Load prediction and energy-carbon optimization regulation and control method and system for multi-heat-pump system of high-speed rail station building

Abstract

The invention discloses a load prediction and energy-carbon optimization regulation and control method and system for a multi-heat-pump system of a high-speed rail station building, and the method comprises the steps: collecting temperature field distribution data and energy consumption operation records of the multi-heat-pump system, and generating a station building thermal environment reference map through space temperature gradient feature extraction and energy consumption fluctuation mode recognition by employing a phase encoding and spectrum modulation technology; identifying a hot island and a cold island by using the reference map, and constructing a thermal partition map through thermal polarity identification and polarity neutralization treatment; carbon emission data under different energy consumption levels are obtained for gravity center analysis, carbon emission step points are identified, and an energy-carbon correlation mapping table is established; load response characteristics are extracted based on the thermal inertia difference of the heat island and the cold island, and partition load prediction is carried out by combining an energy-carbon correlation mapping table to formulate a regulation and control sequence; heat pump group power distribution is carried out, a peak-avoiding operation interval is determined by referring to a carbon emission step point, and an optimization regulation and control instruction is output; and setting a heat buffer area, establishing a heat transmission network, generating a cooperative regulation scheme, and realizing intelligent cooperative regulation of the high-speed rail station building.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of building environment control and energy saving technology, in particular to a high-speed rail station multi-heat pump system load prediction and energy-carbon optimization control method and system. BACKGROUND

[0002] As a large public building, high-speed rail station has the characteristics of large space volume, strong personnel mobility and complex indoor environment, and its HVAC system energy consumption accounts for a large proportion of the total building energy consumption. The traditional multi-unit air conditioning control method mainly relies on timing adjustment and experience setting, lacks real-time response ability to indoor environment changes, and is difficult to adapt to the load fluctuation demand of peak and low valley periods.

[0003] The current building air conditioning control technology faces many challenges when applied to large space buildings such as high-speed rail stations. The existing indoor environment monitoring method is mainly based on point measurement, which cannot fully reflect the temperature distribution law and local environment difference in large space. The load calculation method mainly uses empirical formula or statistical regression, which has insufficient prediction accuracy for traffic buildings with rapid changes in passenger flow density. The device control strategy mainly uses single machine or simple grouping, lacks systematic coordination mechanism, and is prone to mutual interference and energy waste problems between devices.

[0004] Therefore, a method is needed to solve at least one of the above problems. SUMMARY

[0005] The present application provides a high-speed rail station multi-heat pump system load prediction and energy-carbon optimization control method and system, which aims to accurately identify the distribution characteristics of heat islands and cold islands in the station by establishing a thermal environment benchmark map and a zoning atlas, combined with carbon emission gravity shift analysis and step point identification technology, to construct an energy-carbon correlation mapping table, extract load response characteristics based on thermal inertia difference and carry out high-precision zoned load prediction, and then develop peak avoidance operation strategies and collaborative control schemes, to realize intelligent, refined and low-carbon operation and management of the multi-heat pump system.

[0006] The first aspect of the present application provides a high-speed rail station multi-heat pump system load prediction and energy-carbon optimization control method, comprising the following steps: Collecting temperature field distribution data of high-speed rail station and energy consumption operation records of multi-heat pump system, extracting spatial temperature gradient characteristics according to the temperature field distribution data, identifying energy consumption fluctuation mode through the energy consumption operation records, and generating a station thermal environment benchmark map by fusing the spatial temperature gradient characteristics and the energy consumption fluctuation mode; Generating high-temperature aggregation zone and low-temperature aggregation zone by temperature clustering analysis using the station thermal environment benchmark map, identifying heat islands and cold islands in the station according to the high-temperature aggregation zone and the low-temperature aggregation zone, constructing a thermal zoning atlas according to the distribution characteristics of the heat islands and the cold islands, and delineating differentiated control boundaries with the aid of the thermal zoning atlas; obtain carbon emission data at different energy consumption levels from the energy consumption operation record, perform barycentric analysis on the carbon emission data to form a carbon emission change rate curve, identify a carbon emission step point via the carbon emission change rate curve, and establish an energy-carbon correlation mapping table in accordance with the carbon emission step point and the differential control boundary; extract load response characteristics based on the thermal inertia difference between the heat island and the cold island, combine the load response characteristics with the energy-carbon correlation mapping table to carry out partitioned load prediction, generate regional predicted load values through partitioned load prediction, and formulate a preliminary control sequence in accordance with the predicted load values; perform heat pump group power distribution using the preliminary control sequence to form a distribution scheme, determine a peak-avoiding operation interval with reference to the distribution scheme and the carbon emission step point, and adjust the distribution scheme to output optimized control instructions in accordance with the peak-avoiding operation interval; set a thermal buffer region according to the thermal partitioning map, connect the heat island and the cold island via the thermal buffer region to form a heat transfer network, and generate a coordinated control scheme in accordance with the heat transfer network and the optimized control instructions.

[0007] The second aspect of the present application proposes a high-speed railway station multi-heat pump system load prediction and energy-carbon optimization control system, comprising: A data acquisition module is configured to acquire temperature field distribution data of a high-speed railway station and energy consumption operation records of a multi-heat pump system, extract spatial temperature gradient characteristics from the temperature field distribution data, identify energy consumption fluctuation patterns via the energy consumption operation records, and generate a station thermal environment benchmark map by fusing the spatial temperature gradient characteristics and the energy consumption fluctuation patterns. A partition analysis module is configured to perform temperature clustering analysis to generate a high-temperature aggregation zone and a low-temperature aggregation zone using the station thermal environment benchmark map, identify a heat island and a cold island within the station in accordance with the high-temperature aggregation zone and the low-temperature aggregation zone, construct a thermal partitioning map according to the distribution characteristics of the heat island and the cold island, and demarcate differential control boundaries with the aid of the thermal partitioning map. An energy-carbon mapping module is configured to obtain carbon emission data at different energy consumption levels from the energy consumption operation record, perform barycentric analysis on the carbon emission data to form a carbon emission change rate curve, identify a carbon emission step point via the carbon emission change rate curve, and establish an energy-carbon correlation mapping table in accordance with the carbon emission step point and the differential control boundary. A load prediction module is configured to extract load response characteristics based on the thermal inertia difference between the heat island and the cold island, combine the load response characteristics with the energy-carbon correlation mapping table to carry out partitioned load prediction, generate regional predicted load values through partitioned load prediction, and formulate a preliminary control sequence in accordance with the predicted load values. a power distribution module configured to perform heat pump group power distribution using the preliminary control sequence to form a distribution scheme, determine a peak-avoiding operation interval based on the distribution scheme and the carbon emission step point, and output an optimized control instruction according to the distribution scheme adjusted based on the peak-avoiding operation interval; a coordinated control module configured to set a heat buffer region based on the heat partition map, connect the heat island and the cold island to form a heat transfer network through the heat buffer region, and generate a coordinated control scheme based on the heat transfer network and the optimized control instruction.

[0008] The beneficial effects of the present application are embodied in the following aspects: first, by constructing a heat environment benchmark map and a heat partition map, comprehensive perception and fine management of complex heat environments in large spaces are achieved. By using temperature gradient feature extraction and phase coding modulation technology, the spatial distribution and time-varying characteristics of the heat island and cold island in the station building can be accurately identified, solving the technical problems of limited coverage and incomplete environmental feature description of traditional point monitoring methods. Second, an energy-carbon correlation mapping mechanism and a load prediction system based on heat inertia difference are established. Through carbon emission barycenter migration analysis and step point identification technology, the internal law of energy consumption change and carbon emission response is revealed. Combined with load response feature extraction and multi-step extrapolation method, the accuracy of load prediction is improved, and the unified optimization of energy consumption control and carbon emission management is realized. Finally, a multi-level coordinated control system is constructed. Through load balancing processing in heat pump group power distribution, the imbalance problem of single device overload operation and other devices light load is solved, prolonging the service life of the device and reducing the failure rate. The peak-avoiding operation scheduling effectively reduces the carbon emission intensity of the system during the peak passenger flow period by staggering the high-power operation period with the high-emission step period. The heat transfer network realizes the orderly transfer of excess heat from the heat island region to the cold island region by establishing relay sites and optimizing the transmission path in the heat buffer region, reducing the local overheating and overcooling phenomenon, and improving the uniformity of temperature distribution in the station building.

[0009] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present application. BRIEF DESCRIPTION OF DRAWINGS

[0010] The drawings herein show specific examples of the technical solutions described in the present application, and constitute part of the specification together with the specific embodiments, for explaining the technical solutions, principles and effects of the present application.

[0011] Unless specifically stated, the same reference signs in different drawings represent the same or similar technical features, and different reference signs may also be used to represent the same or similar technical features.

[0012] Figure 1This is a flowchart illustrating the method for load prediction and energy-carbon optimization control of multi-heat pump systems in high-speed railway stations according to the present invention.

[0013] Figure 2 This is a structural block diagram of the high-speed railway station multi-heat pump system load prediction and energy-carbon optimization control system of the present invention. Detailed Implementation

[0014] 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.

[0015] 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.

[0016] 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.

[0017] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0018] The technical solutions of the embodiments of this application will be described below.

[0019] like Figure 1 As shown, this embodiment of the invention provides a method for load prediction and energy-carbon optimization control of a multi-heat pump system in a high-speed railway station, including the following steps S110-S160: Step S110: Collect temperature field distribution data and energy consumption operation records of multiple heat pump systems in the high-speed railway station building. Extract spatial temperature gradient features based on temperature field distribution data. Identify energy consumption fluctuation patterns through energy consumption operation records. Integrate spatial temperature gradient features and energy consumption fluctuation patterns to generate a thermal environment baseline map of the station building.

[0020] Specifically, the temperature field distribution data of high-speed railway station and the energy consumption records of multi-heat pump system are collected. Temperature sensors are deployed in key areas such as waiting hall, ticket checking area, and commercial area according to 5m x 5m grid spacing, and are divided into ground layer 0.5m, human body layer 1.5m, and top layer 4.5m in vertical direction. The sensor uses PT1000 platinum resistance with measurement accuracy of ±0.1°C and sampling frequency of 30 seconds / second. Each sensor records data including sensor number (format is area-position-height, such as A01-L1-H2, where A01 represents the first number in area A, L1 represents the first column, and H2 represents the second layer height), timestamp, temperature value, relative humidity, and air flow rate. At the same time, the energy consumption data of multi-heat pump system are collected. Each heat pump unit is equipped with an intelligent electric meter to monitor parameters including unit number, instantaneous power, cumulative power, heating and cooling capacity, COP performance coefficient, operating mode, and load rate. The data collection system converges to the central database through wireless protocol, establishes a unified time reference, and ensures the synchronization of temperature data and energy consumption data. The three-dimensional interpolation of temperature field uses Kriging method to convert discrete monitoring points into continuous temperature field. The interpolation grid resolution is set to 1m x 1m x 1m to improve spatial analysis accuracy. Compared with the interpolation of denser sensor deployment interval, it can capture local temperature change details. The unit-area service relationship mapping is established to determine the influence range and energy consumption contribution of each heat pump.

[0021] According to the temperature field distribution data, the spatial temperature gradient features are extracted. The finite difference method is used to calculate the temperature change rate in three spatial directions: The gradient amplitude quantifies the intensity of temperature spatial change. The gradient direction is indicated by a unit vector, indicating the direction of the fastest temperature increase. The horizontal temperature gradient mainly reflects the temperature difference between different functional areas in the station building, and the vertical temperature gradient reflects the temperature stratification formed by the rising of hot air. The time evolution analysis of gradient features reveals the dynamic characteristics of the temperature field. The gradient increases during morning and evening peak periods, and decreases at night. Key gradient area identification includes the temperature difference area inside and outside the entrance, the thermal convection area near the skylight, and the local high temperature area of equipment room, etc. The statistical characteristics of gradient field include average gradient, maximum gradient, and gradient standard deviation, which represent the uniformity of temperature field. Gradient anomaly detection identifies temperature mutations beyond the normal range, which may correspond to heat source anomalies or poor ventilation.

[0022] The energy consumption fluctuation pattern is identified by energy consumption operation records. The time series analysis method decomposes the energy consumption data into trend, periodic and random terms. The trend term reflects long-term energy consumption changes, the periodic term reflects daily and weekly cycles, and the random term corresponds to sudden load changes. The fluctuation amplitude is calculated by the ratio of the standard deviation to the mean of energy consumption, quantifying the stability of energy consumption. The fluctuation frequency analysis uses the fast Fourier transform to identify the main frequency components of energy consumption changes. Typical frequencies include 24-hour cycles, 12-hour cycles, and weekday / weekend differences. Peak-valley feature extraction identifies the daily energy consumption peak and valley periods, and the peak-valley ratio reflects the degree of load imbalance. Multi-heat pump linkage mode analysis identifies the energy consumption correlation of different heat pumps. Synchronous fluctuations indicate cooperative operation, and reverse fluctuations indicate complementary operation. Abnormal fluctuation detection identifies energy consumption fluctuations that deviate from the normal pattern using statistical process control methods. The causes may include extreme weather, equipment failure, or large passenger flow. Fluctuation patterns include smooth, periodic, sudden, and mixed types, each requiring different control strategies.

[0023] In some embodiments, the fusion of the spatial temperature gradient feature and the energy consumption fluctuation pattern generates a station house thermal environment benchmark map, including: temperature phase coding of the spatial temperature gradient feature to generate a phase feature code; establishing an energy consumption oscillation spectrum based on the energy consumption fluctuation pattern; modulating the energy consumption oscillation spectrum with the phase feature code to form a modulated signal; decoding and reconstructing the modulated signal to generate a station house thermal environment benchmark map.

[0024] The spatial temperature gradient feature is temperature phase coded to generate a phase feature code. The temperature gradient vector is converted to a phase representation, with the horizontal phase angle representing the gradient direction in the horizontal plane, with a value range of [0, 2π]. The vertical phase angle represents the angle between the gradient vector and the vertical direction. Phase quantization discretizes continuous phase values into a finite number of phase levels. The horizontal phase is quantized into 16 directions (one level every 22.5°), and the vertical phase is quantized into 8 levels. The phase feature code uses binary encoding, with 4 bits for the horizontal phase, 3 bits for the vertical phase, and 3 bits for the gradient amplitude level, forming a 10-bit spatial point phase feature code. The time dimension phase coding considers the temperature gradient's time change rate, with fast changes corresponding to high-frequency phases and slow changes corresponding to low-frequency phases. The spatial distribution of the phase feature code forms a phase coding matrix, with each element of the matrix corresponding to the coding value of a grid point in space. Phase shift detection identifies spatial and temporal mutations in the phase feature code, which is used to locate abnormal areas of the thermal environment.

[0025] The energy consumption fluctuation spectrum is established based on the energy consumption fluctuation pattern. The time-domain energy consumption fluctuation data is converted to the frequency domain by Fourier transform to obtain the frequency spectrum distribution of energy consumption. The dominant frequency of energy consumption fluctuation is extracted by the main frequency extraction, which usually includes the daily frequency (1 / 24 hours), half-day frequency (1 / 12 hours), etc. The energy proportion of each frequency component is analyzed by the energy distribution analysis of the spectrum, and the concentrated distribution indicates strong regularity, while the dispersed distribution indicates great randomness. The integer multiple frequency components of the main frequency are identified by the harmonic analysis, and the harmonic content reflects the nonlinear characteristics of energy consumption fluctuation. The coupling spectrum of multiple heat pump systems is obtained by cross-spectrum analysis, which reveals the frequency domain correlation of different heat pump energy consumptions. The time-varying characteristics of the spectrum are analyzed by short-time Fourier transform to establish the time-frequency joint distribution diagram. The abnormal frequency detection identifies the frequency components that do not belong to the normal operation mode, which may correspond to equipment oscillation or control instability. The parametric representation of the oscillation spectrum includes characteristic parameters such as center frequency, bandwidth, and quality factor.

[0026] The energy consumption fluctuation spectrum is modulated by the phase feature code to form a modulated signal. Phase modulation (PM) is used to map the phase feature code to the phase change of the carrier. The carrier selects the main frequency component of the energy consumption fluctuation spectrum, and the modulation depth is dynamically adjusted according to the value of the phase feature code. The modulation function is represented as: where A is the amplitude, f_c is the carrier frequency, β is the modulation index, and P is the phase feature code. Spatial modulation modulates the phase feature codes of different spatial positions onto different frequency components, realizing frequency domain multiplexing of spatial information. Time modulation considers the time-varying characteristics of the phase feature code, and uses a time-varying modulation index to adapt to the dynamic changes of the temperature field. Multi-carrier modulation utilizes multiple peak frequencies of the energy consumption fluctuation spectrum to improve the information transmission capacity. Spectrum analysis of the modulated signal verifies the modulation effect, and the sideband power distribution reflects the rationality of the modulation depth. Nonlinear compensation of the modulation process reduces signal distortion and maintains the accuracy of the temperature-energy consumption relationship.

[0027] The modulated signal is decoded and reconstructed to generate the station building thermal environment benchmark map. The phase information is extracted using the same carrier frequency as the modulation through coherent demodulation method. The phase decoding reversely maps the demodulated phase values to temperature gradient features and energy consumption features. Spatial reconstruction converts the frequency domain information back to the spatial domain through inverse Fourier transform to restore the thermal environment parameters of each spatial point. The data structure of the benchmark map includes temperature field, temperature gradient field, energy consumption distribution field, and comprehensive thermal comfort index. Visual rendering uses pseudo-color coding, with warm colors representing temperature, brightness representing energy intensity, and arrows representing gradient direction. Multi-level information integration superimposes static building structure information and dynamic thermal environment information. Time slicing function supports viewing the thermal environment state at different times. Through the fusion of temperature gradient features and energy consumption fluctuation patterns, the benchmark map reflecting the overall thermal environment of the station building is finally generated.

[0028] Step S120, temperature clustering analysis is performed using the station building thermal environment benchmark map to generate high-temperature aggregation zones and low-temperature aggregation zones, hot islands and cold islands in the station building are identified according to the high-temperature aggregation zones and the low-temperature aggregation zones, a thermal zoning map is constructed according to the distribution characteristics of the hot islands and the cold islands, and a differential control boundary is demarcated with the aid of the thermal zoning map.

[0029] Specifically, temperature clustering analysis is performed using the station building thermal environment benchmark map to generate high-temperature aggregation zones and low-temperature aggregation zones. The clustering analysis takes temperature value, temperature gradient amplitude, spatial coordinates, and gradient direction as feature vectors to enhance the recognition ability of the region with sharp temperature changes. The clustering number is set to 5 levels: high-temperature zone (> 26℃), higher-temperature zone (24-26℃), moderate-temperature zone (20-24℃), lower-temperature zone (18-20℃), and low-temperature zone (< 18℃). The distance measurement of the clustering algorithm uses the Euclidean distance, and the iterative convergence threshold is set to 0.01. High-temperature aggregation zone identification is performed by connected component analysis, which merges adjacent high-temperature zone and higher-temperature zone grid points into continuous regions, and the minimum aggregation area threshold is set to 25 m². The same method is used for low-temperature aggregation zone identification, which merges the continuous regions of low-temperature zone and lower-temperature zone. The shape feature analysis of the aggregation zone includes area, perimeter, compactness, and aspect ratio. Time series analysis is used to analyze the temperature stability and change trend of each aggregation zone, and the duration of more than 2 hours is defined as a stable aggregation zone.

[0030] Hot islands and cold islands in the station building are identified according to the high-temperature aggregation zones and the low-temperature aggregation zones. The hot island identification criteria are: high-temperature aggregation zone area greater than 100 m², temperature difference with the surrounding moderate-temperature zone more than 3℃, and duration more than 4 hours. The cold island identification criteria are: low-temperature aggregation zone area greater than 50 m², temperature difference with the surrounding moderate-temperature zone more than 2℃, and duration more than 2 hours. The hot island intensity is calculated using the temperature contrast method, with the intensity value I_h=(T_c-T_a)×A_h, where T_c is the temperature of the hot island center, T_a is the temperature of the surrounding environment, and A_h is the area of the hot island, with units of ℃·m². The cold island intensity calculation formula is I_c=(T_a-T_c)×A_c, where A_c is the area of the cold island, ensuring that the intensity value is positive. The hot island type classification includes: equipment heat source type (equipment room, power transformation room), personnel aggregation type (waiting area, ticket checking port), sunlight irradiation type (glass curtain wall, skylight area), and poor ventilation type (corner, dead angle area). The cold island type classification includes: air conditioning outlet type (near air conditioning outlet), natural ventilation type (passage, overpass), and external cold source type (region connected with outdoor), and the type information is used to guide the control strategy. Spatial relationship analysis of hot islands and cold islands identifies the mutual influence region pairs, and calculates the distance and influence strength between them.

[0031] In some embodiments, the constructing the thermal zoning map according to the distribution characteristics of the heat island and the cold island comprises: performing thermal polarity identification based on the distribution characteristics of the heat island and the cold island to generate positive polarity thermal zones and negative polarity thermal zones; performing polarity neutralization processing on the negative polarity thermal zones by using the positive polarity thermal zones to form balanced thermal zones; and performing spatial aggregation division on the balanced thermal zones to generate aggregated zones, and constructing the thermal zoning map by boundary solidification on the aggregated zones.

[0032] The thermal polarity identification based on the distribution characteristics of the heat island and the cold island generates positive polarity thermal zones and negative polarity thermal zones. The thermal polarity is defined based on the average temperature of the station building, the heat island region is defined as the positive polarity thermal zone, and the cold island region is defined as the negative polarity thermal zone. The positive polarity thermal zone directly uses the heat island intensity value I_h to represent the polarity intensity, and the negative polarity thermal zone uses the cold island intensity value I_c to represent the polarity intensity. The polarity weight is adjusted according to the heat island type: the equipment heat source type weight is 1.2, the personnel gathering type weight is 1.0, the sunlight irradiation type weight is 0.8, and the poor ventilation type weight is 1.1. The spatial range of the polarity region is expanded to the influence radius of the heat island and the cold island, which is determined according to the temperature decay law, and the decay coefficient is set to 0.1℃ / m. The positive polarity thermal zone is divided into strong positive polarity (I_h>300℃·m²), medium positive polarity (100-300℃·m²), and weak positive polarity (30-100℃·m²) according to the intensity. The negative polarity thermal zone is divided into strong negative polarity (I_c>200℃·m²), medium negative polarity (80-200℃·m²), and weak negative polarity (20-80℃·m²) according to the intensity. The polarity interaction analysis identifies the adjacency relationship and mutual influence intensity of the positive and negative polarity regions, providing pairing basis for neutralization processing.

[0033] For example, the polarity neutralization processing on the negative polarity thermal zones by using the positive polarity thermal zones to form balanced thermal zones comprises: extracting thermal intensity values from the positive polarity thermal zones and the negative polarity thermal zones; determining a neutralization processing weight based on the thermal intensity values; performing thermal balance adjustment to generate an adjustment parameter according to the neutralization processing weight; and forming balanced thermal zones by using the adjustment parameter for polarity neutralization.

[0034] The thermal intensity values are extracted from the positive polarity thermal zones and the negative polarity thermal zones. The thermal intensity value of the positive polarity thermal zone uses I_h, and the thermal intensity value of the negative polarity thermal zone uses I_c, which ensures consistency with the foregoing heat island and cold island analysis. The time weighting processing considers the duration of the polarity region, and the longer the duration, the greater the weight. The weight coefficient w_t=√(duration hours / 24). The type weight w_p is determined according to the heat island and cold island type. The combination weight of the equipment heat source type and the air conditioner outflow type is the highest (1.5), because the two types have the best neutralization effect. The comprehensive intensity value is calculated as I_w=I×w_t×w_p, where I is the original intensity value (I_h or I_c). The spatial distribution analysis of the intensity value identifies high-intensity concentration zones and low-intensity dispersion zones, guiding the development of neutralization strategies.

[0035] The neutralization processing weight is determined based on the thermal intensity value. The neutralization weight distribution follows the intensity matching principle, and the region pair with similar positive and negative polarity intensity values has a higher neutralization weight. The weight calculation formula is W = 1 / (1+|I_hw-I_cw| / min(I_hw,I_cw)), where I_hw is the positive polarity weighted intensity, and I_cw is the negative polarity weighted intensity. The distance decay weight considers the spatial distance of the positive and negative polarity regions, and the distance weight is W_d = exp(-d² / 400), where d is the region center distance (m). The comprehensive weight is W_f = W × W_d, which ensures that the region pair with spatial proximity and intensity matching is preferentially neutralized. The weight normalization processing ensures that the total weight of all region pairs is 1.0. The priority ranking lists the region pair with the highest weight as the priority neutralization target, and the weight threshold is set to 0.1. The region pair below the threshold does not participate in the neutralization process. The neutralization pairing result records the corresponding negative polarity region and its weight distribution for each positive polarity region.

[0036] The thermal balance adjustment generates adjustment parameters according to the neutralization processing weight. Based on the principle of heat conservation, the heat transferred from the positive polarity region is equal to the heat received by the negative polarity region. The adjustment intensity parameter a = W_f × min(I_hw,I_cw) / max(I_hw,I_cw). The temperature adjustment amplitude AT_h = -a × (T_h-T_m) / A_h for the positive polarity region cooling, and AT_c = a × (T_m-T_c) / A_c for the negative polarity region heating, where T_h is the hot zone temperature, T_c is the cold zone temperature, and T_m is the average temperature. The spatial diffusion parameter s = sqrt(A_r / π) defines the influence range of the adjustment effect, where A_r is the area of the region. The time response parameter t is determined according to the heat capacity of the region, t = 0.2 × V^(1 / 3), where V is the volume of the region (m³). The adjustment rate parameter controls the speed of temperature change, and the rate v = AT / t. The rationality verification of the adjustment parameter ensures that the temperature change does not exceed ±2℃ / h.

[0037] The temperature of the positive polarity region is adjusted by a step of ΔT_h, and the temperature of the negative polarity region is adjusted by a step of ΔT_c. The temperature gradient transition is established at the boundary of the adjustment region by spatial gradual processing to avoid temperature mutation. The adjustment speed is controlled by the time response parameter τ, and the temperature change follows the exponential decay law T(t)=T_0+ΔT×(1-exp(-t / τ)), where T_0 is the initial temperature. The temperature change in the adjustment process is monitored in real time to track the neutralization effect, and the adjustment is stopped when the temperature difference between the positive and negative polarity regions is less than 1℃. The balanced heat zone is identified as the region with a polarity intensity of less than 30℃·m². The residual polarity processing records the residual intensity value of the incomplete neutralization region to provide a reference for the next round of adjustment. The balanced quality evaluation calculates the temperature variance change rate before and after neutralization, and the change rate greater than 60% is considered as high-quality balance.

[0038] The balanced heat zone is spatially aggregated to generate aggregated partitions. Based on the spatial adjacency relationship and temperature similarity of the balanced heat zone, adjacent balanced regions with a temperature difference of less than 0.5℃ are merged into an aggregated partition. The region growing method is used to expand and merge adjacent regions that meet the conditions around the seed point of the balanced heat zone. The minimum aggregated area threshold is set to 200m², and isolated balanced regions with an area less than the threshold are merged into adjacent larger aggregated partitions. The encoding of the aggregated partition uses the form of B+ serial number, such as B01, B02, etc., which is convenient for management and identification. Partition attribute statistics include aggregated area, average temperature, temperature standard deviation, number of original balanced regions contained, residual polarity intensity, etc. Partition connectivity analysis identifies the connection channels and isolation boundaries between aggregated partitions. Partition stability evaluation is analyzed through time series, and the aggregated partition with high stability is suitable for long-term control, and the frequently changing partition needs dynamic adjustment. The partition feature vector construction contains temperature, area, shape, stability and other multi-dimensional information.

[0039] The boundary of the aggregated partition is solidified to construct a thermal partition map. The irregular boundary of the aggregated partition is smoothed into a regular geometric shape by using isotherm fitting method. The boundary line simplification is performed by Douglas-Peucker algorithm, and the simplification threshold is set to 2 meters to reduce the number of nodes while maintaining the shape of the boundary. The data structure of the partition map includes geometric information, attribute information and correlation information. The geometric information records the boundary coordinates, area, perimeter and other spatial parameters of the partition. The attribute information includes partition code, temperature statistics, thermal characteristics, control requirements, residual polarity, etc. The correlation relationship describes the adjacency relationship, influence relationship and dependence relationship between partitions. The map visualization uses different colors to represent different types of partitions, and the color depth reflects the temperature, and the boundary thickness represents the importance of the partition. Multi-scale display supports zooming in and out from the overall view of the station building to the local details. The partition boundary and attributes are dynamically adjusted according to the real-time temperature data to maintain the timeliness of the map.

[0040] The differential control boundary is delineated by means of a thermal zoning map. Based on the temperature characteristics of the zones, control requirements, and the energy consumption fluctuation pattern identified in step S110, zones with similar control requirements are merged into a unified control area. The high-temperature control zone includes all the aggregated zones that need to be cooled and the overlapping part of the high-energy consumption peak area, and the control target is to reduce the temperature by 2-4°C and cut peak energy consumption. The low-temperature control zone includes aggregated zones that need to be heated and the low-energy consumption valley area, and the control target is to raise the temperature by 1-3°C and improve equipment utilization. The balanced control zone includes aggregated zones with suitable temperature and stable energy consumption, and the control target is to maintain the status quo. The spatial range of the control boundary takes into account the service radius of the HVAC system and the energy consumption allocation, and each control area corresponds to a specific air conditioning unit or control valve. A boundary buffer zone is set at the junction of different control areas, with a buffer width of 5-10 meters to avoid control conflicts. The control priority is determined by considering the temperature deviation and the degree of energy consumption anomaly, with the area with the highest comprehensive score being given the highest priority. The control strategy mapping establishes the correspondence between the control boundary and specific adjustment measures, including air volume adjustment, temperature setting, operation mode, etc., and finally determines the differential control boundary in the station building.

[0041] In step S130, carbon emission data at different energy consumption levels is obtained from the energy consumption operation record, and a carbon emission change rate curve is formed by barycentric analysis of the carbon emission data. The carbon emission change rate curve is used to identify the carbon emission step point, and an energy-carbon correlation mapping table is established according to the carbon emission step point and the differential control boundary.

[0042] Specifically, carbon emission data at different energy consumption levels is obtained from the energy consumption operation record. The carbon emission calculation formula is C = E × F, where C is the carbon emission (kgCO2), E is the electric energy consumption (kWh), and F is the grid carbon emission factor (kgCO2 / kWh). The current national grid carbon emission factor is 0.5810. The energy consumption level is classified according to the total power of the heat pump system: low energy consumption level (<500kW), medium energy consumption level (500-1000kW), high energy consumption level (1000-1500kW), and peak energy consumption level (>1500kW). Different energy consumption levels correspond to different barycentric migration characteristics. The barycenter is stable at low energy consumption, and the barycenter migrates quickly at peak energy consumption. The carbon emission data corresponding to each energy consumption level includes instantaneous carbon emission rate, cumulative carbon emission, carbon emission intensity per unit area, and carbon emission efficiency. Time series carbon emission data is collected at 5-minute intervals to establish energy consumption-carbon emission data pairs, with a data format of [timestamp, energy consumption value, carbon emission value, spatial location, energy consumption level]. The spatial distribution of carbon emission data is allocated according to the service area of the heat pump unit, and the carbon emission of each aggregated zone is allocated proportionally according to the service area.

[0043] In some embodiments, the center of gravity analysis on the carbon emission data forms a carbon emission change rate curve, including: performing carbon emission center of gravity identification on the carbon emission data to determine a carbon emission center of gravity position; obtaining a center of gravity migration trajectory based on the carbon emission center of gravity position; calculating a carbon emission dynamic offset amount using the center of gravity migration trajectory; and forming a carbon emission change rate curve according to the change slope of the carbon emission dynamic offset amount.

[0044] The carbon emission center of gravity identification on the carbon emission data determines a carbon emission center of gravity position. Based on the spatial distribution characteristics of the carbon emission data, the station building is divided into a grid system, and the grid resolution is set to 10 m x 10 m. The carbon emission value of each grid is obtained by accumulating the carbon emission amounts of all devices in the area, while considering the energy consumption level weight, with a peak energy consumption level weight of 1.5, a high energy consumption weight of 1.2, a medium energy consumption weight of 1.0, and a low energy consumption weight of 0.8. The center of gravity calculation uses a weighted centroid algorithm, with the X-direction center of gravity coordinates being X_c = Σ(C_i x W_i x X_i) / Σ(C_i x W_i), and the Y-direction center of gravity coordinates being Y_c = Σ(C_i x W_i x Y_i) / Σ(C_i x W_i), where C_i is the carbon emission amount of the i-th grid, W_i is the energy consumption level weight, and X_i and Y_i are the grid center coordinates. The stability of the center of gravity position is determined by the change amplitude of the center of gravity position in consecutive time periods, and a change of less than 5 meters is identified as a stable center of gravity. The center of gravity position data structure includes [timestamp, X_c, Y_c, stability marker, dominant energy consumption level].

[0045] For example, the center of gravity migration trajectory is obtained based on the carbon emission center of gravity position, including: evaluating the station building carbon emission dynamic characteristics based on the carbon emission center of gravity position to determine the tracking accuracy requirement, the station building carbon emission dynamic characteristics including peak-valley emission difference, regional contribution ratio, and passenger flow influence degree; recalibrating the carbon emission monitoring point based on the tracking accuracy requirement to form a monitoring point layout; identifying a trajectory conflict point using the monitoring point layout; and obtaining the center of gravity migration trajectory through the trajectory conflict point.

[0046] The dynamic characteristics of carbon emissions of the station building are evaluated based on the carbon emission barycenter position to determine the tracking accuracy requirement. The historical data and stability markers of the barycenter position are used to calculate the difference between the daily maximum carbon emissions and the daily minimum carbon emissions, and the corresponding energy consumption level changes are recorded. The regional contribution ratio analyzes the influence weight of each functional area on the barycenter position. The contribution ratio is calculated by the product of the regional carbon emissions and the distance from the barycenter, and the proportion of the total weight. The passenger flow influence degree is obtained by the correlation analysis of passenger flow density and carbon emission intensity, reflecting the influence degree of personnel activities on carbon emission distribution. The tracking accuracy requirement is set according to the strength of the dynamic characteristics: when the peak-valley emission difference is greater than 100 kgCO2, the accuracy requirement is ±2 meters, when it is 50-100 kgCO2, it is ±5 meters, and when it is less than 50 kgCO2, it is ±10 meters. The accuracy level is selected based on ensuring that 95% of the trajectory point positioning errors are within the allowable range. When the regional contribution ratio difference is greater than 0.3, the accuracy level is increased, and when the passenger flow influence degree coefficient is greater than 0.7, the time resolution requirement is increased. The comprehensive accuracy level is divided into three grades: high precision (±2 meters, 1 minute), medium precision (±5 meters, 5 minutes), and low precision (±10 meters, 15 minutes).

[0047] The carbon emission monitoring points are recalibrated according to the tracking accuracy requirement to form the monitoring point layout. According to the historical trajectory range and accuracy requirement of the barycenter position, the monitoring point density in the high precision area is 1 per 100 square meters, 1 per 200 square meters in the medium precision area, and 1 per 500 square meters in the low precision area. The point layout uses a grid optimization algorithm to ensure that the key areas that the barycenter may pass through are adequately monitored. The monitoring point numbering rule is MP+accuracy level+serial number, such as MP-H-01 for the first monitoring point in the high precision area, MP-M-01 for the medium precision point, and MP-L-01 for the low precision point. The point attribute record includes monitoring point number, spatial coordinates, accuracy level, monitoring radius, data collection frequency, etc. The monitoring radius is determined according to the accuracy requirement: 20 meters for high precision points, 40 meters for medium precision points, and 80 meters for low precision points. The point redundancy design increases the monitoring points at the intersection of the barycenter historical trajectory, and the redundancy rate is set to 20% to ensure the monitoring reliability of the key positions.

[0048] The trajectory conflict point is identified by the layout of the monitoring points. The overlap index is used for evaluation, and the overlap index is the ratio of the overlapping area to the combined area. The conflict point classification includes: strong conflict point (overlap index > 0.6), medium conflict point (overlap index 0.3-0.6), weak conflict point (overlap index 0.1-0.3). The spatial coordinates of the conflict point are determined by the geometric center of the overlapping area, and the conflict point number is derived from the monitoring point number: CF-MP-H-01-MP-H-02 represents the conflict point formed by two high-precision monitoring points. The conflict impact analysis evaluates the influence of the conflict point on the trajectory tracking accuracy. Strong conflict points may cause trajectory judgment errors, and auxiliary monitoring points or monitoring radius adjustment are needed. Conflict resolution strategies include increasing monitoring point density, adjusting monitoring radius, setting priority rules, etc. to ensure accurate trajectory tracking.

[0049] The barycenter migration trajectory is obtained by the trajectory conflict point. Based on the spatial distribution and time sequence of the conflict points, the discrete barycenter positions are connected into a continuous trajectory. The trajectory nodes include the measured barycenter position, conflict point position and interpolation node, and the shortest path connection principle is used between nodes. When it is impossible to avoid, the path selection adopts probability weight, and the weight is related to the energy consumption level. The trajectory smoothing processing adopts spline interpolation to eliminate the jagged fluctuation of the trajectory, and maintains the continuity and rationality of the trajectory. The multi-trajectory fusion processing deals with the case of multiple possible trajectories at the same time, and the optimal trajectory is obtained by confidence weighted average. The trajectory data structure includes trajectory number, node coordinate sequence, timestamp sequence, energy consumption level sequence, confidence, etc. The trajectory quality is evaluated by integrity, continuity and rationality to ensure that the obtained trajectory can accurately reflect the barycenter migration law.

[0050] The carbon emission dynamic displacement is calculated by the barycenter migration trajectory. The Euclidean distance formula is used for calculation, which reflects the deviation of the barycenter from the reference position. The displacement direction angle represents the direction characteristics of the displacement, which is calculated by the inverse tangent function. The displacement velocity is defined as the change rate of the displacement at adjacent time, which reflects the speed of barycenter migration and is related to the change of energy consumption level. The displacement acceleration reflects the trend of displacement change, and the positive acceleration indicates that the displacement increases, and the negative acceleration indicates that the displacement decreases. The statistical characteristics of dynamic displacement include mean, variance, maximum, minimum, etc., which are grouped according to the energy consumption level. The periodicity analysis of the time sequence of the displacement identifies the regular change mode such as daily cycle and half-day cycle. The abnormal displacement detection identifies the displacement events that exceed the normal range, which provides the basis for system anomaly diagnosis.

[0051] A carbon emission change rate curve is formed according to the change slope of the carbon emission dynamic offset. A difference method is used to divide the offset difference between two time points by the time interval. Slope smoothing processing uses a sliding window average, with a window size of 15 minutes to eliminate short-term noise interference. The vertical coordinate of the change rate curve is the slope value (meters / minute), and the horizontal coordinate is the time axis (hours), while the corresponding energy consumption level change is marked, which intuitively reflects the relationship between the carbon emission gravity migration speed and the energy consumption level. Curve feature point recognition includes peak points, valley points, inflection points, etc. These feature points correspond to the key moments of carbon emission change and the energy consumption level conversion points. Curve trend analysis identifies rising and falling trends through first-order derivative, and identifies acceleration and deceleration stages through second-order derivative. The piecewise linear fitting of the change rate curve decomposes the continuous curve into several straight line segments, each of which corresponds to a relatively stable change mode and energy consumption level. Curve parameterization includes characteristic parameters such as slope range, change amplitude, duration, and dominant energy consumption level.

[0052] Carbon emission step points are identified through the carbon emission change rate curve. Step points are defined as positions in the change rate curve where the slope suddenly changes, representing the conversion time of the carbon emission mode, usually corresponding to the switching of the energy consumption level. The step detection algorithm uses gradient mutation analysis to calculate the slope difference between adjacent points. Points with a slope difference exceeding a set threshold are marked as candidate step points. The step threshold is dynamically determined according to the overall change amplitude of the curve, using the slope mean plus twice the standard deviation as the judgment standard to ensure that the recognition rate reaches more than 90%. Step strength is evaluated by the ratio of jump amplitude to local change, reflecting the significance of the step. Step point classification includes: positive step (slope suddenly increases, usually corresponding to energy consumption level rise), negative step (slope suddenly decreases, corresponding to energy consumption level decrease), and shock step (continuous positive and negative jumps in a short time, corresponding to frequent start and stop of equipment). The time marker of the step point is accurate to the minute level, and the spatial position is determined by the gravity coordinates at the corresponding time. Step persistence checks whether the state after the step is stable, and short jumps with a duration of less than 10 minutes are not recognized as valid step points. Multiple step processing combines consecutive steps with a time interval of less than 30 minutes into a step group, with the maximum intensity step as the representative point.

[0053] The carbon emission step point and the differential control boundary are established according to the energy-carbon correlation mapping table. Determine the control area to which each step point belongs, and determine it through the geometric algorithm of the point in the polygon. The time correlation analysis identifies the correspondence between the time when the step point occurs and the running state of the heat pump equipment in the control boundary. The data structure of the mapping table includes: step point number, occurrence time, spatial coordinates, step intensity, step type (positive / negative / oscillation), control area, associated equipment, carbon emission change, triggering energy consumption level, and the like. The energy-carbon correlation coefficient is calculated by the ratio of the energy consumption change before and after the step to the carbon emission change, which reflects the carbon emission response degree corresponding to the unit energy consumption change, and is counted according to the step type. The control response delay analysis calculates the time difference between the control action and the carbon emission step, and the average delay of the positive step is 2-3 minutes, the average delay of the negative step is 3-5 minutes, and the oscillation step needs special processing strategy. The mapping accuracy is evaluated by comparing the predicted carbon emission with the actual carbon emission, and the prediction accuracy is calculated according to the step type. Through the accurate correspondence between the step point and the control boundary, and combined with the energy consumption level and the step type information, a complete correlation mapping table of station building energy consumption and carbon emission is finally established.

[0054] In step S140, based on the thermal inertia difference of the heat island and the cold island, the load response characteristics are extracted, the partition load prediction is carried out by combining the load response characteristics and the energy-carbon correlation mapping table, the predicted load value of each area is generated through the partition load prediction, and the preliminary regulation sequence is formulated according to the predicted load value.

[0055] Specifically, the load response characteristics are extracted based on the thermal inertia difference of the heat island and the cold island. The temperature-load response curve analysis is adopted. In the heat island area, due to high building density and strong heat storage capacity, the load response time corresponding to 1℃ temperature change is 20-40 minutes, and in the cold island area, due to open space and small heat capacity, the response time is only 5-15 minutes. The load response time constant is obtained by exponential fitting, which represents the time required for the load to reach 63% of the steady-state value. The response amplitude characteristic is measured by the load change caused by unit temperature change. The response amplitude of the heat island area is 15-25 kW / ℃, and that of the cold island area is 8-15 kW / ℃. The response lag characteristic reflects the delay time from control action to load change. The lag of the heat island area is 10-20 minutes, and that of the cold island area is 3-8 minutes. The response stability is evaluated by the ratio of the load standard deviation to the mean value. The ratio less than 0.1 is a stable response, and the ratio greater than 0.2 is an unstable response. The statistical analysis of the load response characteristics shows that the heat island area has the characteristics of "slow response, large amplitude, and long lag", and the cold island area has the characteristics of "fast response, small amplitude, and short lag".

[0056] In some embodiments, the load prediction based on the partition of the load response feature and the carbon-energy correlation mapping table comprises: extracting a load inertia factor and a load sensitivity factor from the load response feature; performing inertia correction on the carbon-energy correlation mapping table using the load inertia factor to generate a corrected mapping table; performing multi-step load inference based on the load sensitivity factor and the corrected mapping table; and performing partitioned load prediction based on the multi-step load inference results classified by region.

[0057] The load inertia factor and the load sensitivity factor are extracted from the load response feature. Based on the normalized calculation of the response time constant, the inertia factor I = τ / (τ+τ_ref), where τ is the regional response time constant (minutes), and τ_ref is the reference time constant, which is 15 minutes. The value is determined based on the statistical mean of the station building thermal inertia. The inertia factor value ranges from 0 to 1. The inertia factor of the heat island region is usually 0.6-0.8, and the inertia factor of the cold island region is usually 0.2-0.4. The load sensitivity factor is based on the comprehensive evaluation of response amplitude and stability. The sensitivity factor S = A / (1+σ), where A is the response amplitude coefficient (kW / ℃), and σ is the fluctuation coefficient (dimensionless). The sensitivity factor reflects the sensitivity of the load to temperature changes. The sensitivity factor of the heat island region is 12-20, and the sensitivity factor of the cold island region is 6-12. The spatial distribution of the inertia factor corresponds to the heat island and cold island distribution in S120. The heat island intensive area has high inertia characteristics, and the cold island intensive area has low inertia characteristics. The time-varying characteristics of the sensitivity factor take into account the influence of passenger flow and equipment operating state. The sensitivity factor increases by 20%-30% during the daytime peak period and decreases by 15%-25% during the night low period.

[0058] The inertia correction factor is used to correct the energy-carbon correlation mapping table to generate a corrected mapping table. Based on the correlation coefficient in the previously established energy-carbon correlation mapping table, an inertia influence correction factor is introduced, and the correction formula is η_new=η_original×(1+k×I), wherein η_new is the corrected energy-carbon response coefficient (kW / kgCO2), η_original is the original energy-carbon response coefficient (i.e., the inverse of the ratio of energy consumption change to carbon emission change), k is the correction intensity coefficient, which is 0.6, and this value is determined by historical data regression analysis to ensure the minimum prediction error, and I is the load inertia factor (dimensionless). The response coefficient in the high inertia region (I>0.6) is adjusted upwards, reflecting the cumulative effect and lagging characteristics of load changes. The response coefficient in the low inertia region (I<0.4) is adjusted downwards, reflecting its fast response and short-term impact characteristics. The inertia correction of the response delay time uses the method of delay time multiplied by (1+I), and the delay time in the high inertia region is correspondingly lengthened. The inertia correction of the carbon emission step point intensity considers the influence of inertia on the response amplitude of carbon emission, and the corrected intensity uses the inverse proportion adjustment method. The corrected mapping table maintains the original data structure and fields, and increases the inertia correction identifier and correction parameter record. The spatial interpolation method extends the correction parameters of discrete step points to continuous regions, and the Kriging interpolation is used to ensure spatial continuity.

[0059] Based on the load sensitivity factor and the corrected mapping table, a multi-step load deduction is performed. A 15-minute single step is used to deduce the load change process for a 4-hour prediction period. The initial load value uses the measured load of each region at the current time as the starting point of deduction. The k-step deduction formula is L_k+1=L_k+S×ΔT_k×η_new×F_carbon, wherein L_k is the k-step load value (kW), S is the load sensitivity factor (kW / ℃), ΔT_k is the k-step predicted temperature change (℃), η_new is the corrected energy-carbon response coefficient (kW / kgCO2), and F_carbon is the carbon emission constraint factor (kgCO2), reflecting the load adjustment amount under carbon emission restriction. The temperature change prediction is based on the historical temperature trend and external environmental factors, and the ARIMA time series method is used for 15-minute resolution temperature prediction. The physical constraints of the load are considered in the deduction process, and the load value cannot be negative and cannot exceed 120% of the rated power of the device. The time-varying characteristics of the sensitivity factor are adjusted by a time correction function, with a daytime correction coefficient of 1.2 and a nighttime correction coefficient of 0.8. The propagation control of the deduction error uses error resetting at regular intervals, and the cumulative error is corrected using real-time monitoring data. The multi-step deduction result includes the load prediction sequence at each time point in the prediction period, and each time point records the load value, prediction accuracy and confidence interval.

[0060] The multi-step load deduction results are classified by region to carry out zoned load forecasting. The regional classification is based on the aggregated zones B01, B02, etc. coded in S120, and each aggregated zone is taken as an independent load forecasting unit. The load is summarized to calculate the weighted average load of all deduction points in each zone, and the weight coefficient is determined according to the area contribution of the deduction point in the zone. The weight of the boundary point is 0.5, and the weight of the center point is 1.0. The load density is calculated by dividing the total load of the zone by the area of the zone, which is used to evaluate the spatial concentration of the zone load. The load peak identification is determined by the maximum value in the deduction sequence, and the peak time and peak size are taken as the zone characteristic parameters. The load growth rate is calculated using the linear fitting slope, which reflects the change trend of the zone load. The verification of the zone load forecasting results is compared with the historical data of the same period to ensure the rationality of the prediction. The zoned load forecasting considers the constraints of equipment service radius and control boundary to ensure that the prediction results match the physical system.

[0061] The zoned load forecasting results are standardized and format-converted. The time series standardization corresponds the deduction results to the standard time axis, and the time resolution remains 15 minutes. The unit of load value is unified to kilowatt (kW), and the decimal point is retained to one place to ensure data accuracy. The abnormal value processing corrects the prediction values that exceed the reasonable range, which is defined as 30%-200% of the historical load average. The prediction accuracy evaluation is based on the historical prediction error statistics, and the mean absolute percentage error (MAPE) of each region is calculated as the accuracy index, with the target accuracy controlled within 15%. The confidence interval calculation is based on the statistical distribution of prediction errors, and the 95% confidence interval is the prediction value plus or minus 1.96 times the standard error. The spatial consistency test of load values ensures the continuity of load changes in adjacent regions and avoids unreasonable spatial gradients. The output format of each region's predicted load value is [region number, time series, load value, confidence interval, prediction accuracy].

[0062] The preliminary control sequence is formulated according to the predicted load value. The control sequence is formulated based on the deviation analysis of the predicted load value and the target load and the control strategy. The target load is determined according to the indoor comfort requirement, and the temperature control target is set to 22±2℃. The corresponding target load is obtained through historical data statistics. The load deviation is calculated by subtracting the target load from the predicted load. A positive deviation indicates that the cooling needs to be increased, and a negative deviation indicates that the cooling needs to be reduced. The control priority is sorted according to the absolute value of the deviation. The areas with a deviation exceeding 15kW are listed as the first-level control, the areas with a deviation of 5-15kW are listed as the second-level control, and the areas with a deviation less than 5kW are listed as the third-level control. The control action quantification is based on the load deviation divided by the area sensitivity factor to determine the specific control amplitude. The control timing arrangement considers the load inertia and response delay. The high inertia area starts the control 30 minutes in advance, and the low inertia area starts the control 10 minutes in advance. The control sequence adopts a step adjustment strategy, which is adjusted every 15 minutes, and the single adjustment amplitude does not exceed 10%, avoiding the influence of drastic changes on the stability of the system. The sequence time span covers the 4-hour prediction period, forming a complete control instruction sequence containing [time point, area number, control action, control amplitude]. Multi-objective optimization balances the comfort, energy saving and equipment protection in the control sequence, and adopts a weighted summation method for comprehensive evaluation, with weights of 0.5, 0.3 and 0.2 respectively. Through system analysis and multi-objective optimization design of the predicted load value, the preliminary control sequence of each area of the station building is finally formed.

[0063] In step S150, the preliminary control sequence is used to perform heat pump group power distribution to form a distribution scheme, a peak-shaving operation interval is determined according to the distribution scheme and the carbon emission step point, and the distribution scheme is adjusted according to the peak-shaving operation interval to output an optimized control instruction.

[0064] In some embodiments, the use of the preliminary control sequence to perform heat pump group power distribution to form a distribution scheme includes: establishing a heat pump power demand matrix based on the preliminary control sequence, the heat pump power demand matrix including a basic power demand, a peak power demand and a standby power demand; performing load balancing processing on the heat pump power demand matrix to generate an equalization matrix; distributing target power of each heat pump device according to the equalization matrix; and arranging the target power in time sequence to form a distribution scheme.

[0065] Based on the preliminary control sequence, a heat pump power demand matrix is established. Based on the control amplitude and control mechanism of each region in the control sequence, a matrix with dimensions of [time step x region x demand type] is established, covering a 4-hour prediction period. The number of regions corresponds to the number of aggregated partitions, and the demand types include basic, peak, and standby. The basic power demand is determined according to the daily load of the region, and the calculation formula is P_base = L_average x η_base, where P_base is the basic power demand (kW), L_average is the average load of the region (kW), η_base is the basic load coefficient, and the heat island region takes 0.8 (large thermal inertia requires stable power), and the cold island region takes 0.6 (small thermal inertia can be flexibly adjusted). The peak power demand corresponds to the maximum control amplitude in the control sequence, and the calculation formula is P_peak = L_max x η_peak, where P_peak is the peak power demand (kW), L_max is the maximum load of the region (kW), and η_peak is the peak load coefficient, with a value of 1.5 for the first-level control region, 1.3 for the second-level control region, and 1.2 for the third-level control region. The standby power demand is set according to the system safety margin, with standby power being 15%-25% of the basic power, 25% for the first-level control region, and 15% for the second and third-level control regions. The time dimension of the demand matrix reflects the power change process of the control sequence, and the spatial dimension corresponds to the geographical distribution of each aggregated partition. The demand priority is classified as first, second, and third level control in S140, with the first-level demand having the highest priority and the third-level demand having the lowest priority. The demand matrix data format is [timestamp, region number, basic demand, peak demand, standby demand, priority].

[0066] The load balancing of the heat pump power demand matrix generates an equilibrium matrix. Load balancing aims to avoid the unbalanced state of individual heat pumps being overloaded while other heat pumps being lightly loaded, while considering the demand priority to ensure the power supply of important areas. A priority-weighted load variance minimization method is adopted, and the objective function is min∑[w_i×(P_i-P_avg)²], where w_i is the priority weight of the i-th area (1.5 for the first level, 1.2 for the second level, and 1.0 for the third level), P_i is the power of the i-th heat pump, and P_avg is the average power of the group. The load transfer rule preferentially transfers power from low-priority areas to high-priority areas, and the transfer amount does not exceed 80% of the remaining capacity of the receiving heat pump. The balancing constraints include: the power of a single heat pump does not exceed the rated power, the power transfer does not violate the service area boundary, the total power remains unchanged after transfer, and the power demand of high-priority areas must be met. The balancing process is carried out in groups, and the load balancing is first performed within each group, and then a secondary balancing is performed between groups. The equilibrium matrix maintains the original time and space dimensions, and adds balancing identification and transfer record fields. The iterative balancing process is performed for a maximum of 5 rounds, and the load distribution is recalculated after each balancing. When the load rate of all heat pumps is within the range of 40%-90%, it is considered to be well balanced.

[0067] According to the equilibrium matrix, the target power of each heat pump device is allocated. Through an efficiency-first allocation method, the performance of the device and the load balancing requirement are considered comprehensively. The allocation weight calculation formula is W_i=(eff_i×LR_i) / ∑(eff_j×LR_j), where W_i is the allocation weight of the i-th heat pump, eff_i is the operating efficiency of the i-th heat pump under the current load rate (obtained by looking up the device efficiency curve table), LR_i is the load rate after balancing (between 0.4 and 0.9), and ∑ represents the summation of all heat pumps j in the group. The target power calculation formula is P_target_i=P_total×W_i, where P_target_i is the target power of the i-th heat pump (kW), and P_total is the total power demand of the group (kW). The power allocation considers the best efficiency interval of the device, and preferentially runs the heat pump within the range of 60%-85% of the rated power, which has the highest COP value. The minimum power constraint ensures that the heat pump operates above the minimum stable power (30% of the rated power), avoiding frequent start-stop. The maximum power constraint prevents the device from running overloaded, and the upper limit of the power is set to 95% of the rated power. The power adjustment gradient controls the single power change to be no more than 20% of the rated power, reducing the impact on the device. The time series of the target power forms the power command sequence of each heat pump, which includes power value, operating state, adjustment amplitude, and other information at each time point.

[0068] Target power is sequenced to form a distribution scheme. Based on the timeline of the control sequence, the target power of each heat pump is organized in chronological order into a distribution scheme. The data structure of the distribution scheme is [time step, heat pump number, target power, operating state, adjustment direction], covering a 4-hour prediction period. The time synchronization mechanism ensures that the power adjustment actions of all heat pumps are executed at the same time node, avoiding coordination problems caused by time deviation. The power ramping control limits the power change rate between adjacent time steps, with a ramping rate of 5% per minute of rated power, based on the recommended value of the equipment manufacturer. The start-stop timing arrangement takes into account the preheating and precooling time of the equipment. It takes 5 minutes to preheat before starting in cooling mode, and 3 minutes to precool before starting in heating mode. The start-stop stagger arrangement within the group avoids simultaneous start-up causing power grid impact, with an interval of 2-5 minutes, the specific interval being determined according to the transformer capacity. The timing arrangement of the standby heat pump is used as an emergency plan, which automatically switches to the standby equipment when the main heat pump fails, with a switching time of not more than 2 minutes. The distribution scheme is verified by simulation to check the power balance, equipment constraints and timing rationality, ensuring the executability of the scheme.

[0069] The peak-avoiding operation interval is determined with reference to the distribution scheme and the carbon emission step point. The step point time statistics show that the carbon emission steps are mainly concentrated in the 8:00-10:00 and 14:00-16:00 periods, corresponding to the superposition of peak passenger flow and peak electricity consumption. The peak-avoiding strategy shifts the high-power operation period away from the high-emission step period, reducing the carbon emission intensity of the system. The peak-avoiding interval is identified by step point density analysis, and the period with more than 3 step points or cumulative step intensity exceeding 100 kgCO2 per hour is defined as the peak interval. The operation interval adjustment uses load shifting method to transfer 30%-40% of the load in the peak period to the low peak period, and the transfer ratio is determined according to the building thermal inertia. The transfer constraints include: the temperature deviation after transfer is not more than 1℃, the single transfer time is not more than 1 hour (to ensure that multiple transfers are completed within the 4-hour prediction period), and the transfer power is not more than 50% of the original power. The precooling and preheating strategy takes advantage of the building thermal inertia to precool or preheat in the low peak period, and the precooling / preheating time is determined according to the building time constant, generally 30-60 minutes. The peak-avoiding effect is evaluated by comparing the carbon emission intensity, with a target carbon emission intensity reduction of 10%-15%, equivalent to a daily reduction of 200-300 kgCO2. The multi-scenario peak-avoiding scheme considers three typical scenarios: weekdays, weekends, and holidays, each scenario corresponding to different step point distribution and peak-avoiding strategies.

[0070] The optimization control instruction is output according to the peak-shaving interval adjustment distribution scheme. The original distribution scheme is adjusted in time sequence and power based on the peak-shaving interval. The high-power operation in the peak-shaving interval is adjusted in advance or delayed to the non-peak-shaving period. The adjustment range is determined according to the time window allowed by the thermal inertia. The thermal island region can be advanced by 60 minutes, and the cold island region can be advanced by 30 minutes. The power adjustment adopts the peak clipping and valley filling strategy, reduces the peak power by 15%-25%, increases the valley power by 10%-20%, and keeps the total energy supply basically unchanged within 4 hours. The adjustment constraints include the device start-stop times limit (single device start-stop not more than 3 times within 4 hours), continuous operation time limit (minimum continuous operation time of 30 minutes), power change gradient limit (maintain 5% / minute), etc. The optimization objective function comprehensively considers the three targets of carbon emission minimization, energy consumption minimization and comfort guarantee. The objective function is min[0.4xC_carbon+0.3xC_energy+0.3xC_comfort], wherein C_carbon is the carbon emission cost, C_energy is the energy consumption cost, and C_comfort is the comfort deviation cost. The multi-scenario optimization selects the corresponding optimization parameters according to the scene type determined in paragraph 5. The weekday scene emphasizes the peak-shaving effect, the weekend scene emphasizes the comfort, and the holiday scene balances the indicators. The control instruction format includes instruction number, execution time, target device, power setting, operation mode, priority, scene identifier and other fields. The instruction issuing adopts a hierarchical issuing mechanism. The emergency instruction is issued immediately, the regular instruction is issued according to the plan, and the predicted instruction is issued 15 minutes in advance. The instruction execution feedback mechanism tracks the instruction execution state and effect, and provides data support for the next round of optimization. The conflict processing mechanism solves the contradiction between multiple instructions, and schedules according to the priority and time sequence. The instruction validity check ensures that all instructions are within the device capacity and meet the safety specifications, and finally outputs the optimization control instruction considering energy saving and emission reduction and comfort requirements.

[0071] In step S160, the thermal buffer region is set according to the thermal zoning map, the thermal buffer region is connected to the heat island and the cold island to form a heat transfer network, and a coordinated control scheme is generated according to the heat transfer network and the optimization control instruction.

[0072] Specifically, the thermal buffer area is set according to the thermal zoning map. The buffer area setting criteria are: temperature gradient in the range of 0.5-2.0℃ / m, distance from the thermal island or cold island boundary of 5-15 meters, and area not less than 50m². The buffer area types are divided into three types: thermal conduction type, convection type and radiation type. The thermal conduction type buffer area is located at the direct contact surface of adjacent thermal island and cold island, the convection type buffer area is located on the ventilation channel and air flow path, and the radiation type buffer area is located near the glass curtain wall and skylight. Different types will adopt differentiated site configuration strategies. The buffer capacity calculation is based on the heat capacity characteristics of the area, and the heat capacity C=ρ×V×c, where C is the heat capacity (J / ℃), ρ is the air density (kg / m³), V is the buffer area volume (m³), and c is the specific heat capacity (J / (kg·℃)). The buffer response time is determined according to the geometric characteristics and ventilation conditions of the area. The response time of a narrow and long buffer area is 5-10 minutes, and the response time of an open buffer area is 15-30 minutes. The buffer area code uses the format of BUF+serial number, such as BUF01, BUF02, etc., for easy management and control. The buffer area attribute record includes information such as area code, area type, geometric parameters, heat capacity, response time, and connected thermal island and cold island number.

[0073] In some embodiments, the heat transfer network formed by connecting the thermal island and the cold island through the thermal buffer area includes: establishing heat relay sites in the thermal buffer area; using the heat relay sites to perform segmented heat transfer to generate a transfer chain; optimizing the transfer chain to form an optimized transmission path; and organizing the optimized transmission path to form a heat transfer network.

[0074] Hot relay stations are established in the thermal buffer area. Relay stations are set up at key locations in the buffer area, and the main function is to temporarily store, adjust and transfer heat. The station configuration is designed differently according to the type of buffer area. For the heat conduction type buffer area, the station is mainly configured with temperature sensors and heat conduction enhancement devices, and the station spacing is 10-15 meters. For the convection type buffer area, the station is mainly configured with fans and flow regulating valves, and the station spacing is 20-25 meters. For the radiation type buffer area, the station is mainly configured with sunshade devices and radiation regulators, and the station spacing is 15-20 meters. The site selection of the station considers the convenience of heat transfer and the effectiveness of control, and the geometric center position of the buffer area is preferred. Each buffer area is provided with 1-3 relay stations, and the number of stations is determined according to the area of the buffer area: less than 100m² set 1 station, 100-300m² set 2 stations, and more than 300m² set 3 stations. The heat capacity of the relay station is configured according to 20%-40% of the total heat capacity of the buffer area, and the specific proportion is determined according to the type of buffer area: 40% for heat conduction type (need larger heat capacity buffer), 30% for convection type (rely on flow heat transfer), and 20% for radiation type (mainly adjust radiation). The station code uses the format of buffer area code + station serial number, such as BUF01-S1 representing the first station in the BUF01 area. The station control parameters include target temperature, transmission rate, start-stop conditions, etc., which are set according to the characteristics of the heat island and cold island connected and the type of buffer area.

[0075] The transmission chain is generated by using the relay stations to transfer heat in sections. The heat is transferred from the heat island to the cold island through a series of relay stations in a station-by-station relay manner. The starting point of the transmission chain is the relay station at the boundary of the heat island, and the ending point is the relay station at the boundary of the cold island, and in between passes through several relay stations of the buffer area. The transmission rate control is driven based on the temperature difference, and the transmission rate q = h x A x ΔT, where q is the transmission rate (W), h is the heat transfer coefficient (W / (m²·℃)), A is the heat transfer area (m²), and ΔT is the temperature difference (℃). The chain length is determined according to the distance between the heat island and the cold island, and on average one relay station is set every 20-30 meters to ensure the continuity of the transmission process. The transmission timing arrangement considers the response time and heat capacity of each station, and the heat output of the previous station and the heat input of the next station are kept in time synchronization. The chain classification includes direct transmission chain (heat island directly connected to cold island, station number ≤2), indirect transmission chain (through multiple transfer stations, station number 3-5) and ring transmission chain (forming a closed loop heat circulation, station number ≥4), and different types of chains will adopt differentiated optimization strategies. The transmission efficiency is evaluated by the heat loss rate, and the loss rate is calculated by the difference between the input heat and the output heat divided by the input heat. The direct chain loss rate target is <10%, the indirect chain is <20%, and the ring chain is <15%.

[0076] The path optimization of the transfer chain forms an optimized transmission path. The path optimization aims to minimize the transfer time and heat loss while maximizing the transfer efficiency. The optimization strategy is implemented differently according to the chain type. The shortest path algorithm is used, and the optimization objective is to minimize the path length. The indirect transfer chain uses multi-objective optimization to balance the path length and heat loss. The ring-shaped transfer chain uses a flow balancing algorithm to ensure uniform flow in each section of the loop. The objective function of the optimization algorithm is min(α×L+β×Q_loss), where L is the path length, Q_loss is the heat loss, and the weight coefficients α and β are set according to the chain type: α=0.7, β=0.3 for direct chains, α=0.5, β=0.5 for indirect chains, and α=0.3, β=0.7 for ring-shaped chains. Path selection constraints include: the total path length does not exceed 100 meters (based on the critical distance of the sharp decline in heat transfer efficiency), the number of relay sites does not exceed 5 (based on the control complexity limit), and the path does not pass through the device prohibited area. Multi-path comparison analyzes different transfer paths between the same heat island and cold island pair, and selects the path with the highest comprehensive score as the optimized path. The data structure of the optimized path includes path number, path type, start and end point coordinates, site sequence, path length, transfer time, heat loss rate, and other parameters.

[0077] The optimized transmission path is organized to form a heat transfer network. Graph theory structure is used to represent the nodes as relay sites and the edges as optimized transmission paths. The edge weight is the transmission cost (considering energy consumption and time). The network topology considers the geometric layout and functional partition of the station building, forming a hierarchical transmission structure: the backbone layer carries the main heat transfer (direct transfer chain), the distribution layer realizes regional heat allocation (indirect transfer chain), and the circulation layer maintains local thermal balance (ring-shaped transfer chain). Network connectivity ensures that there is at least one reachable path between any two heat islands and cold islands, and the connectivity index is required to reach more than 95%. Network load balancing avoids overloading some paths while others are idle. Load distribution uses a flow balancing algorithm, with the goal of minimizing the standard deviation of each path load rate. Network capacity planning determines the design capacity of each path based on historical heat transfer demand, with a capacity margin of 20-30% to handle sudden demand. Network control strategies include centralized control and distributed control modes. Centralized control is uniformly scheduled by the central controller, and distributed control is coordinated by each relay site.

[0078] A collaborative control scheme is generated according to the heat transfer network and the optimized control instruction. The power adjustment of the heat pump equipment is combined with the path scheduling of the heat transfer network to maximize the overall energy efficiency. Time coordination ensures that the heat pump control action matches the timing of heat transfer. The heat pump is started 15-30 minutes in advance to prepare for heat transfer. The specific advance time is determined according to the response time of the buffer area. Space coordination corresponds the service area of the heat pump to the coverage of the transmission network to avoid control blind spots and overlapping areas. Power coordination adjusts the output power of the heat pump according to the transmission capacity of the transmission network. The path with strong transmission capacity (direct transmission chain) corresponds to higher heat pump power. Transmission coordination adjusts the path selection and flow distribution of the transmission network according to the operating state of the heat pump. The heat pump running stably corresponds to the main transmission path. The collaborative strategy includes three modes: predictive collaboration (based on load prediction to adjust in advance), responsive collaboration (based on real-time feedback to respond quickly), and adaptive collaboration (based on historical data to learn and optimize). The data structure of the collaborative control scheme includes control schedule, device power distribution, transmission path arrangement, and coordination strategy configuration. Through the deep integration of the heat transfer network and the optimized control instruction, precise collaborative control of the station building thermal environment is realized.

[0079] In order to perform the high-speed rail station multi-heat pump system load prediction and carbon optimization control method corresponding to the above-mentioned method embodiment, the corresponding functions and technical effects are realized. Referring to Figure 2 , Figure 2 The structure block diagram of a high-speed rail station multi-heat pump system load prediction and carbon optimization control system 200 provided by an embodiment of the present application is shown. For ease of illustration, only the parts related to the present embodiment are shown. The high-speed rail station multi-heat pump system load prediction and carbon optimization control system 200 provided by the embodiment of the present application comprises: A data acquisition module 201 is configured to acquire temperature field distribution data of a high-speed rail station and energy consumption operation records of a multi-heat pump system, extract spatial temperature gradient features from the temperature field distribution data, identify energy consumption fluctuation patterns through the energy consumption operation records, and generate a station thermal environment benchmark graph by fusing the spatial temperature gradient features and the energy consumption fluctuation patterns. A partition analysis module 202 is configured to perform temperature clustering analysis to generate a high-temperature aggregation area and a low-temperature aggregation area by using the station thermal environment benchmark graph, identify a heat island and a cold island in the station according to the high-temperature aggregation area and the low-temperature aggregation area, construct a thermal partition map according to the distribution characteristics of the heat island and the cold island, and demarcate a differential control boundary by means of the thermal partition map. A carbon mapping module 203 is configured to obtain carbon emission data at different energy consumption levels from the energy consumption operation records, perform gravity analysis on the carbon emission data to form a carbon emission rate of change curve, identify a carbon emission step point via the carbon emission rate of change curve, and establish an energy-carbon correlation mapping table according to the carbon emission step point and the differential control boundary. The load prediction module 204 is configured to extract a load response feature based on the thermal inertia difference between the heat island and the cold island, combine the load response feature with the carbon-energy correlation mapping table to perform partitioned load prediction, generate regional predicted load values through the partitioned load prediction, and formulate a preliminary control sequence according to the predicted load values; The power allocation module 205 is configured to perform heat pump group power allocation using the preliminary control sequence to form an allocation scheme, determine a peak-avoiding operation interval by referring to the allocation scheme and the carbon emission step point, and adjust the allocation scheme to output an optimized control instruction according to the peak-avoiding operation interval. The cooperative control module 206 is configured to set a thermal buffer region according to the thermal partition map, form a heat transfer network by connecting the heat island and the cold island through the thermal buffer region, and generate a cooperative control scheme according to the heat transfer network and the optimized control instruction.

[0080] The high-speed rail station multi-heat pump system load prediction and energy-carbon optimization control system 200 described above can implement the high-speed rail station multi-heat pump system load prediction and energy-carbon optimization control method of the above method embodiments. The optional items in the above method embodiments are also applicable to this embodiment, which will not be described in detail here. The remaining content of the present embodiment can refer to the content of the above method embodiments, and will not be described in detail in this embodiment.

[0081] The purpose of the above embodiments is to exemplarily reproduce and deduce the technical solutions of the present application, and to completely describe the technical solutions, purposes and effects of the present application. The purpose is to make the public understand the disclosure of the present application more thoroughly and comprehensively, and does not limit the protection scope of the present application.

[0082] The above embodiments are not based on an exhaustive enumeration of the present application, and there can be many other unlisted embodiments. Any substitution and improvement made without violating the concept of the present application is within the protection scope of the present application.

Claims

1. A high-speed railway station building multi-heat pump system load prediction and energy-carbon optimization control method, characterized in that, The method comprises the following steps: Collecting temperature field distribution data of a high-speed railway station building and energy consumption operation records of a multi-heat pump system, extracting spatial temperature gradient characteristics according to the temperature field distribution data, identifying energy consumption fluctuation patterns through the energy consumption operation records, fusing the spatial temperature gradient characteristics and the energy consumption fluctuation patterns to generate a station building thermal environment benchmark map; Using the station building thermal environment benchmark map to perform temperature clustering analysis to generate a high-temperature aggregation area and a low-temperature aggregation area, identifying a heat island and a cold island in the station building according to the high-temperature aggregation area and the low-temperature aggregation area, constructing a thermal zoning map according to the distribution characteristics of the heat island and the cold island, and delineating a differential control boundary by means of the thermal zoning map; Obtaining carbon emission data at different energy consumption levels from the energy consumption operation records, performing gravity analysis on the carbon emission data to form a carbon emission change rate curve, identifying a carbon emission step point through the carbon emission change rate curve, and establishing an energy-carbon correlation mapping table according to the carbon emission step point and the differential control boundary; Extracting load response characteristics based on the thermal inertia difference of the heat island and the cold island, combining the load response characteristics with the energy-carbon correlation mapping table to carry out zonal load prediction, generating regional predicted load values through zonal load prediction, and formulating a preliminary control sequence according to the predicted load values; Using the preliminary control sequence to perform heat pump group power distribution to form a distribution scheme, determining a peak-shaving operation interval by referring to the distribution scheme and the carbon emission step point, and adjusting the distribution scheme to output optimized control instructions according to the peak-shaving operation interval; Setting a thermal buffer area according to the thermal zoning map, connecting the heat island and the cold island through the thermal buffer area to form a heat transfer network, and generating a collaborative control scheme according to the heat transfer network and the optimized control instructions.

2. The method of claim 1, wherein, The method of fusing the spatial temperature gradient characteristics and the energy consumption fluctuation patterns to generate a station building thermal environment benchmark map comprises the following steps: Performing temperature phase coding on the spatial temperature gradient characteristics to generate a phase feature code; Establishing an energy consumption oscillation frequency spectrum based on the energy consumption fluctuation patterns; Using the phase feature code to modulate the energy consumption oscillation frequency spectrum to form a modulated signal; Decoding and reconstructing the modulated signal to generate a station building thermal environment benchmark map.

3. The method of claim 1, wherein, The method of constructing a thermal zoning map according to the distribution characteristics of the heat island and the cold island comprises the following steps: Performing thermal polarity identification based on the distribution characteristics of the heat island and the cold island to generate a positive polarity heat zone and a negative polarity heat zone; Using the positive polarity heat zone to perform polarity neutralization processing on the negative polarity heat zone to form a balanced heat zone; Performing spatial aggregation division on the balanced heat zone to generate an aggregated zone; Performing boundary solidification on the aggregated zone to construct a thermal zoning map.

4. The method of claim 1, wherein, The method of performing gravity analysis on the carbon emission data to form a carbon emission change rate curve comprises the following steps: Performing carbon emission gravity identification on the carbon emission data to determine a carbon emission gravity center position; Obtaining a gravity center migration trajectory based on the carbon emission gravity center position; Using the gravity center migration trajectory to calculate a carbon emission dynamic offset; Forming a carbon emission change rate curve according to the change slope of the carbon emission dynamic offset.

5. The method of claim 1, wherein, The method of combining the load response characteristics with the energy-carbon correlation mapping table to carry out zonal load prediction comprises the following steps: extracting a load inertia factor and a load sensitivity factor from the load response feature; correcting the carbon-related mapping table using the load inertia factor to generate a corrected mapping table; performing multi-step load deduction based on the load sensitivity factor and the corrected mapping table; carrying out zoned load prediction by classifying the multi-step load deduction results by region.

6. The method of claim 1, wherein, the preliminary control sequence is used to form a distribution scheme for heat pump group power distribution, including: a heat pump power demand matrix is established based on the preliminary control sequence, the heat pump power demand matrix including basic power demand, peak power demand, and standby power demand; load balancing processing is performed on the heat pump power demand matrix to generate a balanced matrix; target power for each heat pump device is allocated according to the balanced matrix; the target power is arranged in time sequence to form a distribution scheme.

7. The method of claim 1, wherein, the heat buffer area is connected to the hot island and the cold island to form a heat transfer network, including: a heat relay site is established in the heat buffer area; segmented heat transfer is performed using the heat relay site to generate a transfer chain; the transfer chain is optimized to form an optimized transmission path; the optimized transmission path is organized to form a heat transfer network.

8. The method of claim 3, wherein, the positive polarity heat zone is used to perform polarity neutralization processing on the negative polarity heat zone to form a balanced heat zone, including: thermal intensity values are extracted from the positive polarity heat zone and the negative polarity heat zone; a neutralization processing weight is determined based on the thermal intensity values; thermal balance adjustment parameters are generated according to the neutralization processing weight; polarity neutralization is performed using the adjustment parameters to form a balanced heat zone.

9. The method of claim 4, wherein, the carbon emission gravity center position is used to obtain a gravity center migration trajectory, including: station house carbon emission dynamic characteristics including peak-valley emission difference, regional contribution ratio, and passenger flow influence degree are evaluated based on the carbon emission gravity center position to determine tracking accuracy requirements; monitoring point positions are recalibrated according to the tracking accuracy requirements to form a monitoring point position layout; trajectory conflict points are identified using the monitoring point position layout; a gravity center migration trajectory is obtained through the trajectory conflict points.

10. A high-speed railway station building multi-heat pump system load prediction and energy-carbon optimization control system, characterized in that, including: a data acquisition module for acquiring temperature field distribution data of a high-speed rail station and energy consumption operation records of a multi-heat pump system, extracting spatial temperature gradient features from the temperature field distribution data, identifying energy consumption fluctuation patterns from the energy consumption operation records, and fusing the spatial temperature gradient features and the energy consumption fluctuation patterns to generate a station thermal environment benchmark graph; a zoned analysis module for performing temperature clustering analysis using the station thermal environment benchmark graph to generate high-temperature aggregation zones and low-temperature aggregation zones, identifying hot islands and cold islands within the station based on the high-temperature aggregation zones and the low-temperature aggregation zones, and constructing a thermal zoning map based on the distribution characteristics of the hot islands and the cold islands to delineate differentiated control boundaries. The carbon mapping module is configured to obtain carbon emission data at different energy consumption levels from the energy consumption operation records, perform a barycentric analysis on the carbon emission data to form a carbon emission change rate curve, identify a carbon emission step point via the carbon emission change rate curve, and establish an energy-carbon correlation mapping table according to the carbon emission step point and the differentiated control boundary; The load prediction module is configured to extract load response features based on the thermal inertia difference between the heat island and the cold island, perform partitioned load prediction by combining the load response features with the energy-carbon correlation mapping table, generate regional predicted load values through the partitioned load prediction, and formulate a preliminary control sequence according to the predicted load values; The power distribution module is configured to perform heat pump group power distribution to form a distribution scheme using the preliminary control sequence, determine a peak-avoiding operation interval by referring to the distribution scheme and the carbon emission step point, and output optimized control instructions by adjusting the distribution scheme according to the peak-avoiding operation interval; The collaborative control module is configured to set a thermal buffer region according to the thermal partition map, form a heat transfer network by connecting the heat island and the cold island via the thermal buffer region, and generate a collaborative control scheme according to the heat transfer network and the optimized control instructions.

Images