption of the corresponding energy storage medium in the pressure reduction summary to obtain the comprehensive gap. The ratio of the comprehensive gap to the rated capacity of the energy storage device is defined as the gap rate. When the gap rate exceeds 1.0, it indicates that the current capacity of the energy storage device is insufficient to simultaneously replenish peak consumption and restore the target water level within a single valley period. In this case, the comprehensive gap is capped at the rated capacity of the energy storage device minus the currently measured remaining energy. The excess portion is replenished during the next valley period. Before the capping, the gap rate on the ice-making side is typically between 0.35 and 0.60, and the gap rate on the heat recovery side is typically between 0.25 and 0.45. The energy gap distribution uses each period of the valley period as the time axis, distributing the comprehensive gap according to the maximum chargeable power of each period over the entire valley period. The maximum chargeable power is constrained by both the rated ice-making rate of the ice-making system and the nighttime available capacity of the building transformer. The smaller of these two constraints determines the maximum chargeable power for each period. The four hours from 1 a.m. to 5 a.m. typically cover 55%-70% of the total energy gap distribution. The time-period density difference in the energy gap distribution directly drives the priority direction for identifying subsequent key compensation periods.

[0060] A compensation priority sequence is generated based on the identification of critical compensation periods according to the energy gap distribution. Periods where the gap density exceeds 1.4 times the average density during the entire off-peak period are identified as critical compensation periods. Under typical off-peak conditions for buildings, critical compensation periods are usually concentrated in two intervals: 1:00 AM to 5:00 AM and 6:00 AM to 7:00 AM the following day. The former corresponds to the deep valley low electricity price window, and the latter corresponds to the last high-intensity charging opportunity for the ice-making system before the peak period. Critical compensation period identification is performed using a sliding window method on the energy gap distribution. The window width is set to 1 hour, corresponding to 2 periods. Consecutive periods with an average gap density exceeding a threshold within the window are merged into one critical compensation interval. The number of merged critical compensation intervals is 2-4 under typical off-peak conditions for buildings. The compensation priority sequence is sorted according to the reciprocal of the product of the average gap density of the critical compensation interval and the corresponding electricity price. The larger the reciprocal of the product, the lower the unit compensation cost and the higher the urgency of the gap. The deep valley period in the early morning usually receives the highest compensation priority due to the lowest electricity price, while the 6:00 AM to 7:00 AM period receives the second highest priority due to its proximity to the peak period, ensuring sufficient energy storage levels before the peak period. After all critical compensation intervals are sorted, the compensation priority sequence is organized using four pieces of information: interval number, time period range, average gap density, and priority score. Adjacent intervals with priority scores differing by less than 5% are merged to reduce the number of filling switches and lower equipment wear caused by frequent start-ups and shutdowns of the ice-making system.

[0061] The energy gap distribution is segmented and calibrated using a compensation priority sequence to generate segmented filling amounts. The highest priority valley interval in the compensation priority sequence is the early morning valley interval, which is calibrated first. The segmented filling amount for this interval is capped at the cumulative energy gap distribution within that interval. Simultaneously, it is constrained by the physical constraint that the icing system requires a 15-minute maintenance interval after more than 4 hours of continuous operation. This constraint shortens the actual filling time for the valley interval to 3 hours and 45 minutes, corresponding to a segmented filling amount approximately equal to the rated icing rate multiplied by 225 minutes. The segmented filling target calibration is performed sequentially for each key compensation interval in the compensation priority sequence according to priority. The segmented filling amount for each interval is determined by the smaller of the cumulative energy gap distribution within that interval and the physical constraint's maximum filling limit. After calibration, the calibrated segmented filling amount is deducted from the total energy gap distribution. The remaining gap amount is then calibrated for the next priority interval until the total energy gap distribution is fully allocated or all key compensation intervals are calibrated. After the segmented charging amount is calibrated in each critical compensation interval, the upper limit of the energy storage device capacity is checked. The cumulative value of the segmented charging amount on the ice-making side and the heat recovery side shall not exceed the difference between the rated capacity of the corresponding energy storage device and the current measured remaining energy. If it exceeds the limit, the segmented charging amount of the lower priority interval shall be reduced in reverse order of priority until the capacity constraint is met.

[0062] The total off-peak charging volume is determined by summing the segmented charging volumes. The calibration values ​​of each key compensation interval on the ice-making side are accumulated one by one, while the calibration values ​​of each interval on the heat recovery side are accumulated simultaneously. The accumulated results of the two types of energy storage media form the total off-peak charging volume for the ice-making side and the total off-peak charging volume for the heat recovery side, respectively. The sum of these two constitutes the total charging target for the building's off-peak energy storage system. After the summation is calculated, a closure check is performed against the total energy gap distribution. The check difference is the difference between the sum of the segmented charging volumes and the total energy gap distribution. A positive difference indicates that the calibrated charging volume exceeds the actual gap, and the excess is proportionally reduced from the lower priority intervals. A negative difference indicates that there is a residual gap not covered by the key compensation periods. The residual gap is supplemented by continuous low-power charging during normal periods, with the supplementary power set at 20% of the ice-making system's rated power to reduce charging costs during normal periods. After the total charging volume during the off-peak period is determined on both the ice-making and heat recovery sides, it is written into the total amount field of the off-peak charging sequence plan. The total amount field also records the estimated total electricity cost of the charging during this off-peak period. The estimated total electricity cost is calculated by multiplying the charging volume in each key compensation interval by the corresponding electricity price and then summing the results. After the collaborative optimization configuration is established, this value is compared with the electricity cost saved by peak period reduction. The difference between the two is the estimated net economic benefit of this peak-valley arbitrage.

[0063] The total charging capacity during off-peak periods is allocated according to the different time periods within the off-peak period to generate a time-based charging allocation table. The total charging capacity during off-peak periods is weighted by the real-time electricity price level for each time period. The maximum charging power is allocated to the deepest off-peak period (typically from 1:00 AM to 5:00 AM), during which the building's ice-making system operates at 90% of its rated ice-making rate, achieving 60%-70% of the ice-making target within 4 hours. The time-based charging allocation table is constructed with the upper limit of available charging power for each time period during off-peak periods as a constraint. The upper limit of available charging power is the smaller value between the rated charging rate of each energy storage device and the nighttime available capacity of the building's transformer. During the late night, the load of other electrical equipment in the building decreases, so the available transformer capacity is usually 30%-45% higher than during the day, providing sufficient power capacity margin for the ice-making system. The time-slot charging allocation table covers the entire off-peak period in 30-minute intervals. A typical off-peak period spans 16 time slots from 11 PM to 7 AM the next day. The charging power command for each time slot in the allocation table is expressed in kilowatts. The total integral of the charging power across the 16 time slots must equal the total charging amount during the off-peak period. If the integral deviation exceeds 1%, the power of the last time slot is proportionally corrected until the total constraint is met. After the power command for each time slot is determined, the time-slot charging allocation table is supplemented with a predicted state of charge sequence for the energy storage device. The sequence is calculated by accumulating the charging power integrally from the initial remaining energy. The predicted state of charge must not exceed 95% of the rated capacity of the corresponding energy storage device at any time. If it does, the power command for that time slot is truncated to the maximum value that meets the constraint, and the remaining charging amount is carried over to the next time slot.

[0064] Based on the time-based charging allocation table, a time-based priority ranking is performed to generate a peak-period charging sequence plan. During the priority ranking phase, the charging power commands for the 16 time periods in the time-based charging allocation table are re-evaluated according to the real-time electricity price level of each time period. The time period with the lowest electricity price receives the highest priority, ensuring that power commands for low-cost charging periods are prioritized for execution when equipment output is limited. The time-based priority ranking is based on the reciprocal of the product of the charging power command and the corresponding time period's electricity price. The smaller the product, the lower the unit charging cost and the higher the priority. After ranking, the top four time periods with the highest priority are marked as core charging periods. Charging power commands during core charging periods have priority recovery rights in the event of a temporary fault in the energy storage device. Power commands for other time periods are executed sequentially according to priority after the fault is recovered. After prioritizing time periods, the off-peak charging sequence plan integrates the time period charging allocation table and priority sequence to form a complete scheduling plan with time as the horizontal axis and charging power commands and priority labels as the vertical axis. In the plan, the equipment start time of the core charging period is 5 minutes earlier than the planned time for preheating. The preheating command is written into the equipment preparation field of the off-peak charging sequence plan to ensure that the ice-making system is in the rated ice-making state at the beginning of the core charging period rather than waiting for preheating delay. After the off-peak charging sequence plan is generated, a total closure check is performed. The check is based on whether the difference between the expected total charging amount of the off-peak charging sequence plan and the integral sum of the charging power commands of each time period in the time period charging allocation table is less than 5 kWh. After the check passes, the off-peak charging sequence plan is officially issued to the control layer of the ice-making system and energy storage device for execution, and the generation process of off-peak energy storage parameters is triggered simultaneously.

[0065] The off-peak charging sequence scheme is used to initiate ice making and energy storage charging, generating off-peak energy storage parameters. After the off-peak charging sequence scheme is issued to the control layer of the building's ice making system and hot water energy storage device, the two types of energy storage devices are started sequentially according to the charging power instructions for each time period in the scheme. The ice making system operates at 90% of the rated ice making rate during the core charging period, while the hot water energy storage device charges in parallel at 75% of the rated heat exchange power during the same period. The synchronous start-up of the two types of devices ensures that the building's off-peak charging power is maintained in the range of 80%-88% of the transformer's nighttime available capacity during the deep off-peak period, making full use of the low electricity price window. The off-peak energy storage parameters are generated in real time during the charging process of the ice making system and the hot water energy storage device, with the current state of charge of each energy storage medium as the core indicator. The state of charge on the ice making side is defined as the ratio of the current ice storage amount in the ice storage tank to the rated ice storage capacity, and on the heat recovery side, it is defined as the ratio of the current heat storage amount in the hot water tank to the rated heat storage capacity. Both types of state of charge are updated and written into the off-peak energy storage parameters every 5 minutes. When there is a deviation between the preset charging power command for each time period in the off-peak charging sequence plan and the actual ice-making rate of the ice-making system, the difference between the measured charging amount and the planned charging amount for the corresponding time period in the off-peak energy storage parameters is recorded as the charging deviation. When the cumulative charging deviation exceeds 5% of the total charging amount during the off-peak period, the timing plan is dynamically corrected. The correction is based on the available charging power for the remaining off-peak periods, ensuring that the final state of charge in the off-peak energy storage parameters reaches the preset target level at the end of the off-peak period. The off-peak energy storage parameters record the state of charge timing curves of the ice-making side and the heat recovery side throughout the off-peak period. The slope of the curve reflects the actual effect of the off-peak charging sequence plan at each time period. Periods with a slope more than 15% lower than the planned value correspond to insufficient output of the ice-making system, and the off-peak energy storage parameters add an inefficient charging mark for that period.

[0066] A collaborative optimization configuration is constructed by using off-peak energy storage parameters for dynamic peak-valley energy allocation and compensation. The dynamic peak-valley energy allocation and compensation uses the measured state of charge (SOC) of the ice-making and heat recovery sides in the off-peak energy storage parameters as the starting point, and pairs it with the summaries of the reduction amounts of each device in the peak-peak scheduling execution records. The pairing results form an energy balance sheet for the building's energy storage system in this peak-valley cycle. The income item in the balance sheet is the actual total charge recorded in the off-peak energy storage parameters, and the expenditure item is the actual total consumption of the energy storage side in the summaries of reduction amounts. The collaborative optimization configuration implements three types of compensation and allocation based on the balance sheet: First, for ice-making systems with inefficient charging periods in the off-peak energy storage parameters, the charging power limit for that period is increased in the next day's off-peak plan to compensate for the insufficient charging. Second, for energy storage devices whose state of charge exceeds the preset target level in the off-peak energy storage parameters, the energy release priority of such energy storage media is appropriately increased in the next day's peak-peak reduction plan to absorb the overcharged capacity. Third, for cycles where there is still a net deficit after the dynamic allocation and compensation of peak and off-peak energy, a reduction target correction suggestion is fed back to the peak-valley differentiated scheduling path, and the peak-peak reduction target for the next day is lowered to within the actual replenishable amount for the current off-peak period. After the three types of compensation and allocation are completed, the collaborative optimization configuration integrates the state of charge time series curve of the off-peak energy storage parameters with the corrected next-day scheduling parameters to form a complete configuration file covering three dimensions: the operating benchmark of the cold source side, the energy storage charging and discharging plan, and the equipment collaborative constraints. The configuration file is organized with building equipment number and scheduling time as dual indexes, supporting the scheduling execution layer to directly retrieve the corresponding operating parameter instructions by equipment and time.

[0067] Step S150: Based on the collaborative optimization configuration, power grid load monitoring is performed to generate a load distribution curve. The load distribution curve is then subjected to multi-objective optimization of electricity cost and load stability to determine the optimal scheduling time. At the optimal scheduling time, the building and power grid collaborative control is executed to output collaborative optimization control commands.

[0068] Specifically, load distribution curves are generated based on grid load monitoring using collaborative optimization configuration. Grid load monitoring is initiated synchronously after the collaborative optimization configuration is issued. The monitoring targets include the real-time active power at the building transformer outlet and the load of the distribution network feeders connected to the building. The sampling frequency for both types of monitoring data is 1 minute, and the timestamp accuracy is consistent with the time period granularity of the collaborative optimization configuration. The energy storage charging and discharging plan in the collaborative optimization configuration has the most significant impact on the building's total electricity load. During peak hours, the ice-making system operates at full load, causing the active power at the building transformer outlet to decrease by 20%-35% compared to daytime, while simultaneously increasing ice-making electricity consumption by approximately 15%-25%. The net load change resulting from the superposition of these two effects forms a characteristic load peak on the load distribution curve during peak hours. The magnitude of this load peak directly corresponds to the execution status of the ice-making side charging power command in the energy storage parameters during the valley period. The load distribution curve uses the real-time active power sequence at the transformer outlet as the main axis, superimposed with the load sequence of the distribution network feeders, to form a comprehensive description of the impact of building electricity consumption behavior on the power grid. The deviation between the expected load and the measured load in each time period of the collaborative optimization configuration is marked with shaded areas. Periods where the absolute value of the deviation exceeds 10% of the expected load are marked with an anomaly flag in the load distribution curve. The anomaly flagged periods correspond to the visual location of equipment execution deviations in the collaborative optimization configuration. The time coverage of the load distribution curve is set to the entire day within the validity period of the collaborative optimization configuration. The curve uses a minimum display granularity of 5 minutes to support fine-grained analysis of key scheduling periods. During the optimal scheduling time determination phase, local periods of the load distribution curve can be further refined to a 1-minute granularity to improve the accuracy of optimization time calibration. The mathematical model for multi-objective optimization of electricity cost and load stability is as follows. The objective function uses a weighted summation method: min F = w_c × C_total + w_s × σ_L, where C_total = Σ(P_t × λ_t × Δt) is the total electricity cost of the building within the prediction window (yuan), P_t is the active power at time t (kilowatts), λ_t is the corresponding real-time electricity price (yuan / kWh), Δt = 1 / 60 hours; σ_L is the standard deviation of active power within the stabilization prediction time window (kilowatts); w_c = 0.6, w_s = 0.4. Constraints include: (a) chilled water supply temperature 6℃ to 12℃; (b) output of each device within 20% to 100% of rated power; (c) state of charge (SOC) of energy storage ∈ [10%, 95%]; (d) the charging and discharging amount in a single time period does not exceed the rated charging and discharging power multiplied by the time period duration. The solution employs rolling time-domain optimization with a step size of 15 minutes and a prediction window of 60 minutes. Within each step, the linear programming subproblem is solved after linearization at the current working point. Only the optimal solution of the first step is executed before rolling to the next step.

[0069] In some embodiments, the step of performing multi-objective optimization of electricity cost and load stability on the load distribution curve to determine the optimal scheduling time includes: performing time-series differential decomposition on the load distribution curve to generate a load fluctuation decreasing sequence; locating the rapid convergence interval of fluctuations from the load fluctuation decreasing sequence to generate a stabilization prediction time window; estimating the electricity cost of each time point within the stabilization prediction time window and the load distribution curve to generate a cost stability comprehensive score; and determining the optimal scheduling time based on the cost stability comprehensive score.

[0070] A time-series differential decomposition (TSD) is performed on the load distribution curve to generate a load fluctuation decrease sequence. During the building's morning peak startup period, the absolute value of the 1-minute difference in active power on the load distribution curve typically reaches 80-150 kW / min, while after the ice-making system enters steady-state charging, the absolute value of the difference drops back to 5-15 kW / min. The difference amplitude between these two typical periods differs by an order of magnitude. TSD is precisely for extracting a quantitative description of this dynamic convergence process from the load distribution curve. TSD uses the 1-minute granular power sequence of the load distribution curve as input. First-order difference calculations are performed on the sequence to obtain the minute-by-minute power change. The absolute values ​​of the difference results are then arranged chronologically to form the original fluctuation sequence. The original fluctuation sequence is smoothed using a 5-minute moving average to eliminate single-minute acquisition noise, resulting in the load fluctuation decrease sequence. The load fluctuation reduction sequence exhibits a step-like decline during the execution period of the peak-period load reduction sub-path in the collaborative optimization configuration. After each adjustment of the energy storage equipment output, the sequence value briefly rebounds before continuing to decline. The slope of the step-like decline reflects the speed at which the building HVAC system gradually stabilizes under the drive of collaborative optimization configuration. The larger the absolute value of the slope, the more coordinated the execution is due to the strong collaborative constraints in the equipment collaborative parameters. During the deep trough period, the load fluctuation reduction sequence value typically remains at a low plateau of 3-8 kW / min. This low plateau corresponds to the extremely stable state of the load supplied by the ice-making system in steady-state charging. This plateau range serves as a reference benchmark for the completion of sequence convergence in the subsequent identification of the stabilization prediction time window.

[0071] The stabilization prediction time window is generated by locating the rapid convergence interval of the load fluctuation decrease sequence. The continuous decline in load fluctuation decrease sequence values ​​from a high level (above 50 kW / min) to a low plateau (below 10 kW / min) typically lasts 8-18 minutes. This decline segment appears as a monotonically decreasing continuous segment with a continuously negative slope in the load fluctuation decrease sequence. The starting point of the rapid convergence interval is identified when the sequence value exceeds 1.5 times the average of the entire segment, and the ending point is identified when the sequence value is below 0.6 times the average of the entire segment for three consecutive sampling points. The stabilization prediction time window starts at the end of the rapid convergence interval and extends 15 minutes beyond that point. This 15-minute extension covers the typical time required for the chilled water temperature to rebalance under the new steady-state output of the building's chiller units, ensuring that the load distribution curve values ​​within the stabilization prediction time window fully reflect the stable results of the current execution state of the collaborative optimization configuration. The stabilization prediction time window typically identifies 3-6 windows on the daily time axis of the load fluctuation decrease sequence, corresponding to typical operating conditions such as the steady state after the building starts in the morning peak, the steady state during the midday low, the steady state after the building starts in the evening peak, and the steady state during the deep valley ice making. The start and end times of each window are recorded with absolute timestamps and aligned with the time period boundaries of the collaborative optimization configuration, with an alignment accuracy better than 1 minute.

[0072] The cost stability comprehensive score is generated by estimating electricity costs at each moment within the stabilization prediction time window and comparing them with the load distribution curve. If, at a certain moment within the stabilization prediction time window, the measured active power of a building is 380 kW and the corresponding real-time electricity price is 0.92 yuan per kilowatt-hour, the minute-level electricity cost at that moment is 5.84 yuan. However, 15 minutes later within the same window, when the real-time electricity price drops to 0.78 yuan, the minute cost drops to 4.94 yuan. The distribution of the cost difference between these two moments within the stabilization prediction time window constitutes the core input for the cost stability comprehensive score. The electricity cost estimation calculates the 30-minute forward-looking electricity cost for each moment within the stabilization prediction time window. The forward-looking cost is obtained by multiplying the predicted load sequence of the load distribution curve 30 minutes after the current moment by the corresponding real-time electricity price sequence, multiplying by the time step (1 / 60 hour), and summing the results. The unit is yuan. The predicted load sequence is generated by extrapolating the weighted average of the current moment's load value and the target load of the corresponding time period in the collaborative optimization configuration. The comprehensive cost stability score is a linear weighted composite of the forward-looking electricity cost and the absolute value of the 1-minute difference of the load distribution curve at the current moment within the stabilization prediction time window. The cost weight is set to 0.6 and the stability weight is set to 0.4. The scoring formula is Score=0.6×(1-C_forward / C_max)+0.4×(1-W_current / W_max), where C_forward is the forward-looking electricity cost, C_max is the maximum forward-looking cost within the stabilization prediction time window, W_current is the absolute value of the 1-minute difference of the load distribution curve at the current moment, and W_max is the maximum absolute value of the difference within the stabilization prediction time window. The Score ranges from 0 to 1. A higher value indicates that the moment has both low electricity cost and high load stability. When the forward-looking cost is exactly the same at all moments within the stabilization prediction time window, the cost item is included with an equal value of 0.5 for all moments to avoid the zero denominator causing the cost dimension to lose its distinguishing power.

[0073] The optimal scheduling time is determined based on the comprehensive cost stability score. After the comprehensive cost stability score is calculated for each time point within the stabilization prediction time window, the time corresponding to the maximum value of the comprehensive cost stability score sequence is identified as a candidate for the optimal scheduling time. The candidate optimal scheduling time within the stabilization prediction time window corresponding to the steady state after the evening peak typically appears between 21:10 and 21:30. During this period, electricity prices have begun to decline while building load has stabilized sufficiently, and the comprehensive cost stability score reaches its daily peak in this interval. The optimal scheduling time is then verified against grid-side constraints based on the candidate times. Verification items include whether the current load factor of the distribution network feeder connected to the building is below 85%, whether the grid dispatch platform has issued a demand response signal, and whether the current operating temperature of the building transformer is below 90% of its rated temperature. If all three verifications pass, the candidate time is confirmed as the optimal scheduling time. If any verification fails, the optimal scheduling time is postponed to the next time point with the second-highest score within the stabilization prediction time window for re-verification. If all candidate times within the stabilization prediction time window fail verification, the verification is postponed to the highest-scoring time point in the next stabilization prediction time window, and a grid-side constraint anomaly alarm is triggered simultaneously. Once the optimal scheduling time is determined, a trigger signal is sent to the building and power grid collaborative control module. The collaborative control module generates collaborative optimization control instructions based on the equipment operating parameter instructions corresponding to the optimal scheduling time in the collaborative optimization configuration. The instructions encapsulate four types of parameters with the equipment number as the primary key: chiller load rate setpoint, ice storage tank melting rate setpoint, chilled water pump speed setpoint, and air supply temperature setpoint of each terminal air conditioning unit. The collaborative optimization control instructions are synchronously sent to the execution layer of each equipment at the optimal scheduling time to complete the collaborative optimization control closed loop between the building and the power grid.

[0074] At the optimal scheduling time, the building and power grid collaborative control outputs collaborative optimization control commands. After the optimal scheduling time trigger signal arrives at the building and power grid collaborative control module, the collaborative control module retrieves all equipment operating parameters corresponding to the optimal scheduling time from the collaborative optimization configuration. The retrieval results cover three subsystems: the chiller source side, the transmission and distribution side, and the terminal side. The parameters of each subsystem are encapsulated in the collaborative optimization control command in order of equipment response rate from slowest to fastest. The chiller unit load rate adjustment command is encapsulated first, and the terminal air conditioning unit temperature setpoint adjustment command is encapsulated last. This ordering ensures that the slowest responding host equipment receives the maximum action lead. After the collaborative optimization control command is encapsulated, it is synchronously pushed to the building-side interface of the power grid dispatching platform. The pushed content includes the expected change in the building's total power consumption and its duration at this scheduling time. Based on this, the power grid dispatching platform updates the distribution network feeder load forecast and incorporates the building's current collaborative response into the regional load balance calculation on the power grid side. Upon receiving the collaborative optimization control command, each device in the building's execution layer matches the corresponding parameters using the device number as an index and initiates execution. During execution, device behavior data is transmitted back to the collaborative control module in real time. The transmitted data is compared item by item with the parameter settings in the collaborative optimization control command. Devices with deviations exceeding 5% of the corresponding parameter's rated value trigger a secondary command correction. The correction command is recalculated based on the current measured value and then reissued. The execution status of the collaborative optimization control command is written to the execution log after the building and power grid collaborative control is completed. The log records the command issuance time, the response time of each device, the measured power change, and the measured load change of the power grid feeder. The consistency of these four data points constitutes the evaluation basis for the building and power grid collaborative control effect. After accumulating sufficient historical samples, the execution log provides data support for the iterative optimization of parameters in the collaborative optimization configuration.

[0075] The feasibility of the present invention is illustrated below through numerical examples.

[0076] [Example] A typical summer workday scenario is used for a 50,000 m² Grade A commercial building in Wuxi. The building is equipped with two 400 kW chiller units, an 800 kWh ice storage tank (initial SOC = 28%), and a 400 kWh hot water storage device (initial SOC = 35%), with a target water level of 75%. Peak hours are 18:00-21:00 (average electricity price 1.05 yuan / kWh), and off-peak hours are 23:00-07:00 the next day (average electricity price 0.38 yuan / kWh).

[0077] S110: Peak-hour average power 520 kW, daily average 397 kW, normalized deviation φ=+0.31; valley-hour φ=−0.24, total peak-valley deviation 0.55.

[0078] S120: V_fluctuation=0.42, P75=14.2 kW / min, Threshold_dynamic=16.6 kW / min; 16:52 triggered peak-load strong trend confirmation signal (monotonism ratio 1.48), chiller unit scheduling switchover timed ahead of schedule at 16:38. S130: r_freq=0.19, margin reserve 41.0 kW; peak-period ice storage tank released 95 kWh of cooling, heat recovery 42 kWh, actual reduction 137 kWh, execution score 0.91. Deviation verification: 19:30 outdoor temperature surge caused a 28% deviation, triggering a secondary response—output increased for the remaining period, and the target for ice production during off-peak periods decreased by 19 kWh; after 20:00, the reduction returned to 96% of the planned amount.

[0079] S140: The original calculation of the comprehensive energy gap on the cooling side was (75%−28%)×800+95=471 kWh. After the S130 second-level deviation response, the gap was reduced by 19 kWh to 452 kWh. During the peak period, 405 kWh were generated from 01:00 to 05:00, and 47 kWh were added from 06:00 to 07:00. The total electricity cost during the peak period was approximately 181 yuan.

[0080] S150: Entered the stabilization window at 21:14, achieved the highest score (Score=0.87) at 21:22, passed all grid-side constraints, and output coordinated control commands. Peak electricity costs were reduced by 91.8 yuan, and the load standard deviation decreased from 46 kW to 8 kW.

[0081] To implement the building and smart grid collaborative optimization control method corresponding to the above method embodiments, and to achieve the corresponding functions and technical effects. See also Figure 2 , Figure 2 This paper illustrates a structural block diagram of a building and smart grid collaborative optimization control system 200 provided in an embodiment of this application, comprising:

[0082] Data acquisition module 201 is used to collect building HVAC system operation status data and electricity market price information, and to perform peak-valley energy consumption status matching analysis on the operation status data and the electricity price information to generate an energy consumption status baseline;

[0083] The strategy generation module 202 is used to perform peak-valley electricity price optimization analysis based on the energy consumption status baseline to generate an electricity price response strategy, implement electricity price interval layering to extract peak-valley electricity price time boundaries, and identify trend turning points in advance based on the peak-valley electricity price time boundaries to establish peak-valley differentiated scheduling paths.

[0084] Peak scheduling module 203 is used to perform collaborative response analysis on the peak-valley differentiated scheduling path to generate equipment collaborative parameters, and based on the equipment collaborative parameters, retain the frequency regulation response margin to start the energy storage and heat recovery system to generate peak scheduling execution records;

[0085] The off-peak energy storage module 204 is used to generate an off-peak charging sequence scheme by back-calculating the off-peak charging amount based on the actual reduction amount in the peak scheduling execution record, and to start ice making and energy storage charging using the off-peak charging sequence scheme to generate off-peak energy storage parameters. The off-peak energy storage parameters are used to perform dynamic peak-valley energy allocation and compensation to build a collaborative optimization configuration.

[0086] The instruction output module 205 is used to generate a load distribution curve by monitoring the power grid load based on the collaborative optimization configuration, perform multi-objective optimization of the load distribution curve on electricity cost and load stability to determine the optimal scheduling time, and execute the building and power grid collaborative control to output collaborative optimization control instructions at the optimal scheduling time.

[0087] The aforementioned building and smart grid collaborative optimization control system 200 can implement the building and smart grid collaborative optimization control method of the above-described method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining content of this application embodiment can be referred to the content of the above method embodiments, and will not be repeated in this embodiment.

[0088] The purpose of the above embodiments is to reproduce and derive the technical solution of the present invention by way of example, and to fully describe the technical solution, purpose and effect of the present invention. The purpose is to enable the public to have a more thorough and comprehensive understanding of the disclosure of the present invention, and not to limit the scope of protection of the present invention.

Claims

1. A method for coordinated optimization control of buildings and smart grids, characterized in that, include: Collect building HVAC system operation status data and electricity market price information, and perform peak-valley energy consumption status matching analysis on the operation status data and the electricity price information to generate an energy consumption status baseline; Based on the energy consumption baseline, peak-valley electricity price optimization analysis is performed to generate an electricity price response strategy. This strategy is then used to extract peak-valley electricity price time-period boundaries through price interval stratification. This includes: performing amplitude deviation analysis on the electricity price response strategy to determine the price fluctuation range; extracting price change rate characteristics based on the price fluctuation range to generate a dynamic electricity price stratification threshold; classifying the electricity price response strategy into price levels using the dynamic price stratification threshold; performing time-period boundary calibration processing based on the price level classification to extract peak-valley electricity price time-period boundaries; and identifying trend turning points in advance based on the peak-valley electricity price time-period boundaries to establish differentiated peak-valley scheduling paths. This includes: extracting price change rate characteristics based on the peak-valley electricity price time-period boundaries to generate a trend change sequence; using the trend change sequence to identify continuous unidirectional inflection points to form a trend confirmation signal and scheduling switchover advance time; prioritizing the switching switchover advance time and the trend confirmation signal to generate switching trigger conditions; and integrating the switching trigger conditions to establish differentiated peak-valley scheduling paths. The process involves: performing collaborative response analysis on the peak-valley differentiated scheduling paths to generate equipment collaborative parameters; and, based on these parameters, initiating the energy storage and heat recovery system with a reserved frequency regulation response margin to generate peak-period scheduling execution records. This includes: generating the available output distribution for each device based on the collaborative parameters; calculating the reserved frequency regulation margin for each device's available output distribution to obtain a margin reserve; determining the effective output configuration by performing tiered startup of the energy storage and heat recovery system based on the reserved margin; and generating peak-period scheduling execution records based on the effective output configuration and execution status recordings. Based on the actual reduction in the peak period scheduling execution record, the valley period charging amount is calculated to generate a valley period charging time sequence scheme. The valley period charging time sequence scheme is used to start ice making and energy storage charging to generate valley period energy storage parameters. The valley period energy storage parameters are used to perform dynamic peak and valley energy allocation and compensation to build a collaborative optimization configuration. Based on the aforementioned collaborative optimization configuration, a load distribution curve is generated by monitoring the power grid load. The load distribution curve is then subjected to multi-objective optimization of electricity cost and load stability to determine the optimal scheduling time. At the optimal scheduling time, the building and power grid collaborative control is executed to output collaborative optimization control commands.

2. The method according to claim 1, characterized in that, The step of performing peak-valley energy consumption status matching analysis between the operating status data and the electricity price information to generate an energy consumption status baseline includes: Energy consumption feature distribution is generated by extracting energy consumption features from the operational status data according to peak and valley time periods; The electricity price information is divided into peak and valley ranges to generate peak and valley electricity price range labels; The energy consumption characteristic distribution is compared and matched with the peak-valley electricity price range label to generate an energy consumption deviation label. An energy consumption status baseline is generated based on the energy consumption deviation marker.

3. The method according to claim 1, characterized in that, The method of generating a valley period loading schedule by reverse-engineering the valley period loading amount based on the actual reduction amount in the peak period scheduling execution record includes: Extract the actual reduction amount of each device from the peak scheduling execution record to generate a reduction amount summary; Based on the sum of the reduction amounts, the total filling amount during the valley period is determined by reverse calculation of the filling target during the valley period; The total amount of water added during the trough period is allocated according to the water added amount in each time period of the trough period to generate a time period water added allocation table; Based on the time period filling allocation table, a time period priority sorting is performed to generate a valley period filling time sequence scheme.

4. The method according to claim 1, characterized in that, The process of performing multi-objective optimization of electricity cost and load stability on the load distribution curve to determine the optimal scheduling time includes: The load distribution curve is decomposed by time-series difference to generate a load fluctuation decrease sequence; From the load fluctuation decreasing sequence, locate the rapid convergence interval of fluctuations and generate a stabilization prediction time window; The electricity cost is estimated by comparing each moment within the stabilization prediction time window with the load distribution curve to generate a comprehensive cost stability score. The optimal scheduling time is determined based on the comprehensive cost stability score.

5. The method according to claim 1, characterized in that, The process of using the trend change sequence to identify consecutive inflection points in the same direction to form a trend confirmation signal and advance scheduling switchover time includes: Based on the trend change sequence, adjacent inflection points that change in the same direction continuously are extracted to generate inflection point candidate pairs; The candidate inflection points are used to generate trend confirmation signals through directional consistency verification. Based on the trend confirmation signal, the corresponding trigger time is extracted to generate the initial switching time; The initial switching time is compensated in advance to determine the advance time of the scheduling switch.

6. The method according to claim 3, characterized in that, The process of determining the total filling amount during the trough period by back-calculating the filling target based on the sum of the reduction amounts includes: The energy deficit distribution is generated by performing peak-valley energy balance verification on the sum of the compression reduction amounts. Based on the energy gap distribution, key compensation periods are identified, and a compensation priority sequence is generated. The energy gap distribution is segmented and the target is calibrated using the compensation priority sequence to generate segmented filling amounts. The total amount of water injected during the trough period is determined by summing up the segmented injection amounts.

7. A building and smart grid collaborative optimization control system, characterized in that, include: The data acquisition module is used to collect the operating status data of the building's HVAC system and the electricity market price information, and to perform peak-valley energy consumption status matching analysis on the operating status data and the electricity price information to generate an energy consumption status baseline; The strategy generation module is used to generate an electricity price response strategy based on the energy consumption status baseline by performing peak-valley electricity price optimization analysis. This strategy is then used to extract peak-valley electricity price time-period boundaries by implementing price interval stratification. This includes: performing amplitude deviation analysis on the electricity price response strategy to determine the electricity price fluctuation range; extracting electricity price change rate characteristics based on the price fluctuation range to generate a dynamic electricity price stratification threshold; classifying the electricity price response strategy into price levels using the dynamic price stratification threshold; performing time-period boundary calibration processing based on the price level classification to extract peak-valley electricity price time-period boundaries; and identifying trend turning points in advance based on the peak-valley electricity price time-period boundaries to establish differentiated peak-valley scheduling paths. This includes: extracting electricity price change rate characteristics based on the peak-valley electricity price time-period boundaries to generate a trend change sequence; using the trend change sequence to identify continuous unidirectional inflection points to form a trend confirmation signal and scheduling switch advance time; prioritizing the switching based on the scheduling switch advance time and the trend confirmation signal to generate switching trigger conditions; and integrating the switching trigger conditions to establish differentiated peak-valley scheduling paths. The peak-period scheduling module is used to perform collaborative response analysis on the peak-valley differentiated scheduling paths to generate equipment collaborative parameters, and to start the energy storage and heat recovery system based on the equipment collaborative parameters with a reserved frequency regulation response margin to generate peak-period scheduling execution records. This includes: generating the available output distribution of each device based on the equipment collaborative parameters; calculating the frequency regulation margin reservation for the available output distribution of each device to obtain the margin reservation amount; performing graded start-up of the energy storage and heat recovery system according to the margin reservation amount to determine the effective output configuration; and recording the execution status based on the effective output configuration to generate peak-period scheduling execution records. The off-peak energy storage module is used to generate an off-peak charging time sequence scheme by back-calculating the off-peak charging amount based on the actual reduction amount in the peak scheduling execution record, and to start ice making and energy storage charging using the off-peak charging time sequence scheme to generate off-peak energy storage parameters. The off-peak energy storage parameters are used to perform dynamic peak-valley energy allocation and compensation to build a collaborative optimization configuration. The instruction output module is used to generate a load distribution curve by monitoring the power grid load based on the collaborative optimization configuration, perform multi-objective optimization of the load distribution curve on electricity cost and load stability to determine the optimal scheduling time, and execute the building and power grid collaborative control to output collaborative optimization control instructions at the optimal scheduling time.

Images