A tree heat network heat source stability control type PID self-tuning method and device

Abstract

The application discloses a tree-shaped heat network heat source stability control type PID self-tuning method and equipment, and the method comprises the following steps: filtering the real-time heat network operation condition data of a heat supply system to obtain filtered operation condition data; performing fuzzy dynamic constraint calculation on the filtered operation condition data to obtain a safety constraint condition of PID parameter updating; based on the safety constraint condition, iteratively updating the PID parameter through a deep Q network to obtain a global reference PID parameter; determining the branch weight coefficient of each heat exchange station branch in the heat supply system, and based on the branch weight coefficient of each heat exchange station branch, adaptively scaling and correcting the global reference PID parameter to obtain the PID control parameter of each heat exchange station branch; and based on the PID control parameter of each heat exchange station branch, adjusting the primary network valve of each heat exchange station in the heat supply system. In this way, the stability, uniformity and intelligent level of the heat supply system can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of intelligent automatic control technology for centralized heating, and in particular to a heat source stabilization type PID self-tuning method and equipment for a tree-like heating network. Background Technology

[0002] Urban centralized heating systems generally adopt a single heat source tree-like pipe network structure. The primary network connects the heat source to each heat exchange station, and the secondary network connects the heat exchange station to the user. During on-site operation and maintenance, the secondary side valves are kept fixed after commissioning, and heat distribution is completed solely by the primary network valves. The heat source equipment has high requirements for operational stability, and frequent and large fluctuations in flow and temperature are prohibited. The system uniformly adopts a fixed frequency of 30 seconds to collect operational data.

[0003] PID control is the mainstream control method for heating systems. Among related technologies, various PID self-tuning techniques have obvious defects. Traditional PID and ordinary fuzzy PID lack specific stable constraints for heat sources, which can easily cause fluctuations in heat sources. Furthermore, PID schemes based on reinforcement learning are mostly used in scenarios such as building air conditioning and multi-heat source pipe networks. The industry urgently needs a PID self-tuning method that takes into account dynamic constraints, autonomous learning capabilities, and adaptability to on-site working conditions. Summary of the Invention

[0004] This application provides a heat source stabilization-type PID self-tuning method and device for a tree-like heating network, which solves the following technical problem: how to improve the stability, uniformity and intelligence level of the heating system.

[0005] In a first aspect, embodiments of this application provide a heat source stabilization-type PID self-tuning method for a tree-like heating network, the method comprising: The real-time operating condition data of the heating system is obtained, and the real-time operating condition data is filtered to obtain the filtered operating condition data. Fuzzy dynamic constraint calculations are performed on the filtered operating condition data to obtain the safety constraint conditions for updating the PID parameters; Based on the aforementioned safety constraints, the PID parameters are iteratively updated using a deep Q-network to obtain global baseline PID parameters. The branch weight coefficient of each heat exchange station branch in the heating system is determined, and the global benchmark PID parameters are adaptively scaled and corrected based on the branch weight coefficient of each heat exchange station branch to obtain the PID control parameters of each heat exchange station branch. Based on the PID control parameters of each heat exchange station branch, the primary network valves of each heat exchange station in the heating system are adjusted.

[0006] Secondly, embodiments of this application also provide a heat source stabilization type PID self-tuning device for a tree-like heating network. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a heat source stabilization type PID self-tuning method for a tree-like heating network as described above.

[0007] Thirdly, embodiments of this application also provide a computer storage medium storing computer-executable instructions, which, when executed, implement a heat source stabilization type PID self-tuning method for a tree-like heating network as described above.

[0008] The heat source stabilization type PID self-tuning method and device for a tree-like heating network provided in this application embodiment have the following beneficial effects: First, by acquiring and filtering real-time operating data of the heating network, instantaneous noise and random interference during sensor acquisition can be effectively eliminated, providing a stable and reliable data foundation for subsequent control decisions and avoiding erroneous control due to data jumps. Furthermore, based on this, fuzzy dynamic constraint calculations are performed on the filtered data, dynamically generating a safety boundary for PID parameter updates according to load deviations and heat source fluctuation trends. This limits the magnitude and frequency of parameter adjustments from the source. Combined with a subsequent parameter freezing mechanism, this ensures accurate room temperature control while prioritizing the suppression of drastic fluctuations in heat source flow and temperature, thus addressing the pain point of traditional PID control neglecting the stability control requirements of heat source equipment. Subsequently, within the safety constraint boundary, the PID parameters are iteratively updated using a deep Q-network, enabling the system to achieve more stable and reliable control. The system can autonomously learn and approximate the optimal global baseline PID parameters in a data-driven manner without the need to build a complex pipeline network mechanism model. It can balance the adaptive capability under complex and variable operating conditions with the computing power limitations of industrial edge controllers. Then, by determining the weight coefficients of each heat exchange station branch and adaptively scaling and correcting the global baseline PID parameters, it can effectively reduce the coupling interference between branches and improve the problem of uneven heating in the entire network by addressing the differences in heat load demand and pipeline transmission delay among different branches in the compensation tree network. Finally, by precisely adjusting the primary network valves according to the PID control parameters of each branch, it can achieve differentiated and reasonable distribution of heat in the entire network. While ensuring stable operation on the heat source side and maintenance-free secondary valves, it can improve the uniformity of heating and the overall energy efficiency of the system. Attached Figure Description

[0009] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1A flowchart of a heat source stabilization-type PID self-tuning method for a tree-like heating network provided in this application embodiment; Figure 2 This is a schematic diagram of the internal structure of a heat source stabilization type PID self-tuning device for a tree-shaped heating network provided in an embodiment of this application. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] It is understood that in the embodiments of this application, data related to user information (such as user accounts) is involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with relevant laws, regulations and standards.

[0012] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0013] In the following description, the terms “first, second, ...” are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that “first, second, ...” may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0014] Urban centralized heating systems generally adopt a single heat source tree-like pipe network structure. The primary network connects the heat source to each heat exchange station, and the secondary network connects the heat exchange station to the user. During on-site operation and maintenance, the secondary side valves are kept fixed after commissioning, and heat distribution is completed solely by the primary network valves. The heat source equipment has high requirements for operational stability, and frequent and large fluctuations in flow and temperature are prohibited. The system uniformly adopts a fixed frequency of 30 seconds to collect operational data.

[0015] PID control is the mainstream control method for heating systems. Among related technologies, various PID self-tuning techniques have obvious defects. Traditional PID and ordinary fuzzy PID lack specific stable constraints for heat sources, which can easily cause fluctuations in heat sources. Furthermore, PID schemes based on reinforcement learning are mostly used in scenarios such as building air conditioning and multi-heat source pipe networks. The industry urgently needs a PID self-tuning method that takes into account dynamic constraints, autonomous learning capabilities, and adaptability to on-site working conditions.

[0016] Based on this, this application provides a heat source stabilization-type PID self-tuning method for tree-shaped heating networks. It ensures stable operation of the heat source through fuzzy dynamic constraints and parameter freezing mechanism, realizes model-free self-tuning of PID parameters using deep Q network, and eliminates hydraulic coupling interference of tree-shaped pipe network by branch differentiation correction, which can effectively improve the stability, uniformity and intelligence level of heating system.

[0017] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0018] Figure 1 This application provides a flowchart of a heat source stabilization-type PID self-tuning method for a tree-like heating network. This method can be applied to various fields of intelligent automatic control for centralized heating, such as the renovation of existing buildings' centralized heating systems, heating systems with stringent requirements for heat source stability, tree-like heating networks with complex topologies and significant branch differences, and existing heating scenarios lacking accurate network mechanism models. The process can be executed by computing devices in the relevant field (e.g., smart mobile terminals), and some input parameters or intermediate results in the process can be manually adjusted to help improve accuracy.

[0019] This application provides a heat source stabilization-type PID self-tuning method for tree-like heating networks. It should be noted that the execution entity in these embodiments can be a server or any terminal device with data processing capabilities. For example, the server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal device can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, vehicle terminal, etc., but is not limited to these.

[0020] like Figure 1 As shown in the embodiment of this application, a heat source stabilization-type PID self-tuning method for a tree-like heating network specifically includes the following steps: Step 101: Obtain real-time operating data of the heating network of the heating system.

[0021] Here, you can set the sampling period. The system collects on-site operational data (i.e., real-time heating network operating condition data) every 30 seconds at a fixed frequency. The real-time heating network operating condition data may include heat source flow rate, supply and return water temperature, primary network valve opening at heat exchange stations, unit room temperature measurement point data, and outdoor ambient temperature.

[0022] Step 102: Filter the real-time heating network operating condition data to obtain filtered operating condition data.

[0023] Here, the filtering process can be implemented using a first-order moving average filter to eliminate instantaneous interference. For details, please refer to formula (1).

[0024] (1) in, Sampling time, for Real-time operating status data of the heating network at all times. for The operating condition data after time-lapse filtering. For example, the length of the filtering window. This corresponds to four 30-second sampling periods.

[0025] Step 103: Perform fuzzy dynamic constraint calculation on the filtered operating condition data to obtain the safety constraint conditions for updating the PID parameters.

[0026] In some embodiments, the filtered operating condition data includes load deviation and heat source fluctuation trend; step 103 above can be implemented in the following way: quantize and map the load deviation to obtain a first fuzzy value, and quantize and map the heat source fluctuation trend to obtain a second fuzzy value; based on a preset fuzzy rule, perform fuzzy mapping on the first fuzzy value and the second fuzzy value to obtain a third fuzzy value corresponding to the amplitude coefficient and a fourth fuzzy value corresponding to the period coefficient; perform defuzzification operation on the third fuzzy value and the fourth fuzzy value to obtain a real-time amplitude coefficient and a real-time period coefficient, and construct the safety constraints for updating the PID parameters based on the real-time amplitude coefficient and the real-time period coefficient.

[0027] Thus, on the one hand, fuzzy logic is used to transform precise operating parameters into constraint boundaries that adapt to different operational states. The maximum adjustment ratio and update frequency of PID parameters can be dynamically adjusted according to the magnitude of load deviation and the intensity of heat source fluctuations, thereby achieving an adaptive balance between room temperature control requirements and heat source equipment temperature control requirements. On the other hand, through the fuzzy inference mechanism, the rigidity problem of traditional fixed threshold constraints can be effectively avoided, suppressing the violent fluctuations in heat source flow and temperature caused by parameter mutations or frequent iterations from the source. This can significantly reduce fatigue damage and safety risks of thermal equipment. At the same time, defuzzification operations can ensure that the fuzzy inference results are transformed into precise constraint values ​​that can be directly applied in engineering, providing reliable safety constraints for the subsequent parameter self-tuning of deep Q-networks, thereby ensuring the safe and stable operation and long-term stability of the entire heating system under complex and variable operating conditions.

[0028] It should be noted that load deviation refers to the difference between the actual room temperature and the preset target room temperature; heat source fluctuation trend refers to the amount and direction of change of the heat source water supply temperature within a single sampling period, which is used to reflect the current operational stability of the heat source.

[0029] Here, load deviation and heat source fluctuation trend are used as inputs. Safety constraints (including maximum adjustment range and minimum parameter update cycle) of PID parameters are dynamically generated through preset fuzzy rules to define the safety boundary for PID parameter optimization and ensure stable operation of the heat source.

[0030] In some embodiments, the above-described quantization mapping of the load deviation to obtain a first fuzzy value can be achieved in the following ways: in response to the absolute value of the load deviation being greater than or equal to a first temperature threshold, the first fuzzy value is determined to be large; in response to the absolute value of the load deviation being less than the first temperature threshold and the absolute value of the load deviation being greater than a second temperature threshold, the first fuzzy value is determined to be medium; in response to the absolute value of the load deviation being less than or equal to the second temperature threshold, the first fuzzy value is determined to be small.

[0031] In this way, by using clear threshold divisions to transform continuous and precise physical quantities into discrete fuzzy linguistic variables, we can retain the ability of load deviations to represent the degree of room temperature deviation, while eliminating the interference of instantaneous measurement noise or small fluctuations on fuzzy inference. This allows the system to stably distinguish different operating conditions. At the same time, the setting of the two-level thresholds is closely aligned with heating operation standards, ensuring that the fuzzy quantization results are consistent with empirical knowledge. This provides an accurate and reliable input basis for subsequent fuzzy rule inference, thereby supporting the system to make reasonable parameter constraint decisions under different load deviations, thus balancing room temperature control accuracy and heat source operation stability.

[0032] Here, the standard room temperature setting is 20℃. Deviation levels are categorized according to heating operation standards, and the actual physical deviations are mapped to a unified fuzzy domain. 3,3]; When the absolute value of the room temperature deviation (load deviation) is ≤0.5℃ (second temperature threshold), the room temperature is close to the set value, the operating condition is stable, the corresponding fuzzy set is "small" (first fuzzy value), and the mapping domain interval is [ 3, 1]; When the absolute value of the room temperature deviation (load deviation) is between 0.5℃ (second temperature threshold) and 1.5℃ (first temperature threshold), the room temperature has a moderate deviation, corresponding to the fuzzy set "middle" (first fuzzy value), and the mapping domain interval [ 1,1]; When the absolute value of the room temperature deviation (load deviation) is ≥1.5℃ (first temperature threshold), the room temperature deviation is serious and the user experience is poor, corresponding to the fuzzy set "large" (first fuzzy value), and the mapping domain interval [1,3].

[0033] In some embodiments, the above-described quantization mapping of the heat source fluctuation trend to obtain a second fuzzy value can be achieved in the following ways: in response to the heat source fluctuation trend being a temperature decrease greater than or equal to a temperature threshold, the second fuzzy value is determined to be decreasing; in response to the heat source fluctuation trend being a temperature change less than the temperature threshold, the second fuzzy value is determined to be stable; in response to the heat source fluctuation trend being a temperature increase greater than or equal to the temperature threshold, the second fuzzy value is determined to be increasing.

[0034] In this way, by using the temperature change within a single sampling period as the basis for judgment, the short-term dynamic trend of the heat source water supply temperature can be accurately captured, and the operating status of the heat source can be identified in a timely manner. By setting the absolute value threshold, minute temperature fluctuations under normal operating conditions can be effectively filtered out, avoiding misjudgments caused by instantaneous disturbances, and ensuring that only changes with real trend significance will trigger the corresponding constraint strategy. Furthermore, the fuzzy classification method of temperature fluctuation trend and the fuzzy classification method of load deviation complement each other, so that the fuzzy rules can simultaneously take into account the needs of the user side and the safety of the heat source side. Thus, when the heat source shows a significant cooling or heating trend, the parameter adjustment authority is automatically tightened to protect the heat source equipment, while the constraints are appropriately relaxed when the heat source is stable to improve the control sensitivity. Ultimately, an adaptive balance is achieved between priority control of the heat source and dynamic adjustment of room temperature.

[0035] Here, the change in the water supply temperature of the heat source within a single 30-second sampling period is used as the criterion for fluctuation judgment (heat source fluctuation trend). When the temperature decrease (temperature decrease value) within the period is ≥0.3℃ / 30s (temperature threshold), the heat source shows a significant cooling trend, corresponding to the fuzzy set "decline" (second fuzzy value), and the mapping domain interval [ 3, 1]; When the absolute value of temperature change within the period (the absolute value of the amount of temperature change) < 0.3℃ / 30s (temperature threshold), the heat source operates stably, corresponding to the fuzzy set "stable" (second fuzzy value), and the mapping domain interval [ [1,1]; When the temperature rises (temperature rise value) within the period is ≥0.3℃ / 30s (temperature threshold), the heat source shows a clear heating trend, corresponding to the fuzzy set "rise" (second fuzzy value), and the mapping domain interval [1,3].

[0036] In some embodiments, after performing the above-described safety constraints for constructing the PID parameter update based on the real-time amplitude coefficient and the real-time period coefficient, the following processing can also be performed: determining the temperature fluctuation standard deviation based on the heat source water supply temperature in the filtered operating condition data; determining that parameter freezing is triggered in response to the temperature fluctuation standard deviation being greater than a preset fluctuation threshold; and determining that parameter freezing is not triggered in response to the temperature fluctuation standard deviation being less than or equal to the preset fluctuation threshold.

[0037] Thus, by utilizing the standard deviation statistical method within a sliding window, the true fluctuation intensity of the heat source operation can be accurately extracted from the continuous temperature sequence, effectively distinguishing between short-term random noise and continuous severe oscillations, thereby avoiding false freezing caused by instantaneous temperature jumps. When the standard deviation of temperature fluctuation exceeds a preset threshold, parameter freezing is immediately triggered, forcibly pausing the iterative update of the deep Q network and the adjustment of PID parameters. This allows for the rapid cutoff of control actions that may lead to positive feedback deterioration under extreme operating conditions where the heat source is extremely unstable, providing a rigid safety barrier for the heat source equipment. The automatic unfreezing after the fluctuation returns to normal ensures that the system can continue to perform adaptive learning and optimization for most of the time, fundamentally solving the safety hazard of traditional PID control blindly adjusting when the heat source is on the verge of instability, which could lead to the expansion of accidents. This significantly improves the robustness and safety of the heating system under complex and variable operating conditions.

[0038] Here, to protect the heat source under extreme operating conditions, a parameter freezing mechanism is added, based on the standard deviation of the heat source temperature fluctuation. To determine operational stability, please refer to formula (2); at the same time, set a fluctuation safety threshold. (Preset fluctuation threshold), fixed value for engineering projects. The temperature is 1.0℃. The operating condition is determined based on the standard deviation of the heat source fluctuation. If the heat source fluctuation exceeds the limit, freeze all current PID parameters, pause DQN iterative learning, and only perform basic valve adjustment, thus triggering parameter freezing; if The heat source is operating smoothly, the freeze is lifted, and the normal parameter update and self-tuning process is restored, meaning that parameter freeze was not triggered.

[0039] (2) in, For the standard deviation of temperature fluctuation, Let t be the filtered water supply temperature from the heat source. This represents the average temperature within the sliding window. This is for the statistical window length.

[0040] In some embodiments, the above-described fuzzy mapping of the first and second fuzzy values ​​based on preset fuzzy rules to obtain a third fuzzy value corresponding to the amplitude coefficient and a fourth fuzzy value corresponding to the period coefficient can perform the following processing: In response to the first fuzzy value being small and the second fuzzy value decreasing, determine that the third fuzzy value is medium and the fourth fuzzy value is medium; in response to the first fuzzy value being small and the second fuzzy value being stable, determine that the third fuzzy value is small and the fourth fuzzy value is large; in response to the first fuzzy value being small and the second fuzzy value increasing, determine that the third fuzzy value is small and the fourth fuzzy value is large; in response to the first fuzzy value being medium and the second fuzzy value decreasing, determine the third fuzzy value... The fuzz value is medium and the fourth fuzz value is medium; in response to the first fuzz value being medium and the second fuzz value being stable, the third fuzz value is determined to be medium and the fourth fuzz value is determined to be medium; in response to the first fuzz value being medium and the second fuzz value increasing, the third fuzz value is determined to be medium and the fourth fuzz value is determined to be medium; in response to the first fuzz value being large and the second fuzz value decreasing, the third fuzz value is determined to be large and the fourth fuzz value is determined to be small; in response to the first fuzz value being large and the second fuzz value being stable, the third fuzz value is determined to be large and the fourth fuzz value is determined to be small; in response to the first fuzz value being large and the second fuzz value increasing, the third fuzz value is determined to be large and the fourth fuzz value is determined to be small.

[0041] Real-time amplitude coefficient This limits the maximum adjustment ratio of the PID parameters in a single operation, and sets the output domain to... Real-time period coefficient This defines the minimum interval between two PID parameter updates, in units of sampling periods, and sets the output domain to [value missing]. .

[0042] In some embodiments, the above-mentioned defuzzification operation on the third fuzzy value and the fourth fuzzy value to obtain the real-time amplitude coefficient and the real-time period coefficient can be achieved by formula (3).

[0043] (3) in, To provide accurate output values ​​after deblurring (including real-time amplitude coefficient and real-time period coefficient). Let j be the output universe value corresponding to the j-th rule. Let j be the membership degree corresponding to the j-th rule. To determine the effective number of fuzzy rules, here, .

[0044] Obtaining the real-time amplitude coefficient through defuzzification and real-time period coefficient Construct safety constraints for PID parameter updates, as shown in formula (4).

[0045] (4) in, , and This represents the single adjustment amount of the PID parameters. This is the sampling time for this parameter update. This is the sampling time of the last parameter update.

[0046] Step 104: Based on the security constraints, the PID parameters are iteratively updated using a deep Q-network to obtain the global baseline PID parameters.

[0047] In some embodiments, step 104 described above can be implemented as follows: In response to the absence of parameter freezing, a multi-dimensional operating condition state vector is constructed based on the load deviation, the standard deviation of heat source fluctuation, the outdoor ambient temperature in the filtered operating condition data, and the PID parameters; the following processes are continuously executed until parameter freezing is triggered: the multi-dimensional operating condition state vector is forward-computed using the deep Q-network to obtain the Q value corresponding to each candidate PID parameter adjustment action; the target PID parameter adjustment action is determined based on the safety constraints and the Q value corresponding to each candidate PID parameter adjustment action; the multi-dimensional operating condition state vector is updated based on the target PID parameter adjustment action; in response to the triggering of parameter freezing, the PID parameters in the final multi-dimensional operating condition state vector are used as the global baseline PID parameters.

[0048] Thus, by comprehensively characterizing the four operating conditions of user-side demand, heat source stability, environmental disturbances, and current control strategies through multi-dimensional state vectors, deep Q-networks can make decisions based on a complete system operation profile, thereby significantly improving the environmental perception capability and action rationality of reinforcement learning in complex heating scenarios. At the same time, embedding safety constraints into the action selection process ensures that all candidate PID adjustment actions undergo dual verification of amplitude and frequency after being evaluated by the Q-network, eliminating the risk of parameter mutations or high-frequency oscillations from the source. Furthermore, using parameter freezing as the trigger condition for training termination and baseline output ensures that the system can approach the globally optimal PID parameters through continuous iteration most of the time, and can decisively lock the current optimal solution when the heat source is on the verge of instability, avoiding the risk of runaway, thereby ultimately achieving deep synergy between model-free data-driven and engineering safety constraints.

[0049] Here, we first construct a multi-dimensional working condition state vector. The sampling period is updated every 30 seconds to comprehensively characterize the real-time operating status of the heating network. For details, please refer to formula (5). Then, the agent's actions are... Defined as the small iterative adjustment amount of the PID parameters, which can be referred to in formula (6) for details. All actions are strictly limited by the upper-level fuzzy constraints to prevent parameter mutations. At the same time, the action space is combined with the constraint conditions (formula (4)). Real-time interval truncation ensures the safety of regulation; subsequently, to prioritize the stability of the heat source while considering room temperature accuracy and system stability, a weighted comprehensive reward function is constructed. For details, please refer to formula (7), where a larger reward value indicates a better current control state, a more stable heat source, and a smaller load deviation; then, an industrial lightweight dual neural network architecture is adopted, including the current Q network. With the target Q network To adapt to the computing power of edge controllers and avoid computing power redundancy and training oscillations, the network structure is "input layer (6-dimensional state vector) → hidden layer 1 (64 neurons, ReLU activation) → hidden layer 2 (32 neurons, ReLU activation) → output layer (corresponding to the Q value of all actions, linear activation)". The dual-network division of labor mechanism of updating weights in real time through the current network and synchronizing weights with delay in the target network effectively solves the problems of learning oscillation and non-convergence of traditional Q network. Here, the mean squared error loss function is used to complete the backpropagation training of the network (refer to formulas (8) and (9)). Finally, the current state is input every 30 seconds. ,use Greedy strategy selects the optimal action For details, please refer to formula (10) and perform PID parameter adjustment to obtain the new state. With real-time rewards ,Will Store the data in the experience replay pool, batch randomly sample replay samples, update the current Q-network weights, perform fuzzy constraint boundary verification before each action output, and forcibly truncate if the limit is exceeded to ensure heat source safety; when the reward value changes over 20 consecutive sampling periods... When training convergence is determined (or parameter freezing is triggered), output the global baseline PID parameters. When the system is in steady state, the iteration frequency is reduced, and only a small amount of online adaptive optimization is performed.

[0050] (5) (6) (7) (8) (9) (10) in, Let k be the load deviation. Let k be the standard deviation of the heat source fluctuation at time k. Let k be the outdoor ambient temperature. These are the three sets of control parameters for the current PID at time k. , and This represents the single adjustment amount of the PID parameters. The system overshoot at time k is the control overshoot. , and Let be the weight coefficient, and satisfy... Fixed values ​​for engineering ; This represents the current training loss value of the Q-network. This refers to mathematical expectation calculation. This refers to the Q-value output by the target Q-network. This refers to the Q-value output by the current Q-network. That is , That is , That is , That is ; This refers to the state Next, iterate through all possible actions and take the maximum value output by the target Q-network; This is the reward discount factor, representing the weight of future long-term returns, typically 0.9; This is a random number function used to generate a uniformly distributed random number in the interval [0,1). The exploration rate is initially set to 0.9, and decreases linearly to 0.1 during training to balance environmental exploration and optimal decision-making.

[0051] Step 105: Determine the branch weight coefficient of each heat exchange station branch in the heating system.

[0052] In some embodiments, step 105 described above can be implemented as follows: obtaining the average heat load and equivalent pipe network length of each heat exchange station branch in the heating system; for each heat exchange station branch, performing the following processing: taking the ratio of the average heat load of the heat exchange station branch to the overall average heat load of the heating system as the relative heat load percentage of the heat exchange station branch, and taking the ratio of the equivalent pipe network length of the heat exchange station branch to the overall average equivalent length of the heating system as the relative pipe length percentage of the heat exchange station branch; taking the average of the relative heat load percentage and the relative pipe length percentage of the heat exchange station branch as the branch weight coefficient of the heat exchange station branch.

[0053] Thus, by using dimensionless relative proportions to unify the characteristics of different physical quantities into the same dimensional system, the weight bias caused by differences in numerical scale between heat load and pipe length can be eliminated, allowing the weighting coefficients to simultaneously consider the branch characteristics of both dimensions. Furthermore, by simply and efficiently integrating the dual information of load demand and transmission delay through averaging, branches with high loads and long pipe sections receive greater weights to enhance regulation response capabilities, while branches with low loads and short pipe sections receive smaller weights to suppress overshoot and inter-branch coupling interference. At the same time, normalization based on the overall system average value makes the weighting coefficients adaptive across the entire network. Regardless of changes in the scale of the heating system, the weights can accurately reflect the relative position of each branch in the system, thereby providing a scientific and reasonable basis for the subsequent differential scaling of global benchmark PID parameters, effectively improving the hydraulic imbalance and uneven heating problems of the tree-like heating network.

[0054] Here, considering the characteristics of tree-like heat networks—independent branches, large differences in pipe length and load, and susceptibility to hydraulic coupling interference—a lightweight branch weight correction strategy is adopted to achieve differentiated adaptation of PID parameters for each branch. For the i-th heat exchange station branch, the average heat load of the current branch is extracted. Equivalent length of branch network The adaptation weight of the current branch is calculated using both load and pipe length dimensions. For details, please refer to formula (11).

[0055] (11) in, The average heat load of all branches in the network (the overall average heat load of the heating system). The average equivalent pipe length of the entire network (the overall average equivalent length of the heating system).

[0056] Step 106: Based on the branch weight coefficient of each heat exchange station branch, adaptively scale and correct the global benchmark PID parameters to obtain the PID control parameters of each heat exchange station branch.

[0057] In some embodiments, step 106 described above can be implemented as follows: For each heat exchange station branch, the following processing is performed: multiplying the branch weight coefficient of the heat exchange station branch by the proportional coefficient in the global reference PID parameters to obtain the adjusted proportional coefficient; using the ratio of the integral time in the global reference PID coefficient to the branch weight coefficient of the heat exchange station branch as the adjusted integral time; multiplying the branch weight coefficient of the heat exchange station branch by the derivative time in the global reference PID parameters to obtain the adjusted derivative time; and combining the adjusted proportional coefficient, the adjusted integral time, and the adjusted derivative time to obtain the PID control parameters of the heat exchange station branch.

[0058] In this way, the branches with large loads and long pipe sections with weight coefficients greater than 1 obtain stronger proportional and derivative actions to accelerate the response speed to thermal hysteresis and load fluctuations. At the same time, the integral action is relatively enhanced to eliminate steady-state deviations more quickly. Meanwhile, the branches with small loads and short pipe sections with weight coefficients less than 1 have their regulation intensity suppressed to avoid overshoot, oscillation, and coupling interference to other branches. Thus, differentiated control is achieved on the same set of globally optimal PID controllers. Moreover, this asymmetric scaling method has clear physical meaning and extremely low computational cost. It does not require retuning parameters for each branch independently. It can retain the global optimality learned by the deep Q network and naturally compensate for the heterogeneity of heat demand scale and transmission delay among branches in the tree-like heating network. This significantly alleviates the hydraulic imbalance phenomenon of near-end overheating and far-end underheating, thereby effectively improving the uniformity of heating in the whole network and the overall stability of the system.

[0059] Here, the global baseline parameters are adaptively scaled according to the weights to adapt to the different branch lag and inertia characteristics. For details, please refer to formula (12).

[0060] (12) in, The proportionality coefficient characterizing the branch of the i-th heat exchange station The integral time characterizes the i-th heat exchange station branch. The differential time characterizes the branch of the i-th heat exchange station. The proportional coefficient in the global baseline PID parameters. The integral time is the global baseline PID parameter. The derivative time in the global baseline PID parameters. is the weight coefficient of the i-th heat exchange station branch.

[0061] Step 107: Based on the PID control parameters of each heat exchange station branch, adjust the primary network valves of each heat exchange station in the heating system.

[0062] In some embodiments, step 107 described above can be implemented as follows: For each branch of the heat exchange station, the following processing is performed: determining the temperature deviation between the user-side room temperature of the heat exchange station branch and the preset room temperature in the filtered operating condition data; based on the temperature deviation, historical deviation, and the PID control parameters of the heat exchange station branch, determining the opening adjustment amount of the first network valve of the heat exchange station corresponding to the heat exchange station branch through a discrete PID algorithm, so as to adjust the first network valve of the heat exchange station corresponding to the heat exchange station branch based on the opening adjustment amount.

[0063] In this way, by using branch-level temperature deviation as the control input, the adjustment decision of each heat exchange station is directly anchored to the actual heating effect of that branch. This avoids local overcooling or overheating caused by hydraulic imbalance between near and far ends during global unified control. By introducing historical deviations into the integral calculation, long-term accumulated steady-state errors can be automatically eliminated, ensuring that the room temperature can accurately return to the set value when the outdoor temperature fluctuates or the heat load changes. The addition of the derivative term can perform advance suppression based on the deviation change trend, which can effectively reduce overshoot and oscillation. At the same time, this process strictly uses the branch-differentiated PID parameters obtained by weight scaling in the early stage, so that the high-load, long-pipe-section branches can obtain more sensitive valve response to compensate for heat transfer lag, while the low-load, short-pipe-section branches maintain mild regulation to prevent disturbance propagation. Moreover, the process only acts on the primary network valves, and the secondary side valves remain fixed. Without increasing the complexity of on-site operation and maintenance, the heat of each branch of the tree-like heat network can be accurately distributed on demand, ultimately achieving a coordinated control effect of uniform room temperature and stable heat source operation throughout the network.

[0064] Here, a discrete PID algorithm is used to calculate the valve opening adjustment of the primary network. It is compatible with a 30-second discrete sampling system. For details, please refer to formula (13). The secondary valve is fixed throughout the process and does not participate in any adjustment action, which is in complete compliance with the on-site operation and maintenance specifications.

[0065] (13) In some embodiments, for steady-state conditions, i.e., stable heat source fluctuations and slight changes in outdoor temperature, the iteration cycle of the deep Q network is extended to 5 times the sampling cycle, and the number of parameter updates is reduced; for dynamic conditions, i.e., significant load deviations and outdoor temperature fluctuations, the iteration cycle is set to 2 times the sampling cycle to moderately improve the adaptive capability; for high-fluctuation conditions, i.e., triggering the parameter freeze mechanism, only basic valve adjustment is performed, and parameter self-tuning is suspended.

[0066] In some embodiments, the system uses =With a fixed cycle of 30 seconds, the following process is executed cyclically: data filtering preprocessing → fuzzy constraint threshold calculation → DQN (Deep Q Network) parameter self-learning → tree branch parameter correction → primary network valve adjustment → state feedback update, so as to form a fully autonomous closed-loop intelligent control.

[0067] In summary, addressing the shortcomings of traditional fixed-parameter PID control, ordinary fuzzy control, and mechanism modeling control schemes, this application constructs a collaborative control architecture based on fuzzy dynamic constraints, lightweight DQN model-free self-tuning, tree-branch differential correction, and heat source parameter freezing. This architecture is suitable for actual operating conditions with fixed secondary valves, 30-second periodic sampling, and priority heat source control. It significantly improves system stability, operating condition adaptability, energy efficiency, and engineering practicality, specifically including the following beneficial effects: (1) By adaptively adjusting the parameter update amplitude and period through fuzzy dynamic constraints, and combined with the parameter freezing mechanism for heat source fluctuations exceeding limits, the frequent valve operation and heat source condition oscillation can be effectively suppressed, the problem of traditional control focusing on room temperature accuracy and ignoring heat source stability can be improved, the thermal shock of equipment can be reduced, and the system can be guaranteed to operate smoothly. (2) By adopting the DQN modelless data-driven algorithm, there is no need for pipeline mechanism modeling and system identification, and no reliance on precise pipeline infrastructure parameters. The PID parameters can be adaptively tuned using only real-time running data. This is suitable for existing heating scenarios with aging pipelines and time-varying operating conditions, which can reduce the difficulty of on-site commissioning and modification. Its generalization performance is better than traditional mechanism control schemes. (3) In view of the large differences in load, pipe length and time delay of each branch of the tree-shaped heating network, which easily leads to hydraulic coupling interference and heat imbalance, this application makes differential correction to the benchmark PID parameters by combining the branch load and pipe length weights. This can match the operating characteristics of each branch, weaken the coupling interference between branches, improve the uneven heating of the whole network, and thus improve the heat distribution balance. (4) This application sets up a multi-objective reward function that prioritizes heat source stability. By iteratively optimizing the control strategy through reinforcement learning, it can reduce room temperature control deviation and system overshoot while ensuring stable operation of the heat source, reduce ineffective and excessive control behavior, achieve on-demand heat distribution, and thus reduce the energy consumption of pipeline transportation and heat source operation. (5) This application is adapted to the on-site operation and maintenance mode of fixed secondary side valves. There is no need to modify the existing equipment structure. The system can autonomously complete data filtering, constraint calculation, parameter self-tuning and closed-loop control. It can adapt to changes in environment and load conditions, reduce the workload of manual parameter tuning and debugging, and thus improve the level of system automation. (6) This application adopts a lightweight dual-network DQN architecture with low computing power consumption, which is suitable for the deployment conditions of industrial edge controllers. By combining the 30s industrial sampling time sequence with a hierarchical iteration mechanism, it can adapt to different working conditions and update requirements. It has good anti-interference ability and convergence stability, and can meet the long-term continuous control requirements of industrial sites.

[0068] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a heat source stabilization type PID self-tuning device for a tree-like heating network, the structure of which is as follows: Figure 2 As shown.

[0069] Figure 2 This is a schematic diagram of the internal structure of a heat source stabilization type PID self-tuning device for a tree-like heating network, provided as an embodiment of this application. Figure 2 As shown, the device includes: At least one processor 201; And a memory 202 that is communicatively connected to at least one processor; The memory 202 stores instructions executable by at least one processor, which are executed by at least one processor 201 to enable at least one processor 201 to: The real-time operating condition data of the heating system is obtained, and the real-time operating condition data is filtered to obtain the filtered operating condition data. Fuzzy dynamic constraint calculations are performed on the filtered operating condition data to obtain the safety constraint conditions for updating the PID parameters; Based on the aforementioned safety constraints, the PID parameters are iteratively updated using a deep Q-network to obtain global baseline PID parameters. The branch weight coefficient of each heat exchange station branch in the heating system is determined, and the global benchmark PID parameters are adaptively scaled and corrected based on the branch weight coefficient of each heat exchange station branch to obtain the PID control parameters of each heat exchange station branch. Based on the PID control parameters of each heat exchange station branch, the primary network valves of each heat exchange station in the heating system are adjusted.

[0070] Some embodiments of this application provide corresponding to Figure 1 A non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as follows: The real-time operating condition data of the heating system is obtained, and the real-time operating condition data is filtered to obtain the filtered operating condition data. Fuzzy dynamic constraint calculations are performed on the filtered operating condition data to obtain the safety constraint conditions for updating the PID parameters; Based on the aforementioned safety constraints, the PID parameters are iteratively updated using a deep Q-network to obtain global baseline PID parameters. The branch weight coefficient of each heat exchange station branch in the heating system is determined, and the global benchmark PID parameters are adaptively scaled and corrected based on the branch weight coefficient of each heat exchange station branch to obtain the PID control parameters of each heat exchange station branch. Based on the PID control parameters of each heat exchange station branch, the primary network valves of each heat exchange station in the heating system are adjusted.

[0071] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0072] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0073] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0074] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0075] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0076] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0077] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0078] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0079] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0080] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0081] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.

Claims

1. A heat source stabilization-type PID self-tuning method for a tree-like heating network, characterized in that, The method includes: The real-time operating condition data of the heating system is obtained, and the real-time operating condition data is filtered to obtain the filtered operating condition data. Fuzzy dynamic constraint calculations are performed on the filtered operating condition data to obtain the safety constraint conditions for updating the PID parameters; Based on the aforementioned safety constraints, the PID parameters are iteratively updated using a deep Q-network to obtain global baseline PID parameters. The branch weight coefficient of each heat exchange station branch in the heating system is determined, and the global benchmark PID parameters are adaptively scaled and corrected based on the branch weight coefficient of each heat exchange station branch to obtain the PID control parameters of each heat exchange station branch. Based on the PID control parameters of each heat exchange station branch, the primary network valves of each heat exchange station in the heating system are adjusted.

2. The method according to claim 1, characterized in that, The filtered operating condition data includes load deviation and heat source fluctuation trend; The step of performing fuzzy dynamic constraint calculations on the filtered operating condition data to obtain the safety constraints for updating the PID parameters includes: The load deviation is quantized and mapped to obtain a first fuzzy value, and the heat source fluctuation trend is quantized and mapped to obtain a second fuzzy value; Based on preset fuzzy rules, the first fuzzy value and the second fuzzy value are fuzzy mapped to obtain the third fuzzy value corresponding to the amplitude coefficient and the fourth fuzzy value corresponding to the period coefficient. The third and fourth fuzzy values ​​are defuzzified to obtain the real-time amplitude coefficient and the real-time period coefficient. Based on the real-time amplitude coefficient and the real-time period coefficient, the safety constraints for updating the PID parameters are constructed.

3. The method according to claim 2, characterized in that, The step of quantizing and mapping the load deviation to obtain a first fuzzy value includes: In response to the absolute value of the load deviation being greater than or equal to a first temperature threshold, the first fuzzy value is determined to be large; In response to the absolute value of the load deviation being less than the first temperature threshold and the absolute value of the load deviation being greater than the second temperature threshold, the first fuzzy value is determined to be neutral. In response to the absolute value of the load deviation being less than or equal to the second temperature threshold, the first fuzzy value is determined to be small.

4. The method according to claim 2, characterized in that, The step of quantizing and mapping the heat source fluctuation trend to obtain a second fuzzy value includes: In response to the heat source fluctuation trend being that the temperature decrease is greater than or equal to the temperature threshold, the second fuzzy value is determined to be decreasing; In response to the heat source fluctuation trend being that the absolute value of the temperature change is less than the temperature threshold, the second fuzzy value is determined to be stable. In response to the heat source fluctuation trend being that the temperature rise is greater than or equal to the temperature threshold, the second fuzzy value is determined to be rising.

5. The method according to claim 2, characterized in that, After constructing the safety constraints for updating the PID parameters based on the real-time amplitude coefficient and the real-time period coefficient, the method further includes: Based on the heat source water supply temperature in the filtered operating condition data, the standard deviation of temperature fluctuation is determined. In response to the temperature fluctuation standard deviation being greater than a preset fluctuation threshold, the parameter freezing is determined. In response to the temperature fluctuation standard deviation being less than or equal to a preset fluctuation threshold, it is determined that parameter freezing has not been triggered.

6. The method according to claim 5, characterized in that, Based on the aforementioned security constraints, the PID parameters are iteratively updated using a deep Q-network to obtain global baseline PID parameters, including: In response to the absence of parameter freezing, a multi-dimensional operating condition state vector is constructed based on the load deviation, the standard deviation of the heat source fluctuation, the outdoor ambient temperature in the filtered operating condition data, and the PID parameters. Continuously execute the following process until parameter freezing is triggered: The deep Q-network is used to perform forward calculation on the multi-dimensional working condition state vector to obtain the Q value corresponding to the adjustment action of each candidate PID parameter. Based on the aforementioned safety constraints and the Q value corresponding to each candidate PID parameter adjustment action, the target PID parameter adjustment action is determined. Based on the target PID parameter adjustment action, the multi-dimensional operating condition state vector is updated; In response to the trigger parameter freeze, the PID parameters in the final multidimensional operating condition state vector are used as the global baseline PID parameters.

7. The method according to claim 1, characterized in that, Determining the branch weight coefficient of each heat exchange station branch in the heating system includes: Obtain the average heat load and equivalent length of the pipeline network for each branch of the heat exchange station in the heating system; For each of the heat exchange station branches, the following processes are performed: The ratio of the average heat load of the heat exchange station branch to the overall average heat load of the heating system is taken as the relative heat load percentage of the heat exchange station branch, and the ratio of the equivalent length of the pipe network of the heat exchange station branch to the overall average equivalent length of the heating system is taken as the relative pipe length percentage of the heat exchange station branch. The average of the relative heat load ratio of the heat exchange station branch and the relative pipe length ratio of the heat exchange station branch is used as the branch weight coefficient of the heat exchange station branch.

8. The method according to claim 1, characterized in that, The global baseline PID parameters are adaptively scaled and corrected based on the branch weight coefficient of each heat exchange station branch to obtain the PID control parameters for each heat exchange station branch, including: For each of the heat exchange station branches, the following processes are performed: The adjusted proportional coefficient is obtained by multiplying the branch weight coefficient of the heat exchange station branch with the proportional coefficient in the global baseline PID parameters. The ratio of the integral time in the global baseline PID coefficient to the branch weight coefficient of the heat exchange station branch is used as the adjusted integral time. The adjusted differential time is obtained by multiplying the branch weight coefficient of the heat exchange station branch with the differential time in the global benchmark PID parameters. The adjusted proportional coefficient, the adjusted integral time, and the adjusted derivative time are combined to obtain the PID control parameters for the heat exchange station branch.

9. The method according to claim 1, characterized in that, The adjustment of the primary network valves of each heat exchange station in the heating system based on the PID control parameters of each heat exchange station branch includes: For each of the heat exchange station branches, the following processes are performed: Determine the temperature deviation between the user-side room temperature of the heat exchange station branch and the preset room temperature in the filtered operating condition data; Based on the temperature deviation, historical deviation, and PID control parameters of the heat exchange station branch, the opening adjustment amount of the first network valve of the heat exchange station corresponding to the heat exchange station branch is determined by the discrete PID algorithm, so as to adjust the first network valve of the heat exchange station corresponding to the heat exchange station branch based on the opening adjustment amount.

10. A heat source stabilization type PID self-tuning device for a tree-like heating network, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method as described in any one of claims 1-9.

Images