Automatic waste heat recovery method

Abstract

The present application provides an automatic waste heat recovery method, including: obtaining waste heat pressure parameters, component medium and flow characteristic data, classifying the waste heat data, obtaining characteristic vectors of different types of waste heat, and storing the characteristic vectors of different types of preheating in a waste heat characteristic database; analyzing and predicting the randomness and intermittent laws of waste heat generation and the waste heat concentration area, and dynamically adjusting the cold absorption strategy or heat absorption strategy; retrieving corresponding cold absorption or heat absorption operation parameters from the waste heat characteristic database, and controlling the cold absorption or heat absorption equipment to adaptively adjust the operating state according to the parameter settings; real-time monitoring of the cold and heat storage energy during the recovery process, and analyzing the energy loss value of the waste heat recovery; constructing a photovoltaic power storage energy optimization scheduling model, solving the energy distribution strategy, dynamically adjusting the use strategy of the cold and heat storage energy, optimizing the distribution ratio and timing of cold and heat energy, and realizing coordinated optimization control of photovoltaic power storage and waste heat recovery.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an automatic waste heat recovery method. Background Art

[0002] Failure to recover waste heat from air conditioners in office environments or external air compressors results in a waste of energy resources. Furthermore, with rising energy prices, businesses face increasing pressure on operating costs, and failure to recover waste heat directly impacts their economic performance. Through waste heat recovery, previously wasted heat can be converted into usable energy, reducing heating and cooling needs, lowering energy costs, and improving business profitability. Therefore, automated waste heat recovery in office settings presents the technical challenge of accurately analyzing and processing different waste heat data based on their pressure, composition, and flow characteristics. Due to the diverse and dispersed sources of waste heat in office environments, the pressure, composition, and flow parameters of each type of waste heat vary significantly, and the waste heat content is relatively low, making it difficult for the system to uniformly absorb and remove heat. Furthermore, the random and intermittent nature of waste heat generation presents a significant challenge, necessitating real-time system response and dynamic adjustment to match the rhythm of waste heat generation. Furthermore, during the waste heat absorption and removal process, avoiding energy loss and efficiency degradation while ensuring the quality and quantity of stored heat and heat also requires overcoming technical bottlenecks. Although photovoltaic power storage can provide energy for the system, how to optimize the energy scheduling and distribution strategy of photovoltaic power storage according to the actual needs of waste heat recovery to achieve efficient operation and adaptive control of the system still needs further research. Summary of the Invention

[0003] The present invention provides an automatic waste heat recovery method, which mainly includes:

[0004] Obtain waste heat pressure parameters, component medium, and flow characteristic data, classify the waste heat data, obtain characteristic vectors of different types of waste heat, including temperature, pressure, flow, and heat capacity parameters, and store the characteristic vectors of different types of preheating in the waste heat characteristic database;

[0005] Obtain the waste heat distribution inside and outside the office environment, identify the concentrated areas of each waste heat source, collect the generation data of each waste heat source in real time, analyze and predict the randomness and intermittent patterns of waste heat generation and the waste heat concentration areas, and dynamically adjust the cooling or heat absorption strategy;

[0006] Retrieve the corresponding cooling or heating operation parameters from the waste heat characteristic database, and control the cooling or heating equipment to adaptively adjust the operating state according to the parameter settings;

[0007] Real-time monitoring of the stored cold and heat energy during the recovery process is performed, and the energy loss value of the remaining heat recovery is analyzed. If the energy loss exceeds the preset threshold, the fault diagnosis program is triggered to determine the fault type and provide the corresponding optimization control strategy;

[0008] Construct a photovoltaic power storage energy optimization scheduling model, solve the energy allocation strategy, dynamically adjust the use strategy of cold and heat storage energy, optimize the distribution ratio and timing of cold and heat energy, and realize the coordinated optimization control of photovoltaic power storage and waste heat recovery.

[0009] Furthermore, the acquisition of waste heat pressure parameters, component medium and flow characteristic data includes:

[0010] The temperature parameters, pressure characteristics, flow rate data and heat capacity properties of the waste heat recovery system are obtained, and the parameter data are checked for a numerical range. If the parameter value exceeds the preset threshold range, it is marked as abnormal data; the abnormal data is linearly interpolated or eliminated, and the normal data is normalized to obtain standardized parameter values; the waste heat medium is analyzed by gas chromatograph to identify the main chemical components and their contents, and the component data is merged with the standardized parameter value.

[0011] Furthermore, the classification of the waste heat data includes:

[0012] The K-means clustering algorithm is used to classify the merged multidimensional data to obtain feature clusters of different types of waste heat, and the centroid coordinates of each feature cluster are calculated as the feature vector of the waste heat; a waste heat feature database table is established, which includes the waste heat type, temperature vector, pressure vector, flow vector, heat capacity vector and component vector, and the feature vectors of different types of waste heat are inserted into the waste heat feature database table for storage.

[0013] Furthermore, obtaining the waste heat distribution inside and outside the office environment includes:

[0014] Obtaining an ambient temperature distribution map, which is obtained by acquiring real-time data from multiple temperature sensors around the air compressor exhaust port outside the office environment and the air conditioner condenser inside the office environment, and estimating the waste heat distribution inside and outside the office environment using a Kriging interpolation algorithm based on the air compressor exhaust temperature and exhaust flow rate and the air conditioner condenser exhaust temperature and exhaust flow rate parameters;

[0015] A K-means clustering algorithm is executed according to the ambient temperature distribution map to obtain an abnormal temperature area as a waste heat source concentration area; a sensor network is arranged around the waste heat source concentration area, and the sensor network includes a temperature sensor, a flow sensor and a pressure sensor.

[0016] Furthermore, the analysis and prediction of the randomness and intermittent patterns of waste heat generation and waste heat concentration areas include:

[0017] Collect waste heat source data acquired by the sensor network, the waste heat source data including temperature data, flow data and pressure data; perform time series analysis on the waste heat source data, the time series analysis including data preprocessing, trend analysis and seasonal analysis; use the autoregressive moving average model to process the waste heat source data to obtain characteristic curves and prediction models for each waste heat source.

[0018] Furthermore, the dynamic adjustment of the cold absorption strategy or the heat absorption strategy includes:

[0019] The optimal cooling or heat absorption strategy is calculated based on the characteristic curve and the prediction model, and the optimal cooling or heat absorption strategy is obtained through a linear programming method; if the waste heat generation of the prediction model is greater than a preset threshold, the cooling device is started and the cooling power is adjusted; if the waste heat generation of the prediction model is less than the preset threshold, the heat absorption device is started and the heating power is adjusted.

[0020] Furthermore, the step of retrieving corresponding cold absorption or heat absorption operation parameters from the residual heat characteristic database includes:

[0021] Based on the acquired ambient temperature and humidity data, the waste heat feature record with the highest matching degree is retrieved from a waste heat feature database, wherein the waste heat feature database adopts a B-tree index structure. Based on the waste heat feature record, initial cooling or heat absorption operating parameters are obtained, wherein the initial cooling or heat absorption operating parameters include a heat pump speed range and an expansion valve opening range. The initial cooling or heat absorption operating parameters are optimized using a fuzzy logic controller, wherein input variables of the fuzzy logic controller include a temperature deviation and a humidity deviation.

[0022] Furthermore, the control of the cold or heat absorbing device to adaptively adjust the operating state according to parameter settings includes:

[0023] The output frequency of the heat pump inverter is adjusted in real time by a PID controller, which uses the Ziegler-Nichols method for PID parameter self-tuning; the temperature, pressure, and flow data of the condenser and evaporator are obtained, and the temperature, pressure, and flow data are preprocessed using a sliding average filter; based on the preprocessed temperature, pressure, and flow data, the heat pump speed and expansion valve opening are fine-tuned in real time using an adaptive neural fuzzy inference system, which continuously optimizes inference rules through online learning.

[0024] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0025] This invention discloses an automatic waste heat recovery method. By deploying infrared thermal imagers and sensor networks, real-time data such as temperature, pressure, and flow rate of waste heat sources inside and outside the office area is acquired. The distribution characteristics and generation patterns of waste heat are analyzed, and the cooling or heat absorption strategy is dynamically adjusted. Parameters in a waste heat characteristic database are combined to control the operating status of the cooling or heat absorption equipment, enabling adaptive adjustment of the heat pump speed and expansion valve opening. Simultaneously, energy metering devices are deployed in the waste heat transmission pipeline network to monitor the recovery process of cold and heat storage energy in real time. If energy loss exceeds a preset threshold, a fault diagnosis program is triggered and the control strategy is optimized. Furthermore, photovoltaic cell power generation data and the battery state of charge are acquired. Combined with the energy loss of waste heat recovery, a photovoltaic energy storage optimization scheduling model is constructed. The utilization strategy of cold and heat storage energy is dynamically adjusted, achieving coordinated optimization control of photovoltaic energy storage and waste heat recovery. By continuously monitoring system operating data, potential efficiency declines or failure risks are promptly identified, ensuring efficient system operation. Through the coordinated optimization of waste heat recovery and photovoltaic energy storage, this invention achieves efficient energy management in the office environment, reduces energy loss, and improves the overall energy efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The figure is a flow chart of an automatic waste heat recovery method of the present invention.

[0027] Figure 2 Schematic diagram of an automatic waste heat recovery method of the present invention.

[0028] Figure 3 This is another schematic diagram of an automatic waste heat recovery method of the present invention. DETAILED DESCRIPTION

[0029] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments derived by those skilled in the art based on the embodiments in this specification without creative effort shall fall within the scope of protection of this specification.

[0030] like Figure 1-3 In this embodiment, an automatic waste heat recovery method may specifically include:

[0031] Step S101, obtain waste heat pressure parameters, component medium and flow characteristic data, classify the waste heat data, obtain characteristic vectors of different types of waste heat, including temperature, pressure, flow, and heat capacity parameters, and store the characteristic vectors of different types of preheating in the waste heat characteristic database.

[0032] Obtain the temperature parameters, pressure characteristics, flow data and heat capacity properties of the waste heat recovery system, perform a numerical range check on the parameter data, and mark it as abnormal data if the parameter value exceeds the preset threshold range; perform linear interpolation processing or elimination on the abnormal data, and normalize the normal data to obtain standardized parameter values; use a gas chromatograph to perform component analysis on the waste heat medium, identify the main chemical components and their contents, and merge the component data with the standardized parameter values; use the K-means clustering algorithm to classify the merged multidimensional data, obtain characteristic clusters of different types of waste heat, and calculate the centroid coordinates of each characteristic cluster as the characteristic vector of the waste heat; establish the waste heat recovery system. A thermal characteristic database table includes waste heat type, temperature vector, pressure vector, flow vector, heat capacity vector and component vector, and characteristic vectors of different types of waste heat are inserted into the waste heat characteristic database table for storage; the stored characteristic vectors are read from the waste heat characteristic database table, and principal component analysis is performed by calculating the covariance matrix, eigenvalues ​​and eigenvectors, and the first few eigenvectors whose cumulative contribution rate exceeds a preset threshold are selected as main characteristic components; a multivariate linear regression model is established based on the main characteristic components, and the main characteristic components are used as independent variables and the waste heat utilization efficiency is used as the dependent variable. The regression coefficient is calculated to obtain the multivariate linear regression model of the waste heat characteristics.

[0033] Specifically, temperature parameters, pressure characteristics, flow rate data, and heat capacity properties are obtained from the waste heat recovery system. The obtained parameter data is checked for numerical ranges. If the parameter value exceeds the preset threshold range, it is marked as abnormal data. The abnormal data is linearly interpolated or eliminated, and the normal data is normalized to obtain standardized parameter values. The waste heat medium is analyzed for composition using a gas chromatograph to identify the main chemical components and their contents, and the composition data is merged with the standardized parameter values. The merged multidimensional data is classified using the K-means clustering algorithm to obtain characteristic clusters of different types of waste heat. The centroid coordinates of each characteristic cluster are calculated as the characteristic vector of that type of waste heat. The centroid coordinates are obtained by calculating the average value of all data points in the cluster. A waste heat characteristic database table is established, containing fields such as waste heat type, temperature vector, pressure vector, flow rate vector, heat capacity vector, and component vector. The characteristic vectors of different types of waste heat are inserted into the database table for storage. The stored characteristic vectors are read from the database, and principal component analysis is performed by calculating the covariance matrix, eigenvalues, and eigenvectors. The first few eigenvectors whose cumulative contribution rate exceeds the preset threshold are selected as the main characteristic components. A multivariate linear regression model was established based on the main characteristic components, using the main characteristic components as independent variables and waste heat utilization efficiency as the dependent variable. The regression coefficients were calculated to obtain a mathematical model of waste heat characteristics, which was used for subsequent waste heat classification and matching. Temperature, pressure, flow rate, and heat capacity data were collected from the waste heat recovery system. For example, the temperature range was 0-1000°C, the pressure range was 0-10 MPa, the flow rate range was 0-1000 m³ / h, and the heat capacity range was 0-5 kJ / (kg·K). Anomalies were detected in the collected data, with a temperature threshold set at 950°C. Data points exceeding this threshold were corrected using linear interpolation. Normal data were normalized to the range of 0-1. Gas chromatography was used to analyze the composition of the waste heat medium, identifying the main components such as CO₂, H₂O, and N₂, and recording the percentages of each component. The composition data and normalization parameters were combined to form a multidimensional feature vector. The feature vectors were classified using the K-means clustering algorithm, with the number of clusters set to 5 and 50 iterations to achieve stable clustering results. Calculate the centroid coordinates of each cluster. For example, the centroid coordinates of cluster 1 (0.3, 0.5, 0.7, 0.4, 0.6) represent the eigenvector of this type of waste heat. Create a waste heat feature table in the database, containing fields such as ID, type, temperature, pressure, flow rate, heat capacity, CO2 content, H2O content, and N2 content. Store the calculated eigenvectors in the database, such as inserting a record (1, 'high-temperature waste heat', 0.3, 0.5, 0.7, 0.4, 0.6, 0.3, 0.1). Read all eigenvectors from the database and construct a sample matrix. Calculate the sample covariance matrix and solve for the eigenvalues ​​and eigenvectors. Select the top three eigenvectors with a cumulative contribution rate of 95% as the principal components. Establish a multivariate linear regression model based on the principal components, with the three principal component scores as the independent variables and the waste heat utilization efficiency as the dependent variable.The regression coefficients were estimated using the least squares method, resulting in the model equation: y = 0.72x1 + 0.56x2 + 0.33x3 + 0.15, where y represents the waste heat utilization efficiency, and x1, x2, and x3 represent the scores of the three principal components. This model can be used to predict the utilization efficiency of different types of waste heat and assist in waste heat matching and classification.

[0034] In step S102, infrared thermal imagers are deployed inside and outside the office area to obtain the waste heat distribution inside and outside the office environment and identify the concentrated areas of each waste heat source; and a number of sensors are installed to form a sensor network to collect the generation data of each waste heat source in real time. The generation data of each waste heat source includes the temperature, flow rate, and pressure of the waste heat source. Based on the generation data of each waste heat source, the randomness and intermittent pattern of waste heat generation and the waste heat concentration area are analyzed and predicted, and the cooling absorption strategy or the heat absorption strategy is dynamically adjusted.

[0035] Acquire an ambient temperature distribution map, which is obtained by acquiring real-time data from multiple temperature sensors around the air compressor exhaust port outside the office environment and the air conditioner condenser inside the office environment, and using the Kriging interpolation algorithm to estimate the waste heat distribution inside and outside the office environment based on the air compressor exhaust temperature, exhaust flow rate and the air conditioner condenser exhaust temperature and exhaust flow rate parameters; execute the K-means clustering algorithm based on the ambient temperature distribution map to obtain temperature anomaly areas as waste heat source concentration areas; arrange a sensor network around the waste heat source concentration area, the sensor network including temperature sensors, flow sensors and pressure sensors; collect waste heat source data acquired by the sensor network, the waste heat source data including temperature data, flow data and pressure data; perform time series analysis on the waste heat source data Analysis, the time series analysis includes data preprocessing, trend analysis and seasonal analysis; the waste heat source data is processed using an autoregressive moving average model to obtain a characteristic curve and a prediction model for each waste heat source; the optimal cooling or heat absorption strategy is calculated based on the characteristic curve and the prediction model, and the optimal cooling or heat absorption strategy is obtained through a linear programming method; if the waste heat generation of the prediction model is greater than a preset threshold, the cooling device is started and the cooling power is adjusted; if the waste heat generation of the prediction model is less than the preset threshold, the heat absorption device is started and the heating power is adjusted; the operating effect of the cooling device or the heat absorption device is monitored, and the operating effect is collected through a sensor network; the device operating parameters are adjusted according to the deviation between the operating effect and the expected effect, and the device operating parameters include cooling power or heating power.

[0036] Specifically, infrared thermal imagers are installed inside and outside office areas, such as at air conditioners inside the office area or air compressors outside the office area. Thermal imaging technology is used to capture ambient temperature distribution maps. The temperature distribution maps are then processed using the K-means clustering algorithm. Areas with temperatures 20% above the ambient average are identified as abnormally high, representing concentrated waste heat sources. The spatial coordinates of each waste heat source are then recorded. Temperature, flow, and pressure sensors are placed around the identified waste heat source areas. A sensor network is constructed, with a data acquisition frequency set to once per minute. Real-time temperature, flow, and pressure data from each waste heat source are collected and transmitted to a data processing center via a wireless transmission module. Data integrity checks and sensor self-test procedures are implemented to ensure reliable data transmission. Time series analysis of the waste heat source data collected by the sensor network is performed, including data preprocessing, trend analysis, and seasonality analysis. An autoregressive moving average model is used to calculate the periodicity and volatility of each waste heat source data, with a lag period of 12 and a moving average period of 3. The randomness and intermittent nature of waste heat generation are determined, resulting in characteristic curves and prediction models for each waste heat source. Based on the waste heat source characteristic curve and prediction model, combined with current ambient temperature and humidity data, a linear programming method is used to calculate the optimal cooling or heating absorption strategy. If the predicted waste heat generation exceeds a preset threshold, the cooling device is activated and the cooling power is adjusted. If the predicted waste heat generation is less than the preset threshold, the heating device is activated and the heating power is adjusted. After the strategy is implemented, the actual effect is continuously monitored, and the device operating parameters are adjusted in real time based on deviations, forming a dynamic closed-loop control. An infrared thermal imager with a resolution of 640x480 pixels is installed in the office area, capturing thermal imaging data every 10 minutes to generate a temperature distribution map. The temperature distribution map is processed using the K-means clustering algorithm, with the number of clusters set to K=5, and temperature values ​​used as the clustering feature. Areas that are 20% higher than the average ambient temperature are identified as temperature anomalies. For example, if the average ambient temperature is 25°C, areas above 30°C are marked as concentrated waste heat sources. Three temperature sensors, one flow sensor, and one pressure sensor are placed within a 1-meter radius of each identified waste heat source. The sensors collect data every minute and transmit it to the data processing center via the ZigBee wireless network. A data integrity check mechanism is implemented to automatically request retransmission if a received data packet is incomplete. The sensors perform hourly self-tests to check their battery level and communication status. Time series analysis of the collected data begins with data preprocessing, including outlier removal and missing value interpolation. Trend analysis is then performed, using a moving average method to calculate a 7-day sliding mean. Seasonality analysis is performed on the detrended data, using the autocorrelation function to identify periodic patterns. An ARIMA model is used for time series forecasting, with the autoregressive term p = 2, the differencing term d = 1, and the moving average term q = 2. Based on the forecast results and current environmental conditions, the simplex method is used to solve a linear programming problem and calculate the optimal equipment operating parameters.If the predicted waste heat generation exceeds 5kW / h, the cooling device is activated, and the cooling power is dynamically adjusted between 0-10kW. If it is less than 5kW / h, the heat absorption device is activated, and the heating power is adjusted between 0-5kW. The effectiveness of the strategy execution is evaluated every 15 minutes, and the root mean square error (RMS) between the actual temperature and the target temperature is calculated. If the error exceeds 2°C, parameter re-optimization is triggered.

[0037] The residual heat distribution in each area is counted and classified into concentrated residual heat distribution areas. The concentrated residual heat distribution areas are counted to obtain statistical classification results of the concentrated residual heat distribution areas. The statistical classification results are used as the installation location strategy of the sensor network. Based on the residual heat collection results of the sensors, a data collection strategy for residual heat after the air conditioner is turned off in the office area is generated.

[0038] A multi-angle thermal imager is used to scan the temperature of the grid units, and a complete thermal map is synthesized through an image stitching algorithm to obtain a residual heat distribution thermal map; a cluster analysis is performed based on the residual heat distribution thermal map, and a hierarchical clustering algorithm is used to automatically determine the optimal number of clusters, and grid units with similar temperatures are classified into one category; if the number of grid units with similar temperatures is greater than a preset threshold, the temperature variance is less than a preset temperature threshold, the average temperature is higher than a preset difference from the ambient temperature, and the duration exceeds a preset time threshold, the grid units classified into one category are determined to be a highly concentrated area; the geometric center position of the highly concentrated area is obtained, and a temperature sensor is installed at the geometric center position, and the temperature sensor is integrated with the air-conditioning control system; the sampling frequency is dynamically adjusted according to the data change trend, and the collected data is compressed using a data compression algorithm, and the compressed data is transmitted to the central processing unit via a low-power wide area network.

[0039] Specifically, based on a predefined coordinate system for the office area floor plan, the entire area was divided into several grid cells, each measuring 5 meters by 5 meters. A multi-angle thermal imager was used to scan the temperature of each grid cell. An image stitching algorithm was used to synthesize a complete thermal map, eliminating the effects of obstructions. The average temperature of each grid cell was obtained to generate a residual heat distribution thermogram. Cluster analysis was performed on the residual heat distribution thermogram, and a hierarchical clustering algorithm was used to automatically determine the optimal number of clusters. Grid cells with similar temperatures were grouped together to obtain a preliminary classification of areas of concentrated residual heat. Statistical analysis of the preliminary classification results was performed, calculating the number of grid cells, average temperature, temperature variance, and duration for each category. The degree of residual heat concentration was determined based on a comprehensive score. Areas with greater than 10 grid cells, a temperature variance less than 2°C, an average temperature greater than 5°C above the ambient temperature, and a duration exceeding 30 minutes were considered highly concentrated. Based on the spatial distribution of highly concentrated areas, a sensor network installation strategy was developed. Temperature sensors were installed at the geometric center of each highly concentrated area. These sensors were integrated with the air conditioning control system to obtain real-time air conditioning operating status information. A basic sampling interval was set at 5 minutes. When an air conditioning shutdown signal was detected or the temperature rise rate exceeded a preset threshold, the sampling interval was automatically adjusted to 1 minute. Residual heat data was continuously recorded, and the sampling frequency was dynamically adjusted based on data trends. A data compression algorithm was used to reduce storage space usage, and data was transmitted to the central processing unit via a low-power wide area network. A 1,000-square-meter office area was divided into 400 5-meter by 5-meter grid cells using a predefined coordinate system. A 360-degree rotating thermal imager with a resolution of 640 by 480 pixels was used, scanning the entire space every 15 seconds. Image stitching algorithms, such as SIFT feature matching and RANSAC, were used to synthesize multi-angle thermal imaging data into a complete thermal map, addressing obstructions such as pillars and partitions. The average temperature was calculated for each grid cell, generating a 40 by 40 temperature matrix. Hierarchical clustering was performed using Ward's minimum variance method, with a cluster distance threshold of 2°C. The optimal number of clusters was automatically determined. The clustering results were statistically analyzed, and the number of grid cells, average temperature, temperature variance, and duration for each cluster were calculated. A comprehensive scoring formula was established: Score = number of grid cells × 0.3 + (average temperature - ambient temperature) × 0.4 + (10 - temperature variance) × 0.2 + duration × 0.1. Clusters with a score greater than 15 were identified as highly concentrated areas. A temperature sensor with a measurement range of -20°C to 80°C and an accuracy of ±0.5°C was installed at the geometric center of each highly concentrated area. The sensor communicated with the air conditioning control system via the Modbus protocol to obtain the air conditioning on / off status. The basic sampling interval was 5 minutes. The sampling interval was adjusted to 1 minute when an air conditioning off signal was detected or the temperature rise rate exceeded 0.5°C / minute.An adaptive sampling algorithm is used. When the temperature change rate exceeds 1°C / 5 minutes, the sampling interval is halved; when it is less than 0.2°C / 5 minutes, the sampling interval is doubled, with a minimum of 30 seconds and a maximum of 10 minutes. A differential coding compression algorithm is used to compress raw data at a ratio of 5:1. Data is transmitted to the central processor via the LoRaWAN network, operating in the 915MHz frequency band at a transmission rate of 5.5kbps, enabling low-power, long-distance data transmission.

[0040] In step S103, the corresponding cooling or heat absorption operation parameters are retrieved from the waste heat characteristic database, and combined with the dynamically adjusted cooling or heat absorption strategy, the cooling or heat absorption equipment is controlled to adaptively adjust the operating state according to the parameter settings. The operating state includes dynamically optimizing the heat pump speed and the expansion valve opening.

[0041] Based on the acquired ambient temperature and humidity data, the most matching waste heat feature record is retrieved from the waste heat feature database, which utilizes a B-tree index structure. Based on the waste heat feature record, initial cooling or heat absorption operating parameters are obtained. These initial cooling or heat absorption operating parameters include the heat pump speed range and the expansion valve opening range. A fuzzy logic controller is used to optimize these initial cooling or heat absorption operating parameters. The input variables of the fuzzy logic controller include temperature deviation and humidity deviation. The output frequency of the heat pump inverter is adjusted in real time using a PID controller, which uses the Ziegler-Nichols method for PID parameter self-tuning. The temperature, pressure, and flow data of the condenser and evaporator are obtained and preprocessed using a sliding average filter. Based on the preprocessed temperature, pressure, and flow data, an adaptive neural fuzzy inference system is used to fine-tune the heat pump speed and expansion valve opening in real time. The adaptive neural fuzzy inference system continuously optimizes the inference rules through online learning.

[0042] Specifically, based on the current ambient temperature and humidity data, the waste heat feature record with the highest matching degree is retrieved from the waste heat feature database. The matching degree is calculated using cosine similarity, and the corresponding initial cooling or heat absorption operating parameters, including the heat pump speed range and the expansion valve opening range, are extracted. The waste heat feature database uses a B-tree index structure to optimize query efficiency. The initial operating parameters are integrated with the dynamically adjusted cooling or heat absorption strategy, and the parameters are optimized using a fuzzy logic controller. The input variables include temperature deviation and humidity deviation. The rule base is set based on expert experience to obtain the heat pump speed target value and expansion valve opening target value that are suitable for the current environment. The output frequency of the heat pump inverter is adjusted in real time by a PID controller, and the PID parameters are self-tuned using the Ziegler-Nichols method to gradually adjust the heat pump speed to the target value. At the same time, a stepper motor is used to accurately control the expansion valve opening so that it gradually reaches the target opening value. The waste heat recovery system continuously monitors its operating status, collecting parameters such as condenser and evaporator temperature, pressure, and flow rate every 5 seconds. A sliding average filter is used for data preprocessing. An adaptive neural-fuzzy inference system (ANFIS) fine-tunes the heat pump speed and expansion valve opening in real time. The ANFIS inputs the preprocessed temperature, pressure, and flow rate data, and outputs speed and opening adjustments. Online learning continuously optimizes inference rules to ensure the waste heat recovery system maintains optimal operating conditions. In an office building air conditioning system, the current ambient temperature is 26°C and the relative humidity is 60%. The waste heat signature database uses a B-tree index structure and stores 1,000 historical records. The waste heat recovery system calculates matching scores using the cosine similarity algorithm with a threshold of 0.95. The best matching record is retrieved, with an initial heat pump speed range of 30-60 Hz and an expansion valve opening range of 40-80%. The fuzzy logic controller has five input variables and seven output variables, and its rule base contains 25 if-then rules. After optimization, the target heat pump speed was 45 Hz, and the target expansion valve opening was 60%. The PID controller was auto-tuned using the Ziegler-Nichols method, resulting in parameters Kp = 1.2, Ti = 60 seconds, and Td = 15 seconds. The waste heat recovery system collected operating data every 5 seconds, including condenser temperature, evaporator temperature, system pressure, and refrigerant flow rate. A 10-point sliding average filter was used for data preprocessing. The ANFIS system consists of 5 input nodes, 25 rule nodes, and 2 output nodes. It uses a backpropagation algorithm for online learning with a learning rate of 0.01. After 100 iterations, the ANFIS output fine-tuned the heat pump speed by +2 Hz and the expansion valve opening by -5%. The waste heat recovery system continued to optimize based on feedback, reaching a stable state after 30 minutes, maintaining the indoor temperature at 24.5 ± 0.5°C and the relative humidity at 50 ± 3%.

[0043] In step S104, an energy metering device is installed in the waste heat transmission network to monitor the cold and heat storage energy in the recovery process in real time, and analyze the energy loss value of the waste heat recovery. If it is found that the energy loss exceeds the preset threshold, the fault diagnosis program is triggered to determine the fault type and provide a corresponding optimization control strategy.

[0044] Obtain the hot and cold fluid flow and temperature data of key nodes in the waste heat transmission pipeline network, and the hot and cold fluid flow and temperature data are collected by ultrasonic flow meters and PT100 temperature sensors installed at the key nodes; calculate the energy loss value between adjacent nodes based on the hot and cold fluid flow and temperature data, and the energy loss value is calculated by the central controller based on the heat balance principle; determine whether the energy loss value exceeds the preset energy loss threshold, and the preset energy loss threshold is determined based on statistical analysis of historical data; if the energy loss value exceeds the preset energy loss threshold, trigger a fault diagnosis program, and the fault diagnosis program uses the C4.5 decision tree algorithm to determine the fault type; according to the fault type determined by the fault diagnosis program, call the corresponding control strategy from the optimization control strategy library, and the optimization control strategy library is constructed based on expert knowledge and machine learning algorithms; generate targeted control instructions, and the control instructions are used to adjust the pump speed, change the valve opening or start the backup pipeline.

[0045] Specifically, ultrasonic flowmeters and PT100 temperature sensors are installed at key nodes in the waste heat transmission pipeline network to collect real-time flow and temperature data for both hot and cold fluids. Data is sampled every 10 seconds via a data collector and transmitted to a central controller using the ModBus protocol. The instantaneous heat value at each node is calculated. Temperature sensors and level gauges are also installed on the energy storage devices to monitor changes in stored cold and hot energy. Based on the principle of heat balance, the central controller calculates energy loss between adjacent nodes using the formula Q = mc(T1 - T2), where m is the fluid mass flow rate, c is the specific heat capacity, and T1 and T2 are the adjacent node temperatures. Based on statistical analysis of historical data, a preset energy loss threshold is set. If the energy loss in a particular pipeline section exceeds the threshold, a fault diagnosis program is triggered. This fault diagnosis program utilizes the C4.5 decision tree algorithm, using information gain ratio for feature selection. It combines multi-dimensional data such as temperature, pressure, and flow rate to determine the fault type, classifying it into multiple categories, such as pipeline leakage, insulation damage, and valve failure. Based on the fault type diagnosed, the corresponding optimized control strategy library is invoked. This library is built using a combination of expert knowledge and machine learning algorithms and is continuously updated through online learning. Targeted control instructions are generated, such as adjusting pump speed, changing valve opening, or activating a backup pipeline. The optimized control strategy is then applied to the waste heat recovery system via actuators. In an industrial park waste heat recovery system, the pipeline network is 5 kilometers long and has 20 key nodes. Each node is equipped with an ultrasonic flowmeter with an accuracy of ±0.5% and a PT100 temperature sensor with an accuracy of ±0.1°C. A data logger collects data at 10-second intervals using the ModBus RTU protocol at a transmission rate of 9600 bps. The central controller uses an Intel Core i7 processor operating at 3.6 GHz and 32 GB of memory. For heat balance calculations, the specific heat capacity of water is assumed to be 4.2 kJ / (kg·K). The waste heat recovery system uses the 3σ principle to set an energy loss threshold of 5% based on the past three months of historical data. The C4.5 decision tree algorithm uses information gain ratio for feature selection, setting the minimum number of leaf node samples to 50 and the maximum tree depth to 10. Fault types include pipeline leakage, insulation damage, valve failure, sensor failure, and pump failure. The optimized control strategy library initially contains 100 expert rules and uses a random forest algorithm for online learning, with the model updated every 24 hours. The waste heat recovery system detected that the energy loss in Section A reached 7.2%, triggering fault diagnosis. The decision tree analysis results indicated insulation damage with a confidence level of 0.85. The control strategy automatically generated instructions: reduce the flow rate in this section of the pipeline by 20% and activate the backup pipeline. After receiving the instructions, the actuator adjusted the main pipeline pump frequency from 50 Hz to 40 Hz and the backup pipeline valve opening from 0% to 60%. After 30 minutes of adjustment, the energy loss in Section A was reduced to 4.8%, and the waste heat recovery system resumed normal operation.

[0046] Step S105: Acquire the real-time power generation data of the photovoltaic cell and the charge state data of the battery, and construct a photovoltaic power storage energy optimization scheduling model in combination with the energy loss value of waste heat recovery. The photovoltaic power storage energy optimization scheduling model includes factors such as photovoltaic power generation, battery capacity, and energy loss value of waste heat recovery. The energy allocation strategy is obtained by solving the photovoltaic power storage energy optimization scheduling model.

[0047] Real-time power generation data of photovoltaic cells collected by the photovoltaic inverter is obtained, and the real-time power generation data includes output voltage, output current and output power; based on the real-time power generation data, a support vector regression algorithm is used to perform short-term prediction of photovoltaic power generation within a preset time period in the future to obtain predicted power generation data; based on the predicted power generation data, battery status data and energy loss value, a multi-objective optimization function is constructed, and the multi-objective optimization function is F=w1f1+w2f2-w3f3, where f1 is the photovoltaic utilization rate, f2 is the battery utilization rate, f3 is the energy loss rate, and w1, w2, and w3 are weight coefficients; a genetic algorithm is used to solve the multi-objective optimization function to obtain a Pareto optimal solution set; from the Pareto optimal solution set, a compromise optimal solution is selected through a hierarchical analysis method to determine an energy allocation strategy; according to the energy allocation strategy, a control instruction is generated, and the control instruction includes photovoltaic power generation grid-connected power, battery charge and discharge power and waste heat recovery system operating parameters.

[0048] Specifically, real-time power generation data from photovoltaic cells, including output voltage, current, and power, is collected every five minutes through the photovoltaic inverter. Simultaneously, battery state-of-charge data, including current charge and charge / discharge power, is obtained from the battery management system. This data is transmitted to the central controller via a ModbusTCP communication gateway. A support vector regression algorithm is used to make a short-term forecast of photovoltaic power generation for the next four hours, and battery capacity trends are analyzed based on historical data. Real-time energy loss values ​​are obtained from the waste heat recovery system and used as constraints in the optimization model. Combining photovoltaic power generation data, power generation forecast results, battery status data, and capacity trends, a multi-objective optimization function F = w1f1 + w2f2 - w3*f3 is constructed, where f1 is the photovoltaic utilization rate, f2 is the battery utilization rate, and f3 is the energy loss rate. w1, w2, and w3 are weight coefficients, respectively. This establishes a photovoltaic power storage energy optimization scheduling model. A genetic algorithm was used to solve the photovoltaic (PV) energy storage optimization scheduling model, with a population size of 100, a crossover rate of 0.8, and a mutation rate of 0.1. After 500 iterations, a Pareto-optimal solution set was obtained. The analytic hierarchy process (AHP) was then used to select the optimal compromise solution as the energy allocation strategy. Based on the resulting energy allocation strategy, specific control instructions were generated, including the PV grid-connected power, battery charge and discharge power, and waste heat recovery system operating parameters. These instructions were then distributed to the corresponding execution devices via a controller, achieving optimal energy scheduling. A 1MW PV power generation system and a 500kWh lithium battery energy storage system were installed in a smart industrial park. The PV inverters collected data every five minutes, recording output voltage, current, and power. The battery management system simultaneously collected battery state of charge (SOC), including current charge and charge and discharge power. This data was transmitted to a central controller via the Modbus TCP protocol at a rate of 9600 bps. A support vector regression algorithm used seven days of historical data and weather forecast information to predict PV power generation for the next four hours, with an average prediction error within ±10%. Analysis of battery capacity trends shows that after every 100 complete charge-discharge cycles, the capacity decays by approximately 1%. The real-time energy loss of the waste heat recovery system fluctuates between 5 and 15 kW. In the multi-objective optimization function, the weight coefficients w1 = 0.5, w2 = 0.3, and w3 = 0.2. The genetic algorithm was set with a population size of 100, a crossover rate of 0.8, and a mutation rate of 0.1. After 500 iterations, 20 Pareto optimal solutions were obtained. The analytic hierarchy process was used to construct a judgment matrix. After a consistency test, the solution with the highest overall score was selected as the final energy allocation strategy. After the control instructions were generated, the PLC set the photovoltaic grid-connected power to 850 kW, the battery charging power to 50 kW, and the waste heat recovery system to 70% load, achieving optimal overall energy scheduling. The waste heat recovery system recalculates every 15 minutes and dynamically adjusts its strategy.

[0049] In step S106, according to the energy allocation strategy obtained by the solution, the strategy for using cold and heat storage energy is dynamically adjusted to optimize the allocation ratio and timing of cold and heat energy, thereby achieving coordinated optimization control of photovoltaic power storage and waste heat recovery.

[0050] The real-time temperature, humidity, sunshine intensity data provided by the weather station and the energy load forecast results based on the ARIMA model are obtained; based on the data provided by the weather station and the energy load forecast results, a dynamic programming algorithm is used to calculate the time series of cold and heat storage energy usage in the future preset time period, and the amount of cold and heat energy released in each time period is obtained; the dynamic programming algorithm uses energy balance as a constraint condition, takes the minimum total energy consumption as the objective function, the state variable is the energy level of the energy storage device, and the decision variable is the cold and heat energy released per hour; based on the cold and heat energy released, a fuzzy controller is used to dynamically adjust the operating parameters of the chiller and hot water unit; the fuzzy controller The controller's input variables are temperature deviation and load change rate, and the rule base contains IF-THEN rules; the real-time energy efficiency ratio of each device is obtained, and the distribution ratio of cold and hot energy is dynamically optimized based on the real-time energy efficiency ratio; photovoltaic power generation and power load are predicted in real time through a three-layer feedforward neural network, the hidden layer of which contains neurons, and the network is trained using the Levenberg-Marquardt algorithm; if the photovoltaic power generation is insufficient, the battery is discharged; if the battery power level is lower than a preset threshold, the operating load of the waste heat recovery system is increased, and the power of non-critical energy-consuming equipment is reduced according to a preset priority order.

[0051] Specifically, based on the obtained energy allocation strategy, combined with real-time temperature, humidity, and sunshine intensity data from the weather station and energy load forecasts based on the ARIMA model, a dynamic programming algorithm is used to calculate the time series of cold and hot energy storage usage over the next 24 hours, determining the amount of cold and hot energy released in each time period. The dynamic programming algorithm uses energy balance as a constraint and minimizing total energy consumption as the objective function. The state variable is the energy level of the energy storage device, and the decision variable is the hourly cold and hot energy released. Based on the calculated cold and hot energy release, a fuzzy controller is used to dynamically adjust the operating parameters of the chiller and water heater, including outlet water temperature, flow rate, and operating time. The fuzzy controller input variables are temperature deviation and load change rate, and the rule base contains 25 if-then rules. Furthermore, based on the real-time energy efficiency ratio of each device, the cold and hot energy allocation ratio is dynamically optimized to improve the overall efficiency of the waste heat recovery system. A three-layer feedforward neural network with 20 hidden neurons is used to predict photovoltaic power generation and electricity load in real time. The network is trained using the Levenberg-Marquardt algorithm. Based on the battery state of charge, the energy surplus or deficit is calculated and the operating power of the cooling and heating storage equipment is adjusted accordingly, achieving coordinated optimization of electricity and heat. A multi-level energy balance control strategy is implemented. When photovoltaic power generation is insufficient, battery discharge is prioritized. If the battery charge falls below 30%, the waste heat recovery system's operating load is increased to 90%. Simultaneously, the power of non-critical energy-consuming equipment is reduced according to pre-set priorities, reducing the load by 10%-30% to ensure the overall energy balance of the waste heat recovery system. A smart building complex is equipped with a 2MW photovoltaic power generation system, a 1MWh lithium battery energy storage system, and a 500kW waste heat recovery system. A weather station provides real-time data every 5 minutes, including temperature, humidity, and sunshine intensity. An ARIMA model predicts 24-hour energy load with an average error of ±7%. A dynamic programming algorithm divides the 24-hour period into 96 15-minute time periods, constructing a state transition matrix using 0.5kWh as the energy unit. The fuzzy controller uses a 5×5 rule matrix, and the input variables are discretized into five fuzzy sets. The three-layer neural network has an input layer with eight nodes, including time, temperature, humidity, and solar intensity. The hidden layer has 20 nodes and an output layer with two nodes, including photovoltaic power generation and electricity load. The Levenberg-Marquardt algorithm was trained for 1,000 rounds, and the mean squared error dropped below 0.01. The waste heat recovery system monitors the energy efficiency ratio of the chiller and hot water units in real time. When the cooling and heating load ratio is 3:2, the overall cost-effectiveness (COP) is improved by 8% by increasing the chilled water outlet temperature from 7°C to 9°C and reducing the hot water temperature from 55°C to 52°C. The battery discharge threshold is set at 30%. When the battery charge drops to 28%, the waste heat recovery system load is increased from 70% to 90%. Simultaneously, the lighting load, air conditioning load, and non-critical equipment load are reduced by 10%, 20%, and 30% respectively, according to a pre-set sequence.The waste heat recovery system performs a global optimization calculation every 15 minutes, dynamically adjusting the operating parameters of each subsystem to maximize energy utilization efficiency.

[0052] In step S107, based on the acquired waste heat data, cooling or heat absorption strategy, optimization control strategy, and energy allocation strategy, a waste heat recovery system is constructed and continuously monitored. The operation data of the waste heat recovery system is regularly analyzed to identify potential efficiency degradation or failure risks, and the operation parameters are adjusted or maintenance is performed in a timely manner.

[0053] Acquire operating parameters of the waste heat recovery system, including temperature, pressure, and flow rate; construct a five-layer neural network model based on the operating parameters, wherein the input layer of the neural network model includes 20 nodes, three hidden layers include 50, 30, and 20 neurons, respectively, and an output layer includes 2 nodes; collect operating data of the waste heat recovery system through a real-time data acquisition system, wherein the operating data includes heat exchanger inlet and outlet temperatures, working fluid flow rate, and pressure parameters; preprocess the operating data, including: filtering outliers using a 5-minute sliding window average and a 3σ criterion; input the preprocessed operating data into the five-layer neural network model to obtain the waste heat recovery system efficiency and waste heat recovery failure probability; determine whether the waste heat recovery system efficiency is lower than a preset threshold or whether the failure probability exceeds a preset threshold; if the waste heat recovery system efficiency is lower than the preset threshold or the failure probability exceeds the preset threshold, trigger a long short-term memory network to perform fault diagnosis; and adjust the waste heat recovery system operating parameters, including valve opening and variable frequency pump speed, using a fuzzy PID controller based on the fault diagnosis results of the long short-term memory network.

[0054] Specifically, a five-layer neural network model was constructed based on the acquired waste heat data, the cooling or heating absorption strategy, the optimized control strategy, and the energy allocation strategy. The input layer contained 20 nodes, corresponding to parameters such as temperature, pressure, and flow rate, as well as the policy quantization value. The three hidden layers contained 50, 30, and 20 neurons, respectively, using the Reluctant Unit (ReLU) activation function. The output layer had two nodes to predict the waste heat recovery system efficiency and waste heat recovery failure probability. The model used the Adam optimizer and mean squared error loss function, trained on 50,000 historical data points, with a learning rate of 0.001. The waste heat recovery system's operating data, including heat exchanger inlet and outlet temperatures, working fluid flow rate, and pressure, was collected every minute via a real-time data acquisition system. Outliers were filtered using a 5-minute sliding window average and the 3σ criterion. The preprocessed data was then fed into the neural network model to assess the current waste heat recovery system efficiency and potential failure risk. The waste heat recovery system efficiency was defined as the ratio of output energy to input energy, and the failure risk was calculated based on weighted scores of indicators such as equipment temperature, pressure, and vibration. A multi-level threshold alarm mechanism is implemented. When the waste heat recovery system efficiency falls below 80% or the failure probability exceeds 20%, a long short-term memory (LSTM) network is triggered to perform fault diagnosis, identify the specific fault type and location, and generate corresponding optimization recommendations or maintenance instructions. Based on the fault diagnosis results and optimization recommendations, a fuzzy PID controller automatically adjusts the waste heat recovery system's operating parameters, such as valve opening and variable frequency pump speed, with a control accuracy of ±2%. If the fault cannot be resolved through parameter adjustment, a maintenance instruction is sent to the operation and maintenance management platform. Fault information, historical data, and diagnostic results are stored in a time series database. A waste heat recovery system performance report is automatically generated at 2:00 AM daily, including a 24-hour efficiency curve, energy consumption analysis, and equipment status assessment. Trend analysis is used to predict waste heat recovery system performance changes over the next seven days. In a waste heat recovery system at an industrial park, the input layer of the five-layer neural network model contains 20 nodes, corresponding to 10 temperature sensors, 5 pressure sensors, 3 flow meters, and 2 policy quantization values. The three hidden layers contain 50, 30, and 20 neurons, respectively, using the Reluctant Unit (ReLU) activation function. The two nodes in the output layer predict the waste heat recovery system efficiency and waste heat recovery failure probability, respectively. The Adam optimizer with a learning rate of 0.001 was used, and training took approximately two hours for 50 epochs on an RTX3090 GPU. The real-time data acquisition system collects data every minute, using a 5-minute sliding window average and a 3σ criterion to filter outliers. When the waste heat recovery system efficiency falls below 80% or the failure probability exceeds 20%, the LSTM network is triggered for fault diagnosis. The LSTM contains 64 hidden units and takes as input the past 24 hours of time series data. The fuzzy rule base of the fuzzy PID controller contains 25 rules. The input variables are the deviation and the rate of change of the deviation, and the output is the adjustment amplitude of the control variable. InfluxDB is used as the time series database, writing 200 data points per second.Daily reports include 24-hour efficiency curves, energy consumption analysis, and equipment status assessments. Trend analysis uses the ARIMA model to predict waste heat recovery system efficiency over the next seven days, with an average error of less than ±5%.

[0055] It will be apparent to those skilled in the art that the present application is not limited to the details of the exemplary embodiments described above and that the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the present application is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. An automatic waste heat recovery method, characterized in that: include: Obtain waste heat pressure parameters, component medium, and flow characteristic data, classify the waste heat data, obtain characteristic vectors of different types of waste heat, including temperature, pressure, flow, and heat capacity parameters, and store the characteristic vectors of different types of preheating in the waste heat characteristic database; Obtain the waste heat distribution inside and outside the office environment, identify the concentrated areas of each waste heat source, collect the generation data of each waste heat source in real time, analyze and predict the randomness and intermittent laws of waste heat generation and the waste heat concentrated areas, and dynamically adjust the cooling or heat absorption strategy; retrieve the corresponding cooling or heat absorption operating parameters from the waste heat characteristic database, and control the cooling or heat absorption equipment to adaptively adjust the operating status according to the parameter settings, including: adjusting the output frequency of the heat pump inverter in real time through the PID controller, and the PID controller uses the Ziegler-Nichols method to perform PID parameter self-tuning; obtaining the temperature, pressure, and flow data of the condenser and evaporator, and the temperature, pressure, and flow data are filtered using a sliding average. Preprocessing is performed; based on the preprocessed temperature, pressure, and flow data, an adaptive neural fuzzy inference system is used to perform real-time fine-tuning of the heat pump speed and expansion valve opening. The adaptive neural fuzzy inference system continuously optimizes the inference rules through online learning; the cold and heat storage energy during the recovery process is monitored in real time, and the energy loss value of the remaining heat recovery is analyzed. If it is found that the energy loss exceeds the preset threshold, the fault diagnosis program is triggered, the fault type is determined, and the corresponding optimization control strategy is given; a photovoltaic power storage energy optimization scheduling model is constructed, the energy allocation strategy is solved, the use strategy of the cold and heat storage energy is dynamically adjusted, the distribution ratio and timing of the cold and heat energy are optimized, and the coordinated optimization control of photovoltaic power storage and waste heat recovery is realized.

2. The method according to claim 1, wherein The obtaining of waste heat pressure parameters, component medium and flow characteristic data includes: obtaining temperature parameters, pressure characteristics, flow data and heat capacity properties of the waste heat recovery system, performing a numerical range check on the parameter data, and marking the parameter value as abnormal data if it exceeds a preset threshold range; performing linear interpolation processing or elimination on the abnormal data, and normalizing the normal data to obtain a standardized parameter value; using a gas chromatograph to perform a component analysis on the waste heat medium, identifying the main chemical components and their contents, and merging the component data with the standardized parameter value.

3. The method according to claim 1, wherein The classification of the residual heat data includes: using a K-means clustering algorithm to classify the merged multidimensional data to obtain feature clusters of different types of residual heat, and calculating the centroid coordinates of each feature cluster as a feature vector of the residual heat; establishing a residual heat feature database table, which includes residual heat type, temperature vector, pressure vector, flow vector, heat capacity vector and component vector, and inserting the feature vectors of different types of residual heat into the residual heat feature database table for storage.

4. The method according to claim 1, wherein The method of obtaining the waste heat distribution inside and outside the office environment includes: obtaining an ambient temperature distribution map, wherein the ambient temperature distribution map is obtained by obtaining real-time data from multiple temperature sensors around an air compressor exhaust port outside the office environment and an air conditioner condenser inside the office environment, and estimating the waste heat distribution inside and outside the office environment using a Kriging interpolation algorithm based on the air compressor exhaust temperature and exhaust flow rate and the air conditioner condenser exhaust temperature and exhaust flow rate parameters; executing a K-means clustering algorithm based on the ambient temperature distribution map to obtain temperature anomaly areas as waste heat source concentration areas; and arranging a sensor network around the waste heat source concentration areas, wherein the sensor network includes temperature sensors, flow sensors, and pressure sensors.

5. The method according to claim 1, wherein The analysis and prediction of the randomness and intermittent patterns of waste heat generation and waste heat concentration areas includes: collecting waste heat source data obtained by a sensor network, the waste heat source data including temperature data, flow data and pressure data; performing time series analysis on the waste heat source data, the time series analysis including data preprocessing, trend analysis and seasonal analysis; and processing the waste heat source data using an autoregressive moving average model to obtain a characteristic curve and prediction model for each waste heat source.

6. The method according to claim 1, wherein The dynamic adjustment of the cold absorption strategy or the heat absorption strategy includes: calculating the optimal cold absorption or heat absorption strategy based on the characteristic curve and the prediction model, and the optimal cold absorption or heat absorption strategy is obtained by a linear programming method; if the waste heat generation of the prediction model is greater than a preset threshold, starting the cold absorption device and adjusting the cooling power; if the waste heat generation of the prediction model is less than the preset threshold, starting the heat absorption device and adjusting the heating power.

7. The method according to claim 1, wherein Retrieving corresponding cooling or heat absorption operating parameters from a waste heat characteristic database includes: retrieving a waste heat characteristic record with a highest matching degree from the waste heat characteristic database based on acquired ambient temperature and humidity data, wherein the waste heat characteristic database adopts a B-tree index structure; obtaining initial cooling or heat absorption operating parameters based on the waste heat characteristic records, wherein the initial cooling or heat absorption operating parameters include a heat pump speed range and an expansion valve opening range; and optimizing the initial cooling or heat absorption operating parameters using a fuzzy logic controller, wherein input variables of the fuzzy logic controller include a temperature deviation and a humidity deviation.

Images