Municipal facility state operation and maintenance analysis method for smart city

Abstract

The invention relates to the technical field of smart city operation and maintenance, in particular to a smart city-oriented municipal facility state operation and maintenance analysis method. The method comprises the following steps: collecting multi-source monitoring data of municipal facilities; performing linkage facility state load feature analysis under each collaborative operation based on the municipal facility multi-source monitoring data, and generating linkage facility state load feature data; performing linkage facility state load characteristic trend prediction processing on the linkage facility state load characteristic data to generate linkage facility state load characteristic trend data; establishing a load energy efficiency multi-agent decision model according to the linkage facility state load characteristic data and the linkage facility energy efficiency data; and transmitting the state load characteristic trend data of the linkage facility to a load energy efficiency multi-agent decision-making model to carry out linkage facility intelligent control parameter analysis of global energy efficiency optimization, and generating linkage facility intelligent control parameters. According to the invention, state linkage intelligent operation and maintenance control of smart city municipal facilities is realized.

Description

Technical Field

[0001] This invention relates to the field of smart city operation and maintenance technology, and in particular to a method for analyzing the status of municipal facilities in smart cities. Background Technology

[0002] With the accelerating pace of urbanization, the construction of smart cities has become a key path to improve urban management and the quality of life for residents. Municipal facilities, as the physical foundation of urban operation, directly impact public safety, energy efficiency, and the quality of urban services through their condition-based maintenance. Thanks to the widespread adoption of IoT technology, sensor-based condition monitoring of individual municipal facilities has been initially implemented. However, existing municipal facility condition-based maintenance analysis solutions often focus on individual facility operation data, lacking a systematic analysis of the spatial distribution and functional synergies between facilities. This results in maintenance systems failing to understand the interconnected behavior of facility groups from a macroscopic perspective. Existing predictive models often extrapolate trends based solely on the facility's historical time-series data, neglecting the transmission and dependencies of facility status across spatial networks. This deficiency leads to insufficient accuracy in predicting the overall load and status changes of the municipal facility system. Summary of the Invention

[0003] Based on this, the present invention provides a method for analyzing the status and operation of municipal facilities in smart cities to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a method for analyzing the status and operation of municipal facilities in smart cities includes the following steps: Step S1: Through the city-level Internet of Things platform and the sensing devices integrated with the geographic information system, the municipal facilities are monitored and preprocessed to generate multi-source monitoring data of municipal facilities, which includes municipal facility operation data, municipal facility environmental data and municipal facility geographical location data. Step S2: Based on the multi-source monitoring data of municipal facilities, perform status load characteristic analysis of the interconnected facilities under various collaborative operations, and generate status load characteristic data of the interconnected facilities; Step S3: Perform load characteristic trend prediction processing on the load characteristic data of the linkage facility status to generate load characteristic trend data of the linkage facility status. Step S4: Collect energy efficiency data of the linked facilities by using the status load characteristic data of the linked facilities, and generate energy efficiency data of the linked facilities; establish a multi-agent decision model of load and energy efficiency under each collaborative operation based on the status load characteristic data and energy efficiency data of the linked facilities, and generate a load energy efficiency multi-agent decision model. Step S5: Transmit the status load characteristic trend data of the linked facilities to the load energy efficiency multi-agent decision model for global energy efficiency optimization and analyze the intelligent control parameters of the linked facilities to generate intelligent control parameters of the linked facilities; execute the status operation and maintenance control of municipal facilities in the smart city through the intelligent control parameters of the linked facilities.

[0005] Furthermore, step S1 includes the following steps: Step S11: The city-level Internet of Things platform and the sensing devices integrated with the geographic information system are used to perform multi-source data monitoring and processing of municipal facilities to generate preliminary multi-source monitoring data of municipal facilities. The city-level Internet of Things platform receives the monitoring information of the sensing facilities, and the geographic information system determines the location information of the sensing facilities. Step S12: Obtain municipal facility data monitoring logs from the sensing facilities to obtain municipal facility data monitoring logs; Step S13: Perform data preprocessing on the preliminary multi-source monitoring data of municipal facilities based on the municipal facility data monitoring log to generate multi-source monitoring data of municipal facilities.

[0006] Furthermore, step S2 includes the following steps: Step S21: Analyze the types of municipal facilities based on the municipal facility operation data to generate municipal facility type data; Step S22: Establish a distribution analysis of municipal facility types based on the geographical location data and type data of municipal facilities, generate distribution data of municipal facility types, and perform a linkage distribution analysis of municipal facility types based on the distribution data of municipal facility types to generate linkage distribution data of municipal facility types. Step S23: Map municipal facility operation data and municipal facility environmental data to municipal facility type linkage distribution data to perform linkage facility status load characteristic analysis under various collaborative operations, and generate linkage facility status load characteristic data.

[0007] Furthermore, step S23 includes the following steps: Step S231: Based on the environmental data of municipal facilities, perform an environmental impact facility operation characteristic analysis on the municipal facility operation data to generate environmental impact facility operation characteristic data; Step S232: Based on the environmental impact facility type operation characteristic data, perform quantitative analysis and processing of environmental impact facility operation residual characteristics to generate quantitative data of environmental impact facility operation residual characteristics; Step S233: Map the facility operation characteristic data of environmental impact to the municipal facility type linkage distribution data to perform linkage operation characteristic analysis of municipal facilities with environmental impact, and generate facility linkage operation characteristic data with environmental impact; Step S234: Based on the characteristic data of the coordinated operation of facilities with environmental impact, perform quantitative analysis of the residual characteristics of the coordinated operation of facilities with environmental impact under each coordinated operation, and generate quantitative data of the residual characteristics of the coordinated operation of facilities with environmental impact. Step S235: Analyze the state load characteristics of the linkage facilities under each collaborative operation by using the quantitative data of the residual characteristics of the environmental impact facility operation and the quantitative data of the residual characteristics of the linkage operation of the environmental impact facility, and generate the state load characteristic data of the linkage facilities.

[0008] Furthermore, step S232 includes the following steps: Based on the operational characteristic data of facility types with environmental impact, residual analysis of the operation of environmental impact facilities is performed to generate residual data of the operation of environmental impact facility types. Based on the residual data of environmental impact facility operation, the residual characteristics of environmental impact facility operation are quantified to generate quantitative data of residual characteristics of environmental impact facility operation.

[0009] Furthermore, step S234 includes the following steps: Based on the characteristic data of the coordinated operation of facilities with environmental impact, residual analysis of the coordinated operation of facilities with environmental impact is performed under various collaborative operations to generate residual data of the coordinated operation of facilities with environmental impact. Based on the residual data of the coordinated operation of environmental impact facilities, the residual characteristics of the coordinated operation of environmental impact facilities under each coordinated operation are quantified to generate quantitative data of the residual characteristics of the coordinated operation of environmental impact facilities.

[0010] Furthermore, step S235 includes the following steps: Based on the quantitative data of residual characteristics of environmental impact facilities operation, the state adjustment demand analysis of individual operating facilities is carried out to generate state adjustment demand data of individual operating facilities. Then, the load characteristic analysis of the state adjustment demand data of individual operating facilities is carried out to generate state load characteristic data of individual operating facilities. Based on the single operating facility status load characteristic data as the condition for facility status change, the residual characteristics of the coordinated operation of the environmental impact facilities are quantified to analyze the environmental impact of the coordinated operation facility status adjustment demand data. Based on the data on the adjustment requirements of the status of facilities for coordinated operation, multi-layer convolutional feature analysis of the facility status of coordinated operation characteristics is performed to generate multi-layer convolutional feature data of the status of facilities for coordinated operation. Based on the multi-layer convolutional feature data of the status of the linked operation facilities, a fully connected feature fusion analysis of the linked operation characteristics is performed to generate fully connected status feature data of the linked operation facilities. Based on the full connectivity status characteristic data of the interconnected facilities, the status load characteristic data of the interconnected facilities under each collaborative operation is analyzed, and the status load characteristic data of the interconnected facilities is generated.

[0011] Furthermore, step S3 includes the following steps: Step S31: Collect prior data on the status and load characteristics of the linkage facilities based on the status and load characteristic data of the linkage facilities to obtain prior status and load characteristic data of the linkage facilities. Step S32: Based on the prior data of the state load characteristics of the linkage facilities, perform spatial transmission intensity dependency analysis of the state of the linkage facilities in order to design the spatial transmission dependency graph of the state of the linkage facilities. Step S33: Perform load autocorrelation time series characteristic analysis of the linkage facility status based on the prior linkage facility status load characteristic data, and generate linkage facility status load autocorrelation time series characteristic data. Step S34: Map the autocorrelation time series feature data of the state load of the linkage facility to the transitive dependency graph of the state space of the linkage facility, perform load cross-correlation time series feature analysis of the state of the linkage facility, and generate cross-correlation time series feature data of the state load of the linkage facility. Step S35: Analyze the load time-series change characteristics of the inter-correlation time-series characteristic data of the inter-linkage facility status load to generate inter-linkage facility status load change characteristic data. Step S36: Establish a load trend mapping relationship of the linkage facility status based on the load change characteristic data of the linkage facility status, so as to obtain the load trend prediction model of the linkage facility status. Step S37: Transmit the status feature data of the linkage facility to the load trend prediction model of the linkage facility status for load feature trend prediction processing, and generate load feature trend data of the linkage facility status.

[0012] Furthermore, step S32 includes the following steps: Step S321: Analyze the spatial adjacency relationship and functional coupling of the linkage facilities based on the prior state load characteristic data of the linkage facilities, and obtain the spatial adjacency relationship data and functional coupling data of the linkage facilities respectively. Step S322: Based on the spatial adjacency relationship data and functional coupling data of the interconnected facilities, perform spatial association feature analysis of the interconnected facilities to generate spatial association feature data of the interconnected facilities; Step S323: Analyze the causal relationship of the state space transmission of the linked facilities through the spatial association feature data of the linked facilities, and generate causal relationship data of the state space transmission of the linked facilities; Step S324: Design the spatial transit intensity dependency graph of the state of the linked facilities based on the causal relationship data of the state space of the linked facilities, and obtain the state space transit dependency graph of the linked facilities.

[0013] Furthermore, step S4 includes the following steps: Step S41: Collect energy efficiency data of the linked facilities by using the status load characteristic data of the linked facilities, and generate energy efficiency data of the linked facilities; Step S42: Based on the load characteristic data and energy efficiency data of the linkage facilities, perform load and energy efficiency correlation analysis of the linkage facilities to generate load and energy efficiency correlation data of the linkage facilities. Step S43: Based on the preset reinforcement learning algorithm and the load energy efficiency correlation data of the linkage facilities, design the initial model of multi-agent decision-making for load and energy efficiency under various collaborative operations, so as to obtain the initial load energy efficiency multi-agent decision-making model. Step S44: Set multi-agent hard constraint targets based on the preset facility operation parameter safety boundary, and set multi-agent soft optimization targets based on global energy efficiency optimization and facility load balancing; Step S45: Based on the multi-agent hard constraint objective and the multi-agent soft optimization objective, perform multi-agent objective empowerment processing on the initial load energy efficiency multi-agent decision model to generate the load energy efficiency multi-agent decision model.

[0014] The beneficial effects of this application lie in the fact that, through the deep integration of a city-level Internet of Things platform and a geographic information system, a comprehensive perception system for municipal facility data has been constructed. A data monitoring log-driven preprocessing mechanism effectively solves key technical challenges such as time synchronization, format standardization, and quality verification of multi-source heterogeneous data, forming a standardized multi-source monitoring data set for municipal facilities. The generated multi-source monitoring data for municipal facilities not only includes the operational status of the facilities themselves but also incorporates their environmental context and spatial network relationships, providing the possibility of understanding micro-level facility behavior from a macro-city scale. Through dual analysis of facility type and geographical distribution, a type-linked distribution map of municipal facilities is constructed, defining the collaborative relationship network between facilities from two dimensions: physical connection and functional coupling. Furthermore, by mapping operational data and environmental data to this map, a paradigm shift from isolated facility analysis to group-linked analysis is achieved. Its core lies in quantifying the residual impact of environmental factors on the operational status of individual facilities and facility groups, accurately capturing the deviation between actual operational data and theoretical prediction data, clearly defining the abnormal impact of environmental interference and collaborative operation on facility operation, and providing a quantitative basis for judging whether the operational status of facilities needs adjustment. By analyzing the operational characteristics of facilities affected by environmental factors, operational deviations caused by environmental disturbances are identified at the individual facility level. Subsequently, the analysis is elevated to the facility group level, exploring how environmental disturbances propagate and superimpose within the interconnected facility network. Advanced algorithms such as multi-layer convolution and fully connected fusion are used to extract deep load features that characterize complex collaborative operational relationships. The resulting interconnected facility status load feature data is no longer a simple collection of individual facility states, but rather reveals the collaborative load dynamics of the facility group as a whole system under the influence of internal and external factors, laying a solid feature foundation for accurate prediction and global optimization. The interconnected facility status load feature data undergoes trend prediction processing, collecting prior interconnected facility status load feature data covering different time periods and environmental conditions. This data fully reflects the long-term patterns of interconnected facility load changes, providing sufficient and comprehensive historical data support for subsequent analysis and avoiding prediction biases caused by insufficient data samples. By deeply analyzing the spatial adjacency and functional coupling of interconnected facilities, the spatial correlation strength and functional dependency mechanism between facilities are clarified. Based on these analyses, a spatial transfer intensity dependency diagram is designed, clearly presenting the causal relationship and degree of influence of load transfer between facilities. This fills the gap in traditional prediction methods that neglect the spatial correlation dimension, laying a key foundation for improving prediction accuracy. At the temporal feature analysis level, not only are the autocorrelation patterns of the load of a single interconnected facility at different time scales captured, but these autocorrelation temporal features are also mapped to the spatial transfer dependency diagram, further exploring the interrelationships of load changes among different interconnected facilities.By analyzing the temporal variation characteristics of load under spatial transmission, key features such as the time lag pattern of load peaks and the correlation of load fluctuation amplitude are extracted. Based on these features, a load trend mapping relationship is established, forming a predictive model that can simultaneously integrate spatial and temporal dimensions. The resulting load characteristic trend data of the interconnected facilities can accurately predict the load change trend of the interconnected facilities in the future, including the occurrence time of load peaks and valleys, as well as the core information such as the amplitude of load fluctuations. This effectively avoids the passive situation of operation and maintenance caused by the inability to predict load changes, such as taking measures in advance to deal with the risk of facility overload caused by load peaks. A multi-agent reinforcement learning framework is introduced to construct an intelligent decision-making model that can coordinate global optimization and local constraints. This step first establishes the correlation between the load and energy efficiency of interconnected facilities, providing a clear optimization direction for decision-making. On this basis, the designed load energy efficiency multi-agent decision-making model decouples the complex municipal facility system into multiple cooperating agents. Each agent is responsible for perceiving the local environment and making decisions, thereby distributively handling the high complexity of the system. The model innovatively sets both hard constraint objectives and soft optimization objectives for multi-agent systems. Hard constraints ensure that all facility operating parameters remain within safe boundaries, fundamentally guaranteeing system operational safety. Soft optimization objectives guide the agents to collaboratively explore optimal strategies for global energy efficiency and load balancing among facilities, while meeting safety requirements. This represents a fundamental leap from local optimization of individual facilities to maximizing overall system efficiency. By transmitting the state and load characteristics trend data of interconnected facilities to the multi-agent load efficiency decision model, control strategies are formulated not only based on the current state but also fully consider the future evolution of the system, thus achieving truly predictive control. The intelligent control parameters for interconnected facilities generated by the model represent a refined control scheme that, under all safety constraints, can proactively adapt to future load changes while simultaneously improving global energy efficiency and achieving load balancing.

[0015] Therefore, the municipal facility status operation and maintenance analysis method for smart cities of this invention achieves systematic modeling of the spatial distribution and functional coupling relationship between facilities by constructing interconnected distribution data of municipal facility types, thereby overcoming the limitation of existing technologies that only focus on independent operation data. It accurately depicts the collaborative operation mechanism of facility groups at the macro level and deeply analyzes the propagation path and degree of influence of environmental factors in the facility network, enabling the operation and maintenance system to fully perceive and understand the interconnected behavior between facilities, providing a reliable basis for subsequent prediction and decision-making. By designing a spatial transitive dependency graph of interconnected facility status, it deeply integrates the time-series characteristics of facility load with spatial network dependencies, effectively solving the defect of ignoring spatial transitive effects. This spatiotemporal joint modeling method can accurately capture the chain reactions and spatiotemporal propagation laws triggered by local state changes in the facility network, thereby significantly improving the prediction accuracy and reliability of the overall load and status change trends of the municipal facility system. Based on the trend data of interconnected facility status load characteristics, combined with a multi-agent decision-making model for the collaborative optimization of global energy efficiency and load balance, it achieves a leap from local single-point control to global system optimization. The generated intelligent control parameters take into account the dynamic linkage between facilities, thereby improving the overall efficiency and intelligence level of the status operation and maintenance of smart city municipal facilities. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of a municipal facility status operation and maintenance analysis method for smart cities according to the present invention. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0018] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. Functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods. The term "and / or" as used herein includes any and all combinations of one or more of the associated items listed.

[0019] To achieve the above objectives, please refer to Figures 1 to 2 This invention provides a method for analyzing the status and operation of municipal facilities in smart cities. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart illustrating the steps of a municipal facility status operation and maintenance analysis method for smart cities according to the present invention. The method includes the following steps: Step S1: Through the city-level Internet of Things platform and the sensing devices integrated with the geographic information system, the municipal facilities are monitored and preprocessed to generate multi-source monitoring data of municipal facilities, which includes municipal facility operation data, municipal facility environmental data and municipal facility geographical location data. In this embodiment of the invention, in a smart city municipal infrastructure operation and maintenance project in a provincial capital city, multi-source data monitoring and preprocessing are carried out for three core municipal facilities within a 150-square-kilometer area in the central urban area: water supply and drainage network, road lighting system, and traffic lights. To achieve data acquisition, pressure sensors and flow sensors are deployed at key nodes of the water supply and drainage network. The pressure sensors adopt a diffused silicon structure, with a measurement range of 0-1.6MPa and an accuracy of ±0.2% FS. The flow sensors adopt an electromagnetic principle, supporting a measurement range of 0-500m³ / h. Power sensors are integrated into the distribution boxes of the road lighting facilities to collect real-time data on single-lamp power and total circuit energy consumption. Operating status sensors are installed in the traffic light control boxes to record signal switching cycles and the duration of each light color. Simultaneously, an environmental monitoring device is deployed every 2 kilometers along urban roads, including temperature and humidity sensors (measurement range -40~85℃, measurement accuracy 0~100% RH) and rainfall sensors (measurement range 0-100mm / h). All sensing devices are connected to the city-level IoT platform via the LoRaWAN wireless communication protocol. This platform has the capability to receive 100,000 data entries per second and can store the raw monitoring data uploaded by each sensor in real time. Simultaneously, each sensing device has a built-in BeiDou positioning module with a positioning accuracy of ≤1 meter. Its location coordinate data is synchronously transmitted to the city's geographic information system (GIS) via a dedicated interface. The GIS associates the sensing devices with corresponding municipal facilities based on the coordinate data, generating preliminary multi-source monitoring data for municipal facilities containing "unique facility identifier - sensing device ID - location coordinates - monitoring data - collection timestamp". To ensure data quality, monitoring logs for each sensing device are exported from the city-level IoT platform. The logs record in detail the operating status (normal / faulty), data collection frequency (fixed at 5 minutes / time), data transmission packet loss rate (statistically calculated hourly), and abnormal data markers (such as invalid values ​​when the sensor is offline).Pre-processing of preliminary data based on monitoring logs: First, data corresponding to sensors marked as "faulty" in the logs are filtered out and deleted. For short-term packet loss data marked in the packet loss rate statistics (single packet loss duration not exceeding 30 minutes), linear interpolation is used to calculate interpolation based on three valid data points before and after the packet loss period to fill in missing values. Second, the 3σ criterion is used to detect outliers in all data, calculate the mean and standard deviation of each monitoring indicator, and remove extreme data exceeding the range of "mean ± 3 times standard deviation" (such as rainfall data exceeding the sensor's range during heavy rain). Finally, monitoring data of different dimensions are normalized, mapping the values ​​of indicators such as pressure, flow, and power to the [0,1] interval. The mapping formula is determined based on the rated measurement range of each sensor. Finally, structured multi-source monitoring data of municipal facilities is generated, which includes municipal facility operation data (such as pipeline pressure, lighting power, and traffic light cycle), environmental data (such as temperature and humidity, and rainfall), and geographical location data (such as facility coordinates and administrative region).

[0020] Step S2: Based on the multi-source monitoring data of municipal facilities, perform status load characteristic analysis of the interconnected facilities under various collaborative operations, and generate status load characteristic data of the interconnected facilities; In this embodiment of the invention, based on multi-source monitoring data of municipal facilities, a status load characteristic analysis of interconnected facilities under various collaborative operations is conducted for three types of facilities in the central urban area: water supply and drainage networks, road lighting systems, and traffic lights. Facility type analysis is performed based on the characteristic differences in the municipal facility operation data: facilities containing fields such as "pressure, flow rate, and network number" are classified as water supply and drainage facilities; facilities containing fields such as "power, brightness, and lighting circuit number" are classified as lighting facilities; and facilities containing fields such as "signal cycle, light duration, and intersection number" are classified as traffic lights, generating clear municipal facility type data. Next, a type distribution analysis was conducted by combining the geographical location data and type data of municipal facilities: the K-Means spatial clustering algorithm was adopted, with the geographical location coordinates of the facilities as the clustering feature, and the number of clusters was set to 12 (corresponding to 12 street areas in the central urban area). Through iterative calculation, the spatial distance error of the same type of facilities in each area was minimized, generating municipal facility type distribution data on the quantity, density and concentrated distribution areas of different types of facilities in each street area; on this basis, a linkage and correlation distribution analysis was carried out: graph theory method was used to abstract each type of facility as a node in the graph, generating linkage distribution data of municipal facility types. Subsequently, municipal facility operation data and environmental data are mapped to this linked distributed data to conduct load characteristic analysis: First, based on rainfall, temperature, and humidity data in the environmental data, the operation data is analyzed for environmental impact on facility operation characteristics. A multiple linear regression model is used, with rainfall and temperature as independent variables and pipeline flow and lighting power as dependent variables. The regression coefficients are calculated using the least squares method to clarify the degree of influence of environmental factors on the operation status of a single facility, generating facility operation characteristic data for environmental impact. Based on this data, a quantitative analysis of the residual characteristics of facility operation affected by environmental impact is conducted: First, the difference between the actual operation data of a single facility and the predicted data of the regression model (i.e., the residual) is calculated. For example, if the actual pipeline flow is 120 m³ / h at a certain moment and the model predicts the flow as 105 m³ / h, the residual is 15 m³ / h. Then, the standard deviation of all residuals within a certain time period (e.g., 24 hours) is calculated. The standard deviation is used as a quantitative indicator of residual characteristics to generate quantitative data of residual characteristics of facility operation affected by environmental impact. By mapping environmental impact facility operation characteristic data to municipal facility type linkage distribution data, the collaborative operation patterns of different types of facilities under environmental impact are analyzed, generating environmental impact facility linkage operation characteristic data. Based on this data, quantitative analysis of environmental impact facility linkage operation residual characteristics is carried out: similarly, the residual between the actual operation data and the predicted data of the collaborative prediction model (built based on linkage distribution data) under the linkage operation scenario is calculated, such as the difference between the actual collaborative value and the predicted collaborative value of pipeline flow and traffic light green light duration at a certain moment, and then the standard deviation of the residual is calculated to generate quantitative data of environmental impact facility linkage operation residual characteristics.Finally, the status load characteristic data of the linkage facilities is generated by combining the two types of residual characteristic quantitative data: Based on the standard deviation index in the residual characteristic quantitative data of environmental impact facility operation, if the residual standard deviation of a certain type of single facility exceeds the preset threshold (e.g., the standard deviation threshold of pipeline flow residual is set to 8 m³ / h), it is determined that the facility has a status adjustment requirement. The load change magnitude corresponding to the adjustment requirement is analyzed (e.g., when the residual standard deviation exceeds the threshold by 2 m³ / h, the pipeline load needs to be reduced by 10%), generating status load characteristic data of a single operating facility; Using this data as the facility status change condition, the linkage operation residual characteristic quantitative data of the coordinated environmental impact facilities is substituted to analyze the load magnitude that the linkage facilities need to adjust synchronously (e.g., when the pipeline load is reduced by 10%, the load of the linkage pump station needs to be reduced by 8%), generating status adjustment requirement data of the linkage operation facilities; using 3 A multi-layer convolutional neural network is used to perform multi-layer convolutional feature analysis on the adjustment demand data. The first layer has a kernel size of 3×3 to extract short-term fluctuation features of facility load, the second layer has a kernel size of 5×5 to extract medium-term temporal correlation features of load, and the third layer has a kernel size of 7×7 to extract long-term spatial correlation features of load. After each convolution, ReLU activation function is applied to enhance the non-linear expression of features, generating multi-layer convolutional feature data of the coordinated operation facility status. This convolutional feature data is then input into a two-layer fully connected neural network. The first fully connected layer has 128 neurons, and the second layer has 64 neurons. Matrix multiplication is used to fuse the scattered convolutional features into a feature vector of a unified dimension, generating fully connected status feature data of the coordinated operation facility. Based on this feature data, the peak load, load fluctuation frequency, and load coordination coefficient of the facility under each coordinated operation scenario are calculated. Finally, coordinated facility status load feature data containing the load status, coordination correlation strength, and adjustment demand of various coordinated facilities are generated.

[0021] Step S3: Perform load characteristic trend prediction processing on the load characteristic data of the linkage facility status to generate load characteristic trend data of the linkage facility status. In this embodiment of the invention, load characteristic trend prediction processing of the status of interconnected facilities is performed on the load characteristic data of interconnected facilities. First, the load characteristic data of interconnected facilities generated daily in the central urban area over the past 12 months is collected. At the same time, environmental data (such as daily rainfall and average temperature) and municipal facility operation and maintenance records (such as equipment maintenance time and load adjustment records) for the corresponding time period are collected simultaneously. The three are integrated into prior interconnected facility status load characteristic data containing 10,800 sets of samples (360 days × 30 interconnected facility groups). Each data entry in the sample contains a timestamp, load characteristic index, environmental index, and operation and maintenance mark. Based on prior data, a spatial transmission strength dependency analysis of the status of interconnected facilities is conducted: The geographical location information and load characteristic data of facilities in the prior data are correlated to analyze the spatial adjacency relationships of each interconnected facility group. Facilities with a distance ≤ 500 meters are identified as spatially adjacent facilities by converting the latitude and longitude differences between facilities into actual ground distances. Simultaneously, the functional coupling between facilities is analyzed. Pearson correlation coefficients of different facility load characteristic data are calculated; facilities with an absolute correlation coefficient ≥ 0.7 are classified as strongly coupled facilities, while those with a correlation coefficient between 0.3 and 0.7 are weakly coupled facilities. Spatial adjacency relationship data and functional coupling data of interconnected facilities are generated separately. Based on these two types of data, spatial association characteristic analysis is conducted: the distance values ​​corresponding to spatial adjacency relationships are converted into weights in the 0-1 interval (the closer the distance, the greater the weight). The correlation coefficient corresponding to functional coupling is used as the coupling weight. The spatial association strength between each pair of facilities is calculated by weighted summation, generating spatial association characteristic data of interconnected facilities. The Granger causality test method is used to analyze the causal relationship of the spatial association characteristic data, with the load change data of one facility as the independent variable and the load change data of another interconnected facility as the dependent variable. The F-test is then used to analyze the causal relationship of the spatial association characteristic data. The independent variable is tested to determine whether it has a predictive effect on the dependent variable. If the p-value of the test result is <0.05, a causal relationship is determined, and spatial transitive causal relationship data of the linked facilities is generated. Based on this causal relationship data, a spatial transitive dependency graph of the linked facilities is constructed using the Neo4j graph database. The nodes in the graph represent each linked facility group, and the node attributes include facility type and current load value. The edge attributes between nodes include association strength weight and causal relationship label, thus completing the spatial transitive dependency graph design. Load autocorrelation time series feature analysis is carried out based on the prior linked facility state load feature data: using the ARIMA model, with the load feature data of a certain linked facility group as the time series, the model parameters p=3, d=1, q=2 (p is the number of autoregressive terms, d is the difference frequency, and q is the number of moving average terms) are determined through the AIC information criterion. The autocorrelation coefficients of this time series at the 1-hour, 6-hour, and 24-hour time scales are calculated, and the repetitive variation pattern of the load data in different time periods is analyzed to generate the time series feature data of the linked facility state load autocorrelation.Autocorrelation time-series characteristic data are mapped to a spatial transitive dependency graph. For facility node pairs with causal relationships in the graph, the cross-correlation coefficients of their load time-series data are calculated. The response time and response magnitude of the load of one facility after a change in the load of another facility are analyzed to generate cross-correlation time-series characteristic data of the state load of linked facilities. The load time-series change characteristics of spatial state transit are analyzed on the cross-correlation time-series characteristic data: the time-series change curves of facility load under different environmental conditions (such as sunny days and rainy days) are extracted, and characteristic parameters such as peak occurrence time, peak duration, and load rise / fall rate are calculated. The changing patterns of these parameters during spatial transit are analyzed, such as "in rainy days, the peak load of the pipeline network occurs 2 hours earlier than in sunny days, and the peak duration is extended by 1 hour," generating load change characteristic data of linked facilities. Based on this data on load change characteristics, a load trend mapping relationship is established: A Long Short-Term Memory (LSTM) network is used to construct a load trend prediction model for the linked facilities. The input layer of the model contains five types of feature parameters from the load change characteristic data (peak time, duration, rate of increase, rate of decrease, and cross-correlation coefficient), with an input dimension of 32. Two hidden layers are set, with the first layer containing 256 neurons and the second layer containing 128 neurons, using the tanh activation function. The output layer contains 72 neurons, corresponding to the hourly load prediction values ​​for the next 72 hours. The model is trained using 80% of the samples in the prior data, validated using 10% of the samples, and tested using 10% of the samples. The mean squared error is used as the loss function, and the model parameters are iteratively adjusted using the Adam optimizer until the loss function value of the validation set stabilizes and converges, thus obtaining the load trend prediction model for the linked facilities. Input the current status characteristic data of the interconnected facilities (including real-time load values, environmental data, and spatial correlation data) into the prediction model. The model extracts features from the input data and performs time-series correlation calculations to output the hourly load prediction values ​​for each interconnected facility group over the next 72 hours. These values ​​include the load value, load change trend (increasing / decreasing / stable), and load anomaly risk level (determined based on the difference between the predicted value and the rated load), generating interconnected facility status load characteristic trend data.

[0022] Step S4: Collect energy efficiency data of the linked facilities by using the status load characteristic data of the linked facilities, and generate energy efficiency data of the linked facilities; establish a multi-agent decision model of load and energy efficiency under each collaborative operation based on the status load characteristic data and energy efficiency data of the linked facilities, and generate a load energy efficiency multi-agent decision model. In this embodiment of the invention, based on the status load characteristic data of the interconnected facilities, energy efficiency data collection and multi-agent decision-making model establishment are carried out in three types of interconnected facilities in the central urban area: water supply and drainage network, road lighting system, and traffic signal lights. First, energy efficiency monitoring nodes for each facility are determined using the status and load characteristic data of the linked facilities: Smart meters are installed in the power distribution boxes of pump stations in the water supply and drainage network. These meters use electronic metering principles, have a voltage measurement range of 0-400V, a current measurement range of 0-100A, and a metering accuracy of 0.5, and can collect the active power, reactive power, and cumulative power consumption of the pump station in real time. A miniature power metering module is integrated into each street light head in the road lighting system, supporting voltage measurement of 0-220V and current measurement of 0-5A, and can collect the real-time power and energy consumption of a single light. Energy consumption monitoring units are installed in the traffic signal control boxes to collect the total energy consumption of the traffic lights at the entire intersection and the individual energy consumption of each light color. All energy efficiency monitoring devices receive load change trigger signals from the status and load characteristic data of the linked facilities through a city-level IoT platform. When the load value change in the load characteristic data exceeds 5%, the energy efficiency data collection frequency is automatically increased to once per minute (5 times when the load is stable). (Minutes / times) to generate linked facility energy efficiency data containing facility ID, collection time, real-time energy consumption, and cumulative energy consumption. Next, load and energy efficiency correlation analysis is conducted based on the linked facility status load characteristic data and energy efficiency data: using the Pearson correlation coefficient calculation method, correlation analysis is performed on the load characteristic indicators (such as pump station speed, lighting power, signal light cycle) and energy efficiency indicators (such as active power, unit load energy consumption) of each type of facility, calculating the correlation coefficient matrix. Based on the correlation coefficient matrix, the variation law of energy efficiency within different load ranges is identified, generating linked facility load energy efficiency correlation data. An initial model is constructed based on a pre-defined reinforcement learning algorithm (using the Multi-Agent Deep Deterministic Policy Gradient Algorithm, i.e., MADDPG) and load energy efficiency correlation data: Each type of interconnected facility group is treated as an agent (a total of 30 agents are set up, corresponding to 30 interconnected facility groups). The state space of each agent contains the load characteristic data (such as current load value, load change rate) and energy efficiency data (such as current energy consumption, unit load energy consumption) of the facility group. The action space contains the load adjustment amount (such as the adjustment range of pump station speed, the adjustment range of lighting power). The reward function is initially set to a single indicator based on the energy consumption reduction rate. The model network structure is built using the TensorFlow framework. Each agent contains a policy network and a value network. The policy network adopts 3 fully connected layers (the number of neurons is 256, 128, and 64 respectively). The value network adopts the same structure. The model is initially trained using prior data (training rounds are 2000 rounds, each round contains 100 time steps), resulting in the initial load energy efficiency multi-agent decision model.Based on the safety boundaries of the operating parameters of various facilities, multi-agent hard constraint objectives are set, and the actions generated by the model satisfy that all parameters are not lower than or exceed the corresponding safety boundaries. At the same time, soft optimization objectives are set based on global energy efficiency optimization and facility load balancing: the global energy efficiency optimization objective is determined by calculating the ratio of the total energy consumption to the total load of all interconnected facilities in the central urban area, with the objective of reducing this ratio by 15%. The facility load balancing objective is determined by calculating the deviation between the load value and the average load value of each interconnected facility group, with the objective of the absolute value of the deviation being less than 10%. The initial model is empowered with multi-agent objectives based on hard constraint objectives and soft optimization objectives: a hard constraint penalty term is added to the model's reward function; if the actions generated by the model violate the hard constraint objective, the reward value is directly reduced by 100; a soft optimization reward term is added to the reward function; if the ratio of total energy consumption to total load decreases by 1%, the reward value increases by 10; if the absolute value of the load deviation of each facility is less than 10%, an additional reward of 20 is given. The model is then trained again using the adjusted reward function (training 3000 rounds). After each round of training, a test set is used to verify whether the model meets the hard constraint and soft optimization objectives, until the model does not violate the hard constraints in 100 consecutive rounds of testing and the soft optimization objective achievement rate exceeds 90%, generating the final load energy efficiency multi-agent decision model.

[0023] Step S5: Transmit the status load characteristic trend data of the linked facilities to the load energy efficiency multi-agent decision model for global energy efficiency optimization and analyze the intelligent control parameters of the linked facilities to generate intelligent control parameters of the linked facilities; execute the status operation and maintenance control of municipal facilities in the smart city through the intelligent control parameters of the linked facilities.

[0024] In this embodiment of the invention, the load characteristic trend data of the interconnected facilities is transmitted to a multi-agent decision-making model for load energy efficiency, and intelligent control parameter analysis of the interconnected facilities is carried out for global energy efficiency optimization. The model first analyzes the load characteristic trend data, extracts the interconnected facility groups whose predicted load values ​​exceed the rated load set ratio within a specific future time period, and combines the load energy efficiency correlation data of these facility groups to analyze the adjustable load parameter range without violating hard constraint targets. For the water supply and drainage facilities, the model, based on the trend data's prediction that "the network flow will increase from its current value to a value close to the rated load within a specific time period," and combined with the load energy efficiency correlation data's pattern that "for every set percentage decrease in flow, energy consumption decreases by a corresponding percentage," calculates that the network flow needs to be controlled below a specific value. The corresponding control parameters are adjusting the network valve opening from its current value to the target value and adjusting the downstream pump station speed from its current value to the target value. For the lighting facilities, based on the trend data's prediction that "the lighting load will reach a value close to the rated load within a specific time period," and combined with the correlation pattern that "when the power decreases from its current value to the target value, energy consumption decreases by a set percentage and brightness meets safety standards," calculates the lighting power control parameter as the target value, corresponding to adjusting the output current of each street light to the target value. The model uses a multi-agent collaborative decision-making mechanism to verify the impact of these control parameters on related facilities: for example, whether adjusting the opening of water supply and drainage valves causes the upstream pipeline pressure to exceed the minimum allowable pressure (calculated, the pressure remains within a safe range and meets the requirements); whether adjusting lighting power affects the visual recognition effect of surrounding traffic lights (correlation analysis shows that the brightness still meets nighttime driving safety standards), ensuring that all control parameters meet the goals of global energy efficiency optimization and load balancing. Ultimately, it generates intelligent control parameters for the linked facilities, including the equipment ID of each linked facility group, control parameter type (such as valve opening, speed, and power), control parameter value, and execution time window. Through the control command distribution module of the city-level IoT platform, these intelligent control parameters are converted into commands conforming to the communication protocols of each facility controller, completing the status operation and maintenance control of municipal facilities in the smart city.

[0025] Furthermore, step S1 includes the following steps: Step S11: The city-level Internet of Things platform and the sensing devices integrated with the geographic information system are used to perform multi-source data monitoring and processing of municipal facilities to generate preliminary multi-source monitoring data of municipal facilities. The city-level Internet of Things platform receives the monitoring information of the sensing facilities, and the geographic information system determines the location information of the sensing facilities. In this embodiment of the invention, multi-source data monitoring is carried out for three core municipal facilities in the central urban area of ​​a smart city: water supply and drainage network, road lighting system, and traffic lights. Diffused silicon pressure sensors and electromagnetic flow sensors are deployed at key nodes of the main and branch pipes of the water supply and drainage network. The pressure sensors have corrosion-resistant housings to adapt to the humid environment of the network, and the flow sensors are fixed to the inner wall of the pipe using flange connections. Power sensors are integrated into the power distribution boxes of each lamp in the road lighting facilities, collecting the current and voltage of a single lamp circuit through current transformers. Operating status sensors are installed in the traffic light control boxes, recording the lamp color switching cycle and the duration of each lamp color by detecting relay actions. Simultaneously, an environmental monitoring unit is deployed every 2 kilometers on both sides of the road, containing platinum resistance temperature and humidity sensors and tipping bucket rainfall sensors. All sensing devices have built-in LoRaWAN wireless communication modules, using spread spectrum technology to achieve long-distance, low-power data transmission at a fixed rate of 50kbps, with a data transmission interval of 5 minutes per transmission, uploading the monitoring data to the city-level IoT platform in real time. The platform features a distributed storage architecture, supporting the concurrent reception of 100,000 data entries per second. After reception, the raw monitoring data is stored according to the classification of "facility type - device unique identifier". At the same time, each sensing device integrates a Beidou positioning module with a positioning accuracy of ≤1 meter. Its real-time location coordinates are synchronously transmitted to the city's geographic information system through a dedicated data interface. The geographic information system matches the coordinate data with the physical location information of the corresponding municipal facilities (such as the road where the pipeline is located, the road section where the lighting is located, and the intersection where the traffic lights are located), and associates the location information with the monitoring data received by the IoT platform according to the timestamp (accurate to the second) to generate preliminary multi-source monitoring data of municipal facilities.

[0026] Step S12: Obtain municipal facility data monitoring logs from the sensing facilities to obtain municipal facility data monitoring logs; In this embodiment of the invention, data monitoring logs of municipal facilities are obtained from the log management module of a city-level IoT platform. This module has the function of filtering logs by time range and device identifier, and the log extraction range is completely consistent with the time range of the preliminary data. The logs are classified with the unique identifier of the sensing device as the core dimension. The log entries for each device contain complete operational information: clearly recording the sensor's working status, data acquisition frequency, data transmission packet loss rate, and abnormal data markers. Each log entry carries a timestamp and device identifier consistent with the preliminary multi-source monitoring data of municipal facilities, ensuring that each monitoring record in the preliminary data can be accurately associated with the real-time operating log of its acquisition device, clearly indicating the device's operating condition at the time of data generation.

[0027] Step S13: Perform data preprocessing on the preliminary multi-source monitoring data of municipal facilities based on the municipal facility data monitoring log to generate multi-source monitoring data of municipal facilities.

[0028] In this embodiment of the invention, based on the acquired municipal facility data monitoring logs, systematic preprocessing is performed on the preliminary multi-source monitoring data of municipal facilities: First, data corresponding to sensors marked as "fault" in the logs are directly removed, as the values ​​collected by sensors under fault conditions cannot reflect the true operating status of the facilities; For short-term packet loss data marked in the logs (single packet loss duration not exceeding 30 minutes), linear interpolation is used. Based on three consecutive valid data points before and after the packet loss period, the theoretical data values ​​at each time point within the packet loss period are derived by calculating the slope of change of adjacent valid data, thus filling in missing records; Subsequently, the 3σ criterion is used for outlier detection. First, the overall mean and standard deviation of each type of monitoring indicator (such as pipeline pressure and lighting power) are calculated, and outliers exceeding "mean ± 3 times standard deviation" are detected. Extremely abnormal data within the specified range are deleted to avoid invalid data interference from sudden events such as heavy rain and voltage fluctuations. Finally, monitoring data of different dimensions are normalized. Using the rated measurement range of each sensor as a benchmark, the original values ​​of indicators such as pressure, flow rate, and power are uniformly mapped to the [0,1] interval through linear transformation. The transformation process strictly follows the rated upper and lower limits in the sensor's technical specifications to ensure the comparability of different data types. After the above processing, a well-structured, complete, and numerically comparable multi-source monitoring data of municipal facilities is generated, providing a reliable data foundation for subsequent analysis of the status and load characteristics of interconnected facilities.

[0029] Furthermore, step S2 includes the following steps: Step S21: Analyze the types of municipal facilities based on the municipal facility operation data to generate municipal facility type data; In this embodiment of the invention, operational data fields from multi-source monitoring data of municipal facilities are analyzed to establish a facility type feature tag library: operational data containing fields such as "pipeline pressure, pipeline flow, and pump station speed" are tagged as "water supply and drainage facilities"; data containing fields such as "single lamp power, loop current, and brightness value" are tagged as "lighting facilities"; and data containing fields such as "signal switching cycle, light color duration, and intersection number" are tagged as "traffic signal facilities". A field matching method is used to determine the type of each piece of operational data. Each piece of data is checked to see if any feature fields from the tag library exist. If a piece of data contains all the feature fields of a certain type of facility (e.g., both "pipeline pressure" and "pipeline flow"), the facility associated with that data is directly identified as the corresponding type. If any field is missing (e.g., only the "power" field is present), the determination is supplemented by combining the facility's geographical location data (e.g., located on both sides of a road and associated with street light numbers). Once the determination is complete, a unique type identifier is assigned to each facility (e.g., "GP" represents water supply and drainage facilities, "ZM" represents lighting facilities, and "JT" represents traffic signal facilities), and the unique identifier of the facility is associated with the operational data characteristics to generate municipal facility type data.

[0030] Step S22: Establish a distribution analysis of municipal facility types based on the geographical location data and type data of municipal facilities, generate distribution data of municipal facility types, and perform a linkage distribution analysis of municipal facility types based on the distribution data of municipal facility types to generate linkage distribution data of municipal facility types. In this embodiment of the invention, based on municipal facility type data and geographic location data, the K-Means spatial clustering algorithm is used to conduct type distribution analysis: The geographic location coordinates (latitude and longitude) of the facilities are used as clustering features. The number of clusters is determined according to the city's administrative divisions (e.g., corresponding to 10 street areas in the central urban area). Each cluster center is initialized as the geometric center point of its respective street area. By iteratively calculating the Euclidean distance from each facility to the cluster center, facilities are assigned to the nearest cluster until the cluster center position no longer changes. The number, density (number of facilities per unit area), and concentrated distribution areas of different types of facilities within each cluster are statistically analyzed (e.g., water supply and drainage facilities in a certain street area are mainly distributed along river roads), generating municipal facility type distribution data. Based on this distribution data, a linkage and correlation distribution analysis is conducted: using graph theory, each facility is abstracted as a node, with node attributes including type identifier and coordinates. The correlation of operational data of different types of facilities is analyzed based on the association information of collaborative operation, generating municipal facility type linkage distribution data.

[0031] Step S23: Map municipal facility operation data and municipal facility environmental data to municipal facility type linkage distribution data to perform linkage facility status load characteristic analysis under various collaborative operations, and generate linkage facility status load characteristic data.

[0032] In this embodiment of the invention, municipal facility operation data (such as flow rate of water supply and drainage facilities, and power of lighting facilities) and environmental data (such as rainfall, temperature, and humidity) are extracted from multi-source monitoring data of municipal facilities. These are then matched with nodes in the linked distribution data of municipal facility types according to the unique identifier of the facility, achieving precise mapping between the data and the linked distribution nodes. A multiple linear regression model is used to analyze the impact of environmental data on operation data: with rainfall and temperature as independent variables, and flow rate of water supply and drainage facilities and power of lighting facilities as dependent variables, regression coefficients are calculated using the least squares method to clarify the degree of influence of environmental factors on the operation status of a single facility (such as the magnitude of the impact of rainfall changes on flow rate), generating characteristic data of single facility operation under environmental influence. This data is then substituted into the network of associated nodes in the linked distribution data to analyze the impact of changes in the operation status of a certain facility on associated facilities (such as the speed response law of downstream associated pumping stations after rainfall causes an increase in the flow rate of water supply and drainage facilities in a certain area), generating characteristic data of linked facility operation under environmental influence. Based on the collaborative operation feature data, a sliding window method (with a window duration of 1 hour) is used to extract load features under different collaborative scenarios (such as load peaks and fluctuation amplitudes during peak hours and rainfall periods). Then, a multi-layer convolutional neural network is used to deeply extract the load features: the first convolutional kernel (3×3) extracts short-term load fluctuation features, the second layer (5×5) extracts medium-term temporal correlation features, and the third layer (7×7) extracts spatial correlation features. Each convolutional layer is processed by the ReLU activation function. Finally, a fully connected layer is used to fuse the multi-dimensional features into a unified feature vector. The load coordination coefficient (the degree of synchronization of load changes in associated facilities) under each collaborative scenario is calculated by combining the feature vector, generating the status load feature data of the collaborative facilities.

[0033] Furthermore, step S23 includes the following steps: Step S231: Based on the environmental data of municipal facilities, perform an environmental impact facility operation characteristic analysis on the municipal facility operation data to generate environmental impact facility operation characteristic data; In this embodiment of the invention, environmental data (rainfall, air temperature, relative humidity) and operational data (pipe flow rate and pumping speed of water supply and drainage facilities, single lamp power and loop current of lighting facilities, and switching cycle of traffic signal facilities) are separated from multi-source monitoring data of municipal facilities. The two types of data are aligned by timestamp (accurate to the minute) according to the unique identifier of each facility. An environmental impact analysis model is constructed using a multiple linear regression method, with rainfall, temperature, and humidity in the environmental data as independent variables and core operational parameters of various facilities (such as flow rate of water supply and drainage facilities and power of lighting facilities) as dependent variables. The regression coefficients are solved by the least squares method to determine the influence weight of each environmental factor on the operational parameters (such as the influence of a one-unit change in rainfall on flow rate and the influence of a one-unit change in temperature on power). For each type of facility, group them according to environmental scenarios (such as rainfall scenarios with rainfall >10mm / h and high temperature scenarios with temperature >35℃), calculate the mean, fluctuation range and trend of operating parameters under different scenarios, clarify the quantitative correlation between environmental factors and operating parameters (such as the proportion of increase in the mean flow rate under rainfall scenarios compared to rainless scenarios, and the fluctuation frequency of lighting power under high temperature scenarios), and generate facility operation characteristic data of environmental impact.

[0034] Step S232: Based on the environmental impact facility type operation characteristic data, perform quantitative analysis and processing of environmental impact facility operation residual characteristics to generate quantitative data of environmental impact facility operation residual characteristics; In this embodiment of the invention, based on the constructed multiple linear regression model, the environmental impact facility operation residual is calculated for each operational data record of each type of facility: the environmental data (rainfall, temperature, humidity) corresponding to the data record is substituted into the regression model to obtain the predicted value of the operation parameter. The actual operation parameter value at the same timestamp is subtracted from the predicted value to obtain the residual of a single data point (e.g., if the actual pipeline flow of the water supply and drainage facility at a certain moment is Q1 and the model predicts the flow is Q2, the residual is Q1-Q2). All data records are traversed to generate environmental impact facility type operation residual data, which includes the fields "unique facility identifier - timestamp - actual operation value - predicted operation value - residual value". The generated residual data is quantified: statistical indicators of residuals for each type of facility under different environmental scenarios are calculated, including residual standard deviation (obtained by taking the square root of the sum of the squares of the differences between all residual values ​​and the residual mean, reflecting the overall degree of residual fluctuation), maximum residual (the maximum absolute value of all residual values ​​under a certain scenario, reflecting extreme deviation), and residual fluctuation frequency (the number of times the residual value exceeds the range of "residual mean ± 1 standard deviation" per unit time, reflecting the frequency of residual instability). These quantitative indicators are associated with facility type and environmental scenario to generate quantitative data on the residual characteristics of environmental impact on facility operation.

[0035] Step S233: Map the facility operation characteristic data of environmental impact to the municipal facility type linkage distribution data to perform linkage operation characteristic analysis of municipal facilities with environmental impact, and generate facility linkage operation characteristic data with environmental impact; In this embodiment of the invention, a node association table is extracted from the municipal facility type linkage distribution data. This table includes unique facility identifiers, associated node pairs (such as water supply and drainage network nodes and downstream pumping station nodes, lighting facility nodes and adjacent traffic signal nodes), and association weights. The facility operation characteristic data of environmental impact is matched with the nodes in the node association table according to the facility's unique identifier, thereby achieving a precise mapping between environmental impact operation characteristics and linkage nodes. For each pair of associated nodes, a time series alignment method is used to analyze the linkage response relationship of the associated facility operation characteristics under environmental scene changes. For example, when the increase in rainfall causes the flow of the water supply and drainage network node (node ​​A) to rise in a certain trend, the rotation speed change of the downstream pumping station node (node ​​B) is tracked simultaneously. The time delay from the start of the flow change of node A to the rotation speed response of node B is recorded, and the ratio of the flow change amplitude of node A to the rotation speed change amplitude of node B (reflecting the linkage strength) is calculated. At the same time, the environmental scene corresponding to this linkage relationship is marked (such as moderate rain scene, heavy rain scene). All associated node pairs are classified according to facility type combination (water supply and drainage - water supply and drainage, lighting - traffic signal), and the linkage response delay, linkage intensity mean and linkage stability (linkage intensity fluctuation range) of different type combinations in various environmental scenarios are statistically analyzed to generate facility linkage operation characteristic data of environmental impact.

[0036] Step S234: Based on the characteristic data of the coordinated operation of facilities with environmental impact, perform quantitative analysis of the residual characteristics of the coordinated operation of facilities with environmental impact under each coordinated operation, and generate quantitative data of the residual characteristics of the coordinated operation of facilities with environmental impact. In this embodiment of the invention, based on the determined linkage relationship of related nodes, a linkage operation prediction model is constructed: using the predicted value of the operating parameters of the active node (such as the water supply and drainage network node) in a certain linked node pair (generated by the regression model in step S231) as input, the prediction relationship between the operating parameters of the active node and the operating parameters of the passive node (such as the pump station node) is established through a linear fitting method, thus obtaining the linkage operation prediction model. For the actual operating data of each pair of linked nodes, the linkage operation residual of environmental impact facilities under collaborative operation is calculated: the actual environmental impact operating parameters of the active node are substituted into the linkage operation prediction model to obtain the linkage prediction operating value of the passive node; the linkage prediction operating value is subtracted from the actual operating value of the passive node to obtain the linkage residual value; the residual is recorded according to the timestamp and the linked node pair to generate the linkage operation residual data of environmental impact facilities under collaborative operation. The linkage residual data is quantitatively processed as follows: grouped by environmental scenario and associated node type, the residual covariance (reflecting the linear correlation between associated nodes and residual values, derived by calculating the covariance formula of the residuals of two nodes), residual time synchronization (statistical analysis of the percentage of times the residual values ​​of two nodes simultaneously exceed the allowable range, reflecting the synchronization of residual occurrence), and residual decay rate (statistical analysis of the time required for residual values ​​to drop from peak to mean, reflecting the recovery speed of residuals) are calculated. These quantitative indicators are then associated with scenario and node combinations to generate quantitative data on the linkage operation residual characteristics of collaborative environmental impact facilities.

[0037] Step S235: Analyze the state load characteristics of the linkage facilities under each collaborative operation by using the quantitative data of the residual characteristics of the environmental impact facility operation and the quantitative data of the residual characteristics of the linkage operation of the environmental impact facility, and generate the state load characteristic data of the linkage facilities.

[0038] In this embodiment of the invention, based on the safety operation parameters and historical operation and maintenance data in the technical specifications of various municipal facilities, residual characteristic thresholds are determined. The quantitative data of residual characteristics of environmental impact facilities are compared with these thresholds to identify individual facilities requiring state adjustment. The direction and magnitude of the adjustment are analyzed, and the peak, valley, and fluctuation cycles of the load before and after the adjustment are extracted to generate state load characteristic data for a single operating facility. Next, nodes associated with the single adjustment facility are extracted from the coordinated distribution data of municipal facility types. The adjustment magnitude and target load range of the single facility are used as state change conditions, substituted into the quantitative residual data of coordinated operation, and the residual changes of associated facilities are analyzed through time series alignment to determine the adjustment needs of associated facilities, establish the adjustment sequence and residual control targets, and generate state adjustment demand data for coordinated operation facilities. Subsequently, a three-layer convolutional neural network is constructed (the first layer uses a 3×3 convolutional kernel to extract short-term load fluctuation features, the second layer uses a 5×5 convolutional kernel to extract medium-term temporal correlation features, and the third layer uses a 7×7 convolutional kernel to extract long-term spatial correlation features; each layer is processed by ReLU activation and max pooling). The coordinated adjustment demand data is input into the network in timestamp slices to generate multi-layer convolutional feature data. A two-layer fully connected neural network (128 neurons in the first layer and 64 neurons in the second layer) is then constructed. The convolutional features are concatenated, flattened, and then input into the network for fusion to generate fully connected state feature data. Finally, based on the fully connected feature data, the load coordination coefficient (cosine similarity of the feature vectors of associated nodes), load safety margin ((rated load - actual load) / rated load), and energy efficiency correlation (Pearson correlation coefficient between load adjustment and energy efficiency change) are calculated for each collaborative scenario. Combining the collaborative scenario and the type of associated node, the state load feature data of the linkage facility is generated.

[0039] Furthermore, step S232 includes the following steps: Based on the operational characteristic data of facility types with environmental impact, residual analysis of the operation of environmental impact facilities is performed to generate residual data of the operation of environmental impact facility types. Based on the residual data of environmental impact facility operation, the residual characteristics of environmental impact facility operation are quantified to generate quantitative data of residual characteristics of environmental impact facility operation.

[0040] In this embodiment of the invention, an environmental impact multiple linear regression model established for various types of facilities is invoked to obtain the regression equation of the model (e.g., the flow regression equation for water supply and drainage facilities: flow = α × rainfall + β × temperature + γ, where α, β, and γ are the solved regression coefficients). For each operational data record of each type of facility (including timestamp, actual operational parameter values, and corresponding environmental data), the environmental data (rainfall and temperature) of the record is substituted into the regression equation to calculate the predicted value of the operational parameters at that timestamp. The calculated predicted value is subtracted from the actual operational parameter value of the facility at the same timestamp to obtain the environmental impact facility operation residual for that data point. All operational data records of this type of facility under a certain environmental scenario are traversed, and the residual of each record is calculated one by one. The residual data is organized in the order of timestamp to generate environmental impact facility type operation residual data, ensuring that each residual value can be traced back to the specific environmental conditions and operational status. Based on the environmental impact facility type operational residual data generated in the previous sub-steps, the data is grouped according to facility type (water supply and drainage, lighting, traffic signals) and environmental scenario (rainfall, high temperature, normal), ensuring that each group of data corresponds to the residual record of a specific facility under a specific environment. The quantitative indicators of residual characteristics for each group of data are calculated, including: calculating the residual mean, calculating the residual standard deviation, statistically analyzing the residual extreme values, and calculating the residual fluctuation period. These quantitative indicators are then correlated with the corresponding facility type and environmental scenario to form structured data, generating quantitative data on the operational residual characteristics of environmental impact facilities.

[0041] Furthermore, step S234 includes the following steps: Based on the characteristic data of the coordinated operation of facilities with environmental impact, residual analysis of the coordinated operation of facilities with environmental impact is performed under various collaborative operations to generate residual data of the coordinated operation of facilities with environmental impact. Based on the residual data of the coordinated operation of environmental impact facilities, the residual characteristics of the coordinated operation of environmental impact facilities under each coordinated operation are quantified to generate quantitative data of the residual characteristics of the coordinated operation of environmental impact facilities.

[0042] In this embodiment of the invention, based on the characteristic data of the coordinated operation of facilities with environmental impact, a linkage relationship model of associated node pairs is extracted (e.g., the linkage relationship between water supply and drainage network node A and pump station node B: node B rotation speed = δ × node A flow rate + ε, where δ and ε are coefficients fitted based on historical linkage data). For a certain collaborative operation scenario (e.g., water supply and drainage-pump station coordination during rainfall), the actual operation data of the associated node pairs in this scenario (actual flow rate of node A, actual rotation speed of node B) is obtained, and the data is aligned by timestamp. The actual flow rate of node A is substituted into the linkage relationship model to calculate the predicted rotation speed of node B. The predicted rotation speed is subtracted from the actual rotation speed of node B to obtain the linkage operation residual at that timestamp. All timestamp data of the associated node pairs in this collaborative scenario are traversed, and the linkage operation residual of each data point is calculated one by one, and then sorted in chronological order to generate the linkage operation residual data of environmental impact facilities for collaborative operation. Based on the residual data of the coordinated operation of environmental impact facilities generated in the previous sub-steps, the data is grouped according to the coordination scenario (rainfall coordination, high temperature coordination, morning and evening peak coordination) and the type of associated node pair (water supply and drainage - water supply and drainage, water supply and drainage - lighting, lighting - traffic signal). For each group of data, quantitative indicators of the linkage residual characteristics are calculated: residual covariance (calculated using a formula to measure the covariance of the residual values ​​of the two nodes in the associated node pair, reflecting the degree of linear correlation between the residuals of the two nodes); residual synchronization rate (calculated as the proportion of times the residual values ​​of two nodes simultaneously exceed the range of "residual mean ± 1 standard deviation" out of the total number of data points, reflecting the temporal synchronicity of residual occurrences); residual correction efficiency (calculated as the average time required for residual values ​​to recover from exceeding the threshold to within the threshold, reflecting the recovery speed of the residuals); and residual distribution concentration (calculated using a histogram to show the distribution ratio of residual values ​​in different intervals, reflecting the central tendency of the residuals). These quantitative indicators are then correlated with the coordination scenario and the type of associated node pair to generate quantitative data of the residual characteristics of the coordinated operation of environmental impact facilities.

[0043] Furthermore, step S235 includes the following steps: Based on the quantitative data of residual characteristics of environmental impact facilities operation, the state adjustment demand analysis of individual operating facilities is carried out to generate state adjustment demand data of individual operating facilities. Then, the load characteristic analysis of the state adjustment demand data of individual operating facilities is carried out to generate state load characteristic data of individual operating facilities. Based on the single operating facility status load characteristic data as the condition for facility status change, the residual characteristics of the coordinated operation of the environmental impact facilities are quantified to analyze the environmental impact of the coordinated operation facility status adjustment demand data. Based on the data on the adjustment requirements of the status of facilities for coordinated operation, multi-layer convolutional feature analysis of the facility status of coordinated operation characteristics is performed to generate multi-layer convolutional feature data of the status of facilities for coordinated operation. Based on the multi-layer convolutional feature data of the status of the linked operation facilities, a fully connected feature fusion analysis of the linked operation characteristics is performed to generate fully connected status feature data of the linked operation facilities. Based on the full connectivity status characteristic data of the interconnected facilities, the status load characteristic data of the interconnected facilities under each collaborative operation is analyzed, and the status load characteristic data of the interconnected facilities is generated.

[0044] In this embodiment of the invention, residual characteristic thresholds for various facilities are determined. These thresholds are based on the safe operation parameters and historical operation and maintenance data in the facility's technical specifications. The quantitative data of environmental impact facility operation residual characteristics are compared with the corresponding thresholds. If the residual standard deviation, maximum residual, or fluctuation frequency of a facility exceeds the threshold (e.g., the residual standard deviation of a water supply and drainage network exceeds the threshold by 20%), it is determined that the facility has a status adjustment requirement. The adjustment direction (e.g., if the residual is positive and exceeds the threshold, it is determined that the load needs to be reduced to decrease the deviation between the actual operating value and the predicted value) and the adjustment magnitude (calculated based on the proportion exceeding the threshold; for example, if it exceeds 20%, the load needs to be reduced by 15%) are analyzed, generating single-operation facility status adjustment requirement data. For this adjustment requirement data, load characteristics before and after adjustment are extracted: the peak, valley, and fluctuation period of the load before adjustment are calculated, the target range of the load after adjustment is predicted (calculated by combining the adjustment magnitude and the current load value), and the load change trend corresponding to the adjustment requirement is analyzed, generating single-operation facility status load characteristic data. Nodes associated with a single adjustment facility are extracted from the coordinated distribution data of municipal facilities (e.g., downstream pumping station nodes associated with water supply and drainage network nodes, and adjacent traffic signal nodes associated with lighting facility nodes). The "adjustment range - target load range" from the single operational facility's state load characteristic data is used as the state change condition, and substituted into the quantitative data of the residual characteristics of the coordinated operation of environmental impact facilities. A time series alignment method is used to analyze the changes in residual characteristics of associated facilities under state change conditions: for example, if the water supply and drainage network needs to reduce its load by 15% (target load range determined), the residual standard deviation and residual synchronization rate changes of the downstream pumping station under this condition are calculated simultaneously to determine that the associated facility also has an adjustment need. The correlation relationship, adjustment sequence, and residual control target after adjustment are recorded to generate data on the state adjustment requirements of coordinated operation facilities. A three-layer convolutional neural network was constructed for feature extraction. The first layer, with a kernel size of 3×3 and a stride of 1, was used to extract short-term fluctuation features of the load on interconnected facilities (e.g., load fluctuations within 1 hour). The second layer, with a kernel size of 5×5 and a stride of 2, was used to extract medium-term temporal correlation features (e.g., load response delay of interconnected facilities within 6 hours). The third layer, with a kernel size of 7×7 and a stride of 3, was used to extract long-term spatial correlation features (e.g., load coordination patterns of interconnected facilities in different regions within 24 hours). Each convolutional layer was followed by a ReLU activation function to enhance the non-linear expression of the features, and a max-pooling layer (with a 2×2 kernel size) was used to reduce the feature dimensionality. The data on the adjustment requirements of the interconnected facility status were sorted by timestamp to form time-series data (5-minute intervals), and sliced ​​into data blocks of fixed duration (e.g., 24 hours) as network input.The network uses convolutional operations to transform information such as load adjustment magnitude and associated response time in the time series into feature maps. The first layer outputs a short-term fluctuation feature map (including the time of load peak occurrence and fluctuation magnitude), the second layer outputs a medium-term correlation feature map (including the distribution of response delays of associated facilities), and the third layer outputs a long-term spatial feature map (including the trend of multi-regional coordinated load changes), generating multi-layer convolutional feature data of the status of interconnected facilities. A two-layer fully connected neural network is constructed, with 128 neurons in the first layer and 64 neurons in the second layer. Each layer uses the ReLU activation function, and the output layer is a 64-dimensional feature vector. The short-term, medium-term, and long-term feature maps from the multi-layer convolutional feature data of the status of interconnected facilities are concatenated according to the channel dimension. The two-dimensional feature maps are transformed into one-dimensional feature vectors through flattening (e.g., the short-term feature map dimension is 128×128, and after flattening, it is a 16384-dimensional vector), which is then input into the first layer of the fully connected neural network. The first layer maps the high-dimensional vector to a 128-dimensional space using matrix multiplication, filtering out redundant features. The second layer further maps the 128-dimensional vector to a 64-dimensional space, fusing convolutional features from different layers (such as the weight allocation between short-term fluctuation features and long-term spatial features, determined through gradient descent optimization during network training). The fused 64-dimensional feature vector contains comprehensive information on the load fluctuations, temporal correlations, spatial coordination, and adjustment needs of the interconnected facilities. Each dimension corresponds to a specific feature meaning (e.g., the 10th dimension corresponds to the weight of the linkage adjustment magnitude, and the 25th dimension corresponds to the load coordination stability), generating fully connected state feature data for the interconnected operation facilities. Based on the fully connected status characteristic data of the interconnected facilities, core load indicators are calculated for each collaborative operation scenario: First, the load coordination coefficient is calculated by taking the cosine similarity of the associated nodes to the fully connected feature vector, reflecting the degree of synchronization of load changes in the associated facilities; second, the load safety margin is calculated by subtracting the current actual load from the facility's rated load and then dividing by the rated load to obtain a percentage, reflecting the safe redundancy of load operation; finally, the energy efficiency correlation is calculated by combining energy efficiency data and analyzing the correlation between load adjustment magnitude and energy efficiency changes using the Pearson correlation coefficient. These indicators are then associated with collaborative scenarios (such as rainfall coordination, morning and evening peak coordination), associated node pair types (water supply and drainage - water supply and drainage, lighting - traffic signal), and fully connected feature vectors. The mean, maximum value, and trend of each indicator under different scenarios are statistically analyzed to generate interconnected facility status load characteristic data.

[0045] Furthermore, as an embodiment of the present invention, reference is made to... Figure 2 As shown, Figure 1 A detailed flowchart illustrating the implementation steps of step S3 is provided in this embodiment. Step S3 includes: Step S31: Collect prior data on the status and load characteristics of the linkage facilities based on the status and load characteristic data of the linkage facilities to obtain prior status and load characteristic data of the linkage facilities. In this embodiment of the invention, based on the status and load characteristic data of interconnected facilities, the scope and content of prior data collection are determined: historical status and load characteristic data of interconnected facilities in the central urban area of ​​the corresponding smart city are collected, while simultaneously collecting municipal facility environmental data (such as daily rainfall, average temperature, and humidity) and operation and maintenance records (such as equipment maintenance time, manual load adjustment records, and fault handling logs) within the same time period. During the collection process, the data is classified according to the "unique identifier of the interconnected facility group," and each group contains complete time-series information. The time interval is consistent with the generation frequency of the load characteristic data to ensure the continuity of the data sequence. The collected raw data is filtered to remove abnormal data segments, and finally, the prior data on the status and load characteristics of interconnected facilities is obtained.

[0046] Step S32: Based on the prior data of the state load characteristics of the linkage facilities, perform spatial transmission intensity dependency analysis of the state of the linkage facilities in order to design the spatial transmission dependency graph of the state of the linkage facilities. In this embodiment of the invention, the geographical coordinates and load characteristic indicators of each type of interconnected facility group are extracted from the prior data on the state and load characteristics of interconnected facilities, and a spatial transmission intensity dependency analysis is conducted. First, the spatial adjacency relationship of the interconnected facilities is analyzed. By calculating the latitude and longitude difference between the center coordinates of the facility groups, the distance is converted to a straight-line distance on the ground. Facility groups with a distance ≤ 500 meters are identified as spatially adjacent, and the adjacent facility pairs and their actual distances are recorded. Second, functional coupling is analyzed. The Pearson correlation coefficient calculation method is used to perform correlation analysis on the load synergy coefficients of adjacent facility groups. A correlation coefficient with an absolute value ≥ 0.7 is considered strongly coupled, while a correlation coefficient between 0.3 and 0.7 is considered weakly coupled. Spatial adjacency relationship data and functional coupling data are generated respectively. Based on these two types of data, spatial association characteristic analysis is conducted: spatial distance is converted into an adjacency weight in the range of 0-1 (the closer the distance, the greater the weight, calculated using a linear mapping formula). The functional coupling correlation coefficient is used as the coupling weight, and the spatial association strength of each pair of facility groups is obtained through weighted summation. The Granger causality test method was adopted, using the load characteristic change data of one facility group as the independent variable and the load characteristic change data of another related facility group as the dependent variable. The F-test was used to determine whether the independent variable had a predictive effect on the dependent variable (P < 0.05 indicates a causal relationship), generating spatial transitive causal relationship data. A spatial transitive dependency graph of the interconnected facility states was constructed using the Neo4j graph database. Nodes in the graph represent interconnected facility groups, and node attributes include facility group ID, facility type, and current load safety margin. Edge attributes between nodes include spatial association strength, causal relationship marker (existence / non-existence), and functional coupling type (strong / weak), completing the spatial transitive dependency graph design.

[0047] Step S33: Perform load autocorrelation time series characteristic analysis of the linkage facility status based on the prior linkage facility status load characteristic data, and generate linkage facility status load autocorrelation time series characteristic data. In this embodiment of the invention, time series data of load coordination coefficients are extracted from prior data on the load characteristics of linked facilities, categorized by linked facility group (timestamps accurate to the minute, 30 days of data for a specific facility group are continuously extracted). An ARIMA model is used to conduct time series analysis of load autocorrelation characteristics: the time series data is tested for stationarity (using the ADF unit root test; if the P-value < 0.05, the data is considered stationary; otherwise, differencing is performed until stationarity is achieved); model parameters (number of autoregressive terms p, difference times d, number of moving average terms q) are determined using the AIC information criterion. For example, by iterating through combinations of p = 0-5, d = 0-2, q = 0-5, the combination with the smallest AIC value is selected as the final parameters (e.g., p = 3, d = 1, q = 2). Based on the established ARIMA model, the autocorrelation coefficients of the load coordination coefficient of the facility group at different time scales are calculated: by calculating the correlation coefficients between the time series with different lag orders (e.g., lags of 1 hour, 6 hours, and 24 hours, corresponding to lag steps of 12, 72, and 288) and the original series, the repetitive variation patterns of load characteristics in different time periods are analyzed. Simultaneously, the periodicity of load characteristics is statistically analyzed, such as identifying the principal period of the load coordination coefficient through power spectral density analysis, generating time-series autocorrelation characteristic data of the linked facility status load.

[0048] Step S34: Map the autocorrelation time series feature data of the state load of the linkage facility to the transitive dependency graph of the state space of the linkage facility, perform load cross-correlation time series feature analysis of the state of the linkage facility, and generate cross-correlation time series feature data of the state load of the linkage facility. In this embodiment of the invention, causally related pairs of facilities (such as facility group A and facility group B) are extracted from the transitive dependency graph of the interconnected facility state space. The autocorrelation time-series characteristic data of the load of facility group A (including autocorrelation coefficients and main periods at different time scales) and the autocorrelation time-series characteristic data of facility group B are aligned by timestamps to achieve the mapping of autocorrelation time-series characteristics to the dependency graph. For each pair of related facility groups, the load cross-correlation time-series characteristics are calculated: using the sliding window method (window duration set to 1 hour), the cross-correlation coefficients of the load coordination coefficients of facility group A and facility group B within the window are calculated. The changing trend of the cross-correlation coefficients under different windows is analyzed to determine the optimal response lag time of the load characteristics between related facility groups (e.g., after the load of facility group A changes, facility group B shows the maximum cross-correlation coefficient after 30 minutes, i.e., the response lag time is 30 minutes). At the same time, the mean and fluctuation amplitude of the cross-correlation coefficients under different environmental scenarios (e.g., rain / no rain, weekday / weekend) are statistically analyzed to reflect the impact of the environment on the load cross-correlation characteristics of related facilities. By associating related facility pairs, response lag times, and statistical values ​​of cross-correlation numbers under different scenarios, time-series characteristic data of cross-correlation between the status and load of linked facilities are generated.

[0049] Step S35: Analyze the load time-series change characteristics of the inter-correlation time-series characteristic data of the inter-linkage facility status load to generate inter-linkage facility status load change characteristic data. In this embodiment of the invention, the load coordination coefficient time series of each group of data (including the time series of facility group A and facility group B) is extracted from the time series characteristic data of the cross-correlation between the status and load of the interconnected facilities. The time series is then decomposed into a trend term (long-term trend), a seasonal term (periodic fluctuation), and a residual term (random disturbance) to analyze the load time series change characteristics under spatial status transmission: For the trend term, the synchronization rate of trend changes between the interconnected facility groups is calculated (e.g., the overlap ratio between the upward trend periods of facility group A and facility group B); for the seasonal term, the phase difference of the seasonal cycles of the interconnected facility groups is compared (e.g., the difference between the seasonal peak occurrence time of facility group A and facility group B); for the residual term, the synchronization fluctuation frequency of the residuals of the interconnected facility groups is statistically analyzed (the number of times the residuals of both exceed the allowable range simultaneously per unit time). Simultaneously, key changing parameters under different scenarios are extracted: for example, in a rainfall scenario, the percentage increase in the peak load coordination coefficient of the associated facility group compared to a rainless scenario, and the extent of the extension of the peak duration; in morning and evening peak scenarios, the rate of increase and rate of decrease of the load coordination coefficient. These characteristic parameters are associated with the associated facility group pairs and scenario types to generate data on the load change characteristics of the linked facilities.

[0050] Step S36: Establish a load trend mapping relationship of the linkage facility status based on the load change characteristic data of the linkage facility status, so as to obtain the load trend prediction model of the linkage facility status. In this embodiment of the invention, a load trend prediction model for linked facilities is constructed based on the load change characteristic data of the linked facilities. The model adopts a Long Short-Term Memory (LSTM) network architecture: the input layer dimension is set to 8 (corresponding to 8 key parameters in the load change characteristic data: trend synchronization rate, seasonal phase difference, residual synchronization fluctuation frequency, peak change ratio, response lag time, mean cross-correlation coefficient, autocorrelation coefficient, and load safety margin); two hidden layers are set, with the first layer containing 256 neurons and the second layer containing 128 neurons, both using tanh as the activation function; the output layer contains 72 neurons, corresponding to the predicted load synergy coefficient value for each hour in the next 72 hours. The model is trained using the Adam optimizer (initial learning rate set to 0.001, decaying to 0.9 every 100 training rounds). After each training round, the loss value of the validation set is calculated. Training stops when the loss value of the validation set does not decrease significantly (decrease < 0.0001) for 50 consecutive rounds. After training, the mean absolute error (MAE) and root mean square error (RMSE) of the model are calculated using the test set to ensure that MAE < 0.05 and RMSE < 0.08, which meets the prediction accuracy requirements. Finally, the linkage facility status load trend prediction model is obtained.

[0051] Step S37: Transmit the status feature data of the linkage facility to the load trend prediction model of the linkage facility status for load feature trend prediction processing, and generate load feature trend data of the linkage facility status.

[0052] In this embodiment of the invention, key parameters of the linked facility group to be predicted are extracted from the current status load characteristic data of the linked facilities. These parameters include the current load coordination coefficient, load safety margin, and energy efficiency correlation. Simultaneously, the correlation data (current load status of the linked facility group and spatial correlation strength) of the current environmental data (real-time rainfall and temperature) and the spatial transitivity graph are acquired. These parameters are then organized into an 8-dimensional input vector according to the dimensional requirements of the model input layer. The input vector is transmitted to the trained linked facility status load trend prediction model. The model passes the vector to the hidden layer through the input layer: the first layer LSTM neurons extract long-term dependency features of the input parameters to capture the long-term trend of load changes; the second layer LSTM neurons further extract short-term fluctuation features and combine spatial correlation information to correct the prediction results; the output layer outputs the predicted load coordination coefficient for each hour within the next 72 hours through a linear activation function. Post-processing of the prediction results: Combining the scenario type parameters in step S35, the predicted scenarios for different time periods are labeled; simultaneously, based on the mapping relationship between the load coordination coefficient and the load safety margin, the predicted load safety margin value for the next 72 hours is derived, and periods with load safety margins below a threshold (e.g., 0.8) are marked as high-risk periods. Trend data on the status load characteristics of the linked facilities are generated.

[0053] Furthermore, step S32 includes the following steps: Step S321: Analyze the spatial adjacency relationship and functional coupling of the linkage facilities based on the prior state load characteristic data of the linkage facilities, and obtain the spatial adjacency relationship data and functional coupling data of the linkage facilities respectively. In this embodiment of the invention, the geographical coordinates and unique identifiers of each group of linked facilities are extracted from the prior load characteristic data of the linked facilities. Simultaneously, the load synergy coefficient (a key parameter reflecting the degree of load synergy within the facility group) of the core load characteristic indicators of each group is extracted. When analyzing spatial adjacency, the latitude and longitude distance calculation formula is used to calculate the distance between the center coordinates of any two groups of linked facilities. The calculation result is compared with a preset spatial adjacency threshold. If the ground distance between the two groups of facilities is ≤500 meters, they are considered spatially adjacent, and the spatial adjacency relationship data of the linked facilities is output. When analyzing functional coupling, the load synergy coefficient time series of two groups of facilities over 30 consecutive days is used as the analysis object. The Pearson correlation coefficient calculation method is used. By calculating the ratio of the product of the covariance and standard deviation of the two time series, the functional coupling correlation coefficient is obtained. The absolute value of the correlation coefficient is compared with the coupling strength threshold (strong coupling ≥0.7, weak coupling 0.3-0.7, no coupling <0.3) to determine the coupling strength level of each group of associated facilities, generating the functional coupling data of the linked facilities.

[0054] Step S322: Based on the spatial adjacency relationship data and functional coupling data of the interconnected facilities, perform spatial association feature analysis of the interconnected facilities to generate spatial association feature data of the interconnected facilities; In this embodiment of the invention, a spatial association feature analysis model is constructed based on spatial adjacency relationship data and functional coupling data of interconnected facilities. The "actual distance" in the spatial adjacency relationship data is weighted: a linear mapping formula is used to convert the distance value into an adjacency weight in the range of 0-1, with the mapping rule being "the smaller the distance, the greater the weight." Then, the "coupling correlation coefficient" in the functional coupling data is directly used as the coupling weight. For strong coupling (≥0.7), the coupling weight takes the original value of the correlation coefficient; for weak coupling (0.3-0.7), it takes "correlation coefficient × 0.8" to reduce the weight ratio. Subsequently, the spatial association strength of each group of interconnected facilities is calculated using a weighted summation formula, and the type combination of the interconnected facility groups is marked to generate spatial association feature data of interconnected facilities.

[0055] Step S323: Analyze the causal relationship of the state space transmission of the linked facilities through the spatial association feature data of the linked facilities, and generate causal relationship data of the state space transmission of the linked facilities; In this embodiment of the invention, spatial association feature data of interconnected facilities are used to screen pairs of interconnected facilities with a spatial association strength ≥ 0.5. Causal relationship analysis is then performed on each selected pair. The Granger causality test is employed, using the time series of the load synergy coefficient of one pair of facilities as the independent variable and the time series of the load synergy coefficient of the other pair as the dependent variable. The stationarity of the two time series is tested (using the ADF unit root test; if the data is non-stationary, first-order differencing is performed until stationarity is achieved), and the lag order is determined. The F-statistic and P-value of the null hypothesis "the independent variable is not a Granger cause of the dependent variable" are calculated using the F-test. If the P-value < 0.05, the null hypothesis is rejected, and the independent variable facility group is determined to be a Granger cause of the dependent variable facility group, indicating a one-way causal relationship. If the two sets of facilities are Granger causes of each other, a two-way causal relationship is determined. The causal direction (one-way / two-way), F-statistic, P-value, and test results for each pair of interconnected facilities are recorded to generate state space transit causal relationship data for interconnected facilities.

[0056] Step S324: Design the spatial transit intensity dependency graph of the state of the linked facilities based on the causal relationship data of the state space of the linked facilities, and obtain the state space transit dependency graph of the linked facilities.

[0057] In this embodiment of the invention, a graph database is used to construct a state space transitive dependency graph for interconnected facilities. First, the node attributes of the graph are defined: each node corresponds to a group of interconnected facilities, the node ID is a unique identifier for the facility group, and the node attributes include "facility type - center latitude and longitude coordinates - current load safety margin - number of facilities (number of individual facilities included in the group)". The node location is determined in the map coordinate system based on latitude and longitude coordinates. Then, the edge attributes of the graph are defined: each edge corresponds to a pair of interconnected facilities with a causal relationship, and the edge attributes include "spatial association strength - causal direction - coupling strength level - Granger test p-value". The edge thickness is set according to the spatial association strength. Through the visualization function of the graph database, nodes are distributed according to their actual geographical locations, and edges are connected according to their association relationships. Detailed attribute data can be viewed by clicking on nodes / edges, thus completing the design of the state space transitive dependency graph for interconnected facilities. This graph can intuitively present the spatial transitive relationships and dependency strengths between facility groups.

[0058] Furthermore, step S4 includes the following steps: Step S41: Collect energy efficiency data of the linked facilities by using the status load characteristic data of the linked facilities, and generate energy efficiency data of the linked facilities; In this embodiment of the invention, energy efficiency data of the linked facilities is collected through load characteristic data of the linked facilities and energy efficiency monitoring equipment. All energy efficiency monitoring equipment communicates with the city-level IoT platform via the LoRaWAN protocol. The default collection frequency is set to 5 minutes / time. When the "load coordination coefficient change amplitude > 5%" in the load characteristic data, the platform automatically sends an instruction to the corresponding energy efficiency equipment to increase the collection frequency to 1 minute / time, ensuring the timeliness of energy efficiency data when the load fluctuates, and generating energy efficiency data of the linked facilities.

[0059] Step S42: Based on the load characteristic data and energy efficiency data of the linkage facilities, perform load and energy efficiency correlation analysis of the linkage facilities to generate load and energy efficiency correlation data of the linkage facilities. In this embodiment of the invention, the energy efficiency data and load characteristic data of the linked facilities are aligned using the dual key fields of "linked facility group ID - timestamp" to construct a load-energy efficiency correlation analysis dataset. By calculating the ratio of the product of the covariance and standard deviation of the two time series over 30 days, the correlation coefficient for different load intervals is obtained, clarifying the impact of increased load on energy consumption. For the lighting facility group, using "load safety margin" as the load index and "average power per lamp" as the energy efficiency index, the data is analyzed by environmental scenario grouping. The linear regression equation between the load index and the energy efficiency index within each group is calculated to obtain a quantitative correlation. For the traffic signal facility group, the correlation between load coordination and energy efficiency is analyzed, and the energy consumption changes under different signal cycles are statistically analyzed to generate linked facility load-energy efficiency correlation data.

[0060] Step S43: Based on the preset reinforcement learning algorithm and the load energy efficiency correlation data of the linkage facilities, design the initial model of multi-agent decision-making for load and energy efficiency under various collaborative operations, so as to obtain the initial load energy efficiency multi-agent decision-making model. In this embodiment of the invention, the initial model is constructed using the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm. Each group of interconnected facilities corresponds to one agent, with a total of 30 agents covering all core interconnected facilities in the central urban area of ​​the smart city. The state space of the agents includes five types of features: "load coordination coefficient, load safety margin, real-time energy consumption, cumulative energy consumption, and environmental scene labeling," with a dimension of 32. The action space is defined as "load adjustment amplitude." The model network structure includes a policy network and a value network: the policy network uses three fully connected layers with 256, 128, and 64 neurons respectively, using ReLU as the activation function, and the output layer uses the tanh function to map action values ​​to a preset range; the value network structure is consistent with the policy network, and the output is a state-action value. Using the load-energy efficiency correlation data of the linked facilities as training samples, the training set is used to iteratively update the model parameters, and the validation set is used to evaluate the model fit. The mean squared error loss is minimized by the Adam optimizer (initial learning rate 0.001). After 2000 iterations of training, an initial load-energy efficiency multi-agent decision model with basic load-energy efficiency decision-making ability is generated.

[0061] Step S44: Set multi-agent hard constraint targets based on the preset facility operation parameter safety boundary, and set multi-agent soft optimization targets based on global energy efficiency optimization and facility load balancing; In this embodiment of the invention, the safety boundary of the facility's operating parameters is extracted as the hard constraint objective for the multi-agent system. The hard constraint objective is achieved through model parameter restrictions, ensuring that no action value output by any decision leads to the parameters exceeding the boundary. The soft optimization objective is divided into global energy efficiency optimization and facility load balancing: the global energy efficiency optimization objective is determined by calculating the "total energy efficiency ratio = total functional output / total energy consumption" of all interconnected facilities in the central urban area, with the objective being that this ratio is 15% higher than the historical average; the facility load balancing objective is determined by calculating the "absolute value of the deviation between the load coordination coefficient of each interconnected facility group and the global average load coordination coefficient", with the objective being that the absolute value of the deviation for all groups is <10%. The soft optimization objective is reflected in the reward function of the subsequent model training.

[0062] Step S45: Based on the multi-agent hard constraint objective and the multi-agent soft optimization objective, perform multi-agent objective empowerment processing on the initial load energy efficiency multi-agent decision model to generate the load energy efficiency multi-agent decision model.

[0063] In this embodiment of the invention, the initial load energy efficiency multi-agent decision-making model is target-empowered based on hard constraints and soft optimization objectives. A hard constraint penalty term is added to the model's reward function: when the load adjustment output by an agent causes facility operating parameters to exceed hard constraint boundaries (e.g., pipeline pressure < 0.3 MPa), the reward value is directly reduced by 100, and the action is forcibly invalidated; if the parameters are within the constraint range, the penalty term is counted as 0. A soft optimization reward term is added: when the global energy efficiency ratio increases by 1%, all participating agents receive a total of 10 reward points; when the absolute value of the load coordination coefficient deviation of a certain facility group is < 10%, that agent receives 5 reward points, with an additional 1 reward point added for every 1% reduction in deviation. The load energy efficiency correlation data from step S42 is used to supplement the training samples, and the model is trained for another 3000 rounds. After each round of training, the validation set is used to test whether the model meets the requirements of "no breakthrough in hard constraints and achievement rate of soft optimization objectives ≥90%". If the model meets the requirements for 100 consecutive rounds of testing, training is stopped. If the requirements are not met, the weight of the soft optimization objective in the reward function is adjusted (e.g., the global energy efficiency reward weight is increased from 0.4 to 0.5), and the model is retrained until the requirements are met, thus generating a load energy efficiency multi-agent decision model that meets both safety and optimization objectives.

[0064] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0065] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for analyzing the status and operation of municipal facilities in smart cities, characterized in that, Includes the following steps: Step S1: Through the city-level Internet of Things platform and the sensing devices integrated with the geographic information system, the municipal facilities are monitored and preprocessed to generate multi-source monitoring data of municipal facilities, which includes municipal facility operation data, municipal facility environmental data and municipal facility geographical location data. Step S2: Based on the multi-source monitoring data of municipal facilities, perform status load characteristic analysis of the interconnected facilities under various collaborative operations, and generate status load characteristic data of the interconnected facilities; Step S3: Perform load characteristic trend prediction processing on the load characteristic data of the linkage facility status to generate load characteristic trend data of the linkage facility status. Step S4: Collect energy efficiency data of the linked facilities by using the status load characteristic data of the linked facilities, and generate energy efficiency data of the linked facilities; establish a multi-agent decision model of load and energy efficiency under each collaborative operation based on the status load characteristic data and energy efficiency data of the linked facilities, and generate a load energy efficiency multi-agent decision model. Step S5: Transmit the status load characteristic trend data of the linked facilities to the load energy efficiency multi-agent decision model for global energy efficiency optimization and analyze the intelligent control parameters of the linked facilities to generate intelligent control parameters of the linked facilities; execute the status operation and maintenance control of municipal facilities in the smart city through the intelligent control parameters of the linked facilities.

2. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: The city-level Internet of Things platform and the sensing devices integrated with the geographic information system are used to perform multi-source data monitoring and processing of municipal facilities to generate preliminary multi-source monitoring data of municipal facilities. The city-level Internet of Things platform receives the monitoring information of the sensing facilities, and the geographic information system determines the location information of the sensing facilities. Step S12: Obtain municipal facility data monitoring logs from the sensing facilities to obtain municipal facility data monitoring logs; Step S13: Perform data preprocessing on the preliminary multi-source monitoring data of municipal facilities based on the municipal facility data monitoring log to generate multi-source monitoring data of municipal facilities.

3. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Analyze the types of municipal facilities based on the municipal facility operation data to generate municipal facility type data; Step S22: Establish a distribution analysis of municipal facility types based on the geographical location data and type data of municipal facilities, generate distribution data of municipal facility types, and perform a linkage distribution analysis of municipal facility types based on the distribution data of municipal facility types to generate linkage distribution data of municipal facility types. Step S23: Map municipal facility operation data and municipal facility environmental data to municipal facility type linkage distribution data to perform linkage facility status load characteristic analysis under various collaborative operations, and generate linkage facility status load characteristic data.

4. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 3, characterized in that, Step S23 includes the following steps: Step S231: Based on the environmental data of municipal facilities, perform an environmental impact facility operation characteristic analysis on the municipal facility operation data to generate environmental impact facility operation characteristic data; Step S232: Based on the environmental impact facility type operation characteristic data, perform quantitative analysis and processing of environmental impact facility operation residual characteristics to generate quantitative data of environmental impact facility operation residual characteristics; Step S233: Map the facility operation characteristic data of environmental impact to the municipal facility type linkage distribution data to perform linkage operation characteristic analysis of municipal facilities with environmental impact, and generate facility linkage operation characteristic data with environmental impact; Step S234: Based on the characteristic data of the coordinated operation of facilities with environmental impact, perform quantitative analysis of the residual characteristics of the coordinated operation of facilities with environmental impact under each coordinated operation, and generate quantitative data of the residual characteristics of the coordinated operation of facilities with environmental impact. Step S235: Analyze the state load characteristics of the linkage facilities under each collaborative operation by using the quantitative data of the residual characteristics of the environmental impact facility operation and the quantitative data of the residual characteristics of the linkage operation of the environmental impact facility, and generate the state load characteristic data of the linkage facilities.

5. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 4, characterized in that, Step S232 includes the following steps: Based on the operational characteristic data of facility types with environmental impact, residual analysis of the operation of environmental impact facilities is performed to generate residual data of the operation of environmental impact facility types. Based on the residual data of environmental impact facility operation, the residual characteristics of environmental impact facility operation are quantified to generate quantitative data of residual characteristics of environmental impact facility operation.

6. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 4, characterized in that, Step S234 includes the following steps: Based on the characteristic data of the coordinated operation of facilities with environmental impact, residual analysis of the coordinated operation of facilities with environmental impact is performed under various collaborative operations to generate residual data of the coordinated operation of facilities with environmental impact. Based on the residual data of the coordinated operation of environmental impact facilities, the residual characteristics of the coordinated operation of environmental impact facilities under each coordinated operation are quantified to generate quantitative data of the residual characteristics of the coordinated operation of environmental impact facilities.

7. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 4, characterized in that, Step S235 includes the following steps: Based on the quantitative data of residual characteristics of environmental impact facilities operation, the state adjustment demand analysis of individual operating facilities is carried out to generate state adjustment demand data of individual operating facilities. Then, the load characteristic analysis of the state adjustment demand data of individual operating facilities is carried out to generate state load characteristic data of individual operating facilities. Based on the single operating facility status load characteristic data as the condition for facility status change, the residual characteristics of the coordinated operation of the environmental impact facilities are quantified to analyze the environmental impact of the coordinated operation facility status adjustment demand data. Based on the data on the adjustment requirements of the status of facilities for coordinated operation, multi-layer convolutional feature analysis of the facility status of coordinated operation characteristics is performed to generate multi-layer convolutional feature data of the status of facilities for coordinated operation. Based on the multi-layer convolutional feature data of the status of the linked operation facilities, a fully connected feature fusion analysis of the linked operation characteristics is performed to generate fully connected status feature data of the linked operation facilities. Based on the full connectivity status characteristic data of the interconnected facilities, the status load characteristic data of the interconnected facilities under each collaborative operation is analyzed, and the status load characteristic data of the interconnected facilities is generated.

8. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Collect prior data on the status and load characteristics of the linkage facilities based on the status and load characteristic data of the linkage facilities to obtain prior status and load characteristic data of the linkage facilities. Step S32: Based on the prior data of the state load characteristics of the linkage facilities, perform spatial transmission intensity dependency analysis of the state of the linkage facilities in order to design the spatial transmission dependency graph of the state of the linkage facilities. Step S33: Perform load autocorrelation time series characteristic analysis of the linkage facility status based on the prior linkage facility status load characteristic data, and generate linkage facility status load autocorrelation time series characteristic data. Step S34: Map the autocorrelation time series feature data of the state load of the linkage facility to the transitive dependency graph of the state space of the linkage facility, perform load cross-correlation time series feature analysis of the state of the linkage facility, and generate cross-correlation time series feature data of the state load of the linkage facility. Step S35: Analyze the load time-series change characteristics of the inter-correlation time-series characteristic data of the inter-linkage facility status load to generate inter-linkage facility status load change characteristic data. Step S36: Establish a load trend mapping relationship of the linkage facility status based on the load change characteristic data of the linkage facility status, so as to obtain the load trend prediction model of the linkage facility status. Step S37: Transmit the status feature data of the linkage facility to the load trend prediction model of the linkage facility status for load feature trend prediction processing, and generate load feature trend data of the linkage facility status.

9. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 8, characterized in that, Step S32 includes the following steps: Step S321: Analyze the spatial adjacency relationship and functional coupling of the linkage facilities based on the prior state load characteristic data of the linkage facilities, and obtain the spatial adjacency relationship data and functional coupling data of the linkage facilities respectively. Step S322: Based on the spatial adjacency relationship data and functional coupling data of the interconnected facilities, perform spatial association feature analysis of the interconnected facilities to generate spatial association feature data of the interconnected facilities; Step S323: Analyze the causal relationship of the state space transmission of the linked facilities through the spatial association feature data of the linked facilities, and generate causal relationship data of the state space transmission of the linked facilities; Step S324: Design the spatial transit intensity dependency graph of the state of the linked facilities based on the causal relationship data of the state space of the linked facilities, and obtain the state space transit dependency graph of the linked facilities.

10. The method for analyzing the status and operation of municipal facilities for smart cities according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Collect energy efficiency data of the linked facilities by using the status load characteristic data of the linked facilities, and generate energy efficiency data of the linked facilities; Step S42: Based on the load characteristic data and energy efficiency data of the linkage facilities, perform load and energy efficiency correlation analysis of the linkage facilities to generate load and energy efficiency correlation data of the linkage facilities. Step S43: Based on the preset reinforcement learning algorithm and the load energy efficiency correlation data of the linkage facilities, design the initial model of multi-agent decision-making for load and energy efficiency under various collaborative operations, so as to obtain the initial load energy efficiency multi-agent decision-making model. Step S44: Set multi-agent hard constraint targets based on the preset facility operation parameter safety boundary, and set multi-agent soft optimization targets based on global energy efficiency optimization and facility load balancing; Step S45: Based on the multi-agent hard constraint objective and the multi-agent soft optimization objective, perform multi-agent objective empowerment processing on the initial load energy efficiency multi-agent decision model to generate the load energy efficiency multi-agent decision model.

Images