Energy-saving optimization platform for multi-source integrated energy system based on data mining

Abstract

The present application relates to the technical field of data mining, and discloses a multi-source comprehensive energy system energy-saving optimization platform based on data mining, comprising: obtaining characteristic parameters such as upstream energy composition, node flow, settlement reference value and system routing weight through a data acquisition module; a data processing module identifies the propagation characteristics of each node composition, generates an energy characteristic time sequence and constructs a system-level deviation and energy account residual error sequence; a correlation analysis module extracts hysteresis characteristic quantities under the characteristic frequency, establishes the correlation between the hysteresis characteristic quantities and system energy consumption, and generates energy consumption correlation values; an optimization decision module takes system parameters as variables, minimizes energy cost and energy consumption correlation values under operation constraints; and an execution module applies optimal decision results to the system to realize energy saving and operation optimization.

Description

Technical Field

[0001] This invention relates to the field of data mining technology, and more specifically, to a data mining-based energy-saving optimization platform for multi-source integrated energy systems. Background Technology

[0002] Currently, the operation and management of multi-source energy systems rely on energy management systems or integrated energy optimization platforms. These platforms are typically based on static models, performing energy balance calculations with the assumption that the energy conversion factor for each energy carrier remains constant. For example, in a combined electricity-gas-heat (ECH) system, the calorific value of the gas is considered constant; however, in actual operation, this assumption is often distorted. Due to differences in gas sources (such as different natural gas well areas, hydrogen blending ratios, or gas quality fluctuations), the energy content per unit volume varies over time. Furthermore, gas storage and pressure changes in the pipeline network introduce transmission delays, causing discrepancies between the effective energy actually received by downstream users and the upstream settlement figures.

[0003] The above phenomena lead to systemic biases in energy system dispatching and energy efficiency assessment:

[0004] 1. The same energy supply strategy exhibits different energy efficiencies under different composition conditions;

[0005] 2. Energy-saving renovation measures are difficult to verify accurately;

[0006] 3. Cost and energy efficiency settlements across energy types lack a unified and dynamic energy mapping basis.

[0007] As the proportion of renewable energy and hydrogen energy in integrated energy systems increases, the volatility and time-varying nature of gas sources have significantly increased. When the composition of upstream gas sources changes, the calorific value of the gas and its contribution to the overall energy balance also change. However, the current system's scheduling and settlement algorithms do not reflect this change in real time, but instead use a fixed energy conversion factor. Due to the energy transfer relationships between different energy carriers, this deviation accumulates and amplifies within the system. Specifically:

[0008] 1. Power-side load forecasting is based on outdated energy conversion factors;

[0009] 2. The energy distribution on the thermal side does not take into account the real-time changes in the gas calorific value;

[0010] 3. The energy cost term in the economic dispatch objective function is disconnected from the actual energy supply.

[0011] When these discrepancies accumulate, the system exhibits a significant energy mismatch, meaning the energy supplied recorded at the billing layer does not match the effective energy actually consumed by the terminals. This not only prevents the true reflection of energy-saving potential but may also lead to misleading scheduling strategies, resulting in energy waste and economic losses. Summary of the Invention

[0012] This invention provides an energy-saving optimization platform for multi-source integrated energy systems based on data mining, which solves the technical problem of how to identify energy mismatch patterns caused by changes in energy composition from multi-source energy operation data through data mining, and optimize energy allocation and settlement parameters at the system level, thereby reducing energy-saving losses caused by energy deviation.

[0013] This invention provides a data mining-based energy-saving optimization platform for multi-source integrated energy systems, comprising:

[0014] The data acquisition module collects characteristic parameters of the target system, including: upstream energy composition time series, flow time series of each node, settlement reference value time series, and system routing weight time series;

[0015] The data processing module identifies the component propagation characteristic parameters of each node by combining characteristic parameters, generates the node energy characteristic time series, and constructs the system-level deviation time series and the energy reconciliation residual time series.

[0016] The correlation analysis module extracts the characteristic parameters of the system-level deviation time series and the energy reconciliation residual time series at characteristic frequencies to determine the corresponding hysteresis characteristic quantities; based on historical data, it establishes the correlation between the hysteresis characteristic quantities and system energy consumption, and generates energy consumption correlation values.

[0017] The optimization decision module uses the system routing weight time series and the settlement reference value adjustment time series as decision variables. Under system operation constraints, it minimizes the correlation between system energy cost and energy consumption to obtain the target decision variable.

[0018] The execution module applies the target decision variables to the target system.

[0019] The beneficial effects of this invention are as follows: By introducing data mining and energy consumption correlation analysis mechanisms, dynamic perception and intelligent optimization of the operating status of multi-source integrated energy systems are achieved. By identifying changes in energy composition and their transmission characteristics in real time, a system-level energy deviation model is constructed, quantifying the hysteresis law of energy mismatch, thereby dynamically adjusting routing weights and settlement parameters at the optimization decision-making level. Compared with traditional static energy conversion models, this platform can significantly reduce system performance consumption errors caused by gas quality fluctuations and energy delays, achieving unified and dynamic energy reconciliation and scheduling optimization across energy types. Ultimately, this invention can effectively improve the energy-saving level and economic operating efficiency of integrated energy systems, providing data-driven intelligent energy-saving decision support for multi-energy complementary systems such as electricity, gas, and heat. Attached Figure Description

[0020] Figure 1 This is an architecture diagram of the data mining-based multi-source integrated energy system energy-saving optimization platform of the present invention. Detailed Implementation

[0021] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0022] like Figure 1 As shown, the multi-source integrated energy system energy-saving optimization platform based on data mining includes:

[0023] The data acquisition module collects characteristic parameters of the target system, including: upstream energy composition time series, flow time series of each node, settlement reference value time series, and system routing weight time series;

[0024] The data processing module identifies the component propagation characteristic parameters of each node by combining characteristic parameters, generates the node energy characteristic time series, and constructs the system-level deviation time series and the energy reconciliation residual time series.

[0025] The correlation analysis module extracts the characteristic parameters of the system-level deviation time series and the energy reconciliation residual time series at characteristic frequencies to determine the corresponding hysteresis characteristic quantities; based on historical data, it establishes the correlation between the hysteresis characteristic quantities and system energy consumption, and generates energy consumption correlation values.

[0026] The optimization decision module uses the system routing weight time series and the settlement reference value adjustment time series as decision variables. Under system operation constraints, it minimizes the correlation between system energy cost and energy consumption to obtain the target decision variable.

[0027] The execution module applies the target decision variables to the target system.

[0028] Upstream energy composition time series refers to the energy composition of the upstream energy injection end (such as natural gas gate station, hydrogen blending station) in a multi-source integrated energy system. It is formed by continuously collecting time series data at fixed time intervals, reflecting the dynamic changes in the quality purity / blending ratio of the source energy over time.

[0029] Taking the upstream gate station of a natural gas-hydrogen hybrid energy supply system as an example, with a sampling interval of 15 minutes, the time series of upstream energy composition on a certain working day is as follows:

[0030] 08:00-08:15: Hydrogen molar fraction 18%, methane content 81%, other inert gases 1%;

[0031] 08:15-08:30: Hydrogen molar fraction 17.5%, methane content 81.4%, other inert gases 1.1%;

[0032] 08:30-08:45: Hydrogen molar fraction 19%, methane content 80.2%, other inert gases 0.8%.

[0033] The flow time series of each node refers to the time series data formed by continuously collecting energy volume / mass flowing through all key nodes in the system (such as pipeline distribution stations, user entry nodes, and multi-energy conversion nodes) at fixed time intervals, reflecting the dynamic changes in the amount of energy transported / distributed in the system over time.

[0034] Taking three key nodes of a regional integrated energy system as an example, with a sampling interval of 15 minutes, the flow time series for a certain workday is as follows:

[0035] Node 1 (Gate Station Exit, Natural Gas): 200 cubic meters / hour from 08:00 to 08:15, and 220 cubic meters / hour from 08:15 to 08:30 (transmission volume increases during the morning peak).

[0036] Node 2 (Industrial User Inlet, Natural Gas): 80 cubic meters / hour from 08:00 to 08:15, and 85 cubic meters / hour from 08:15 to 08:30 (Industrial Load Start-up).

[0037] Node 3 (electric heating conversion station inlet, natural gas): 60 cubic meters / hour from 08:00 to 08:15, and 75 cubic meters / hour from 08:15 to 08:30 (heating demand increases during the morning peak).

[0038] Settlement reference value time series refers to the benchmark energy parameters used for energy transaction settlement in the system. It is time series data formed by setting or updating according to a fixed time unit (such as hour / day). It provides a unified standard for accounting for the book value of energy. It is usually determined based on industry norms or contractual agreements. Common parameters are volumetric calorific value and mass calorific value.

[0039] Taking the daily settlement reference value of natural gas in a certain industrial park as an example, the settlement parameter is the volumetric calorific value. The time series of the settlement reference value for a certain week is as follows:

[0040] Monday: 36.5 megajoules per cubic meter (based on the regional average atmospheric conditions set on Sunday).

[0041] Tuesday: 36.3 MJ / m³ (adjusted slightly according to CV contraction rules due to a slight increase in upstream hydrogen blending ratio and a decrease in regional average calorific value on Monday).

[0042] Wednesday: 36.4 MJ / m³ (upstream hydrogen blending ratio declined, regional average calorific value rebounded);

[0043] Thursday: 36.2 MJ / m³ (The reference value was lowered due to upstream gas source switching and gas quality fluctuations on the same day).

[0044] System routing weight time series refers to the time series data formed by adjusting the proportion of upstream input energy (such as natural gas) to different load sides (electricity side, heat side, and cold side) in a multi-source integrated energy system at fixed time intervals, reflecting the scheduling strategy of energy destination.

[0045] Taking the natural gas routing weight of a gas-electricity-heat-cooling multi-energy combined system as an example, with a sampling interval of 15 minutes, the time series of routing weights for a certain workday is as follows:

[0046] 07:00-07:15: Electric side 30%, heating side 50%, cooling side 20% (heating demand is high before the morning peak, cooling demand is low).

[0047] 07:15-07:30: Electricity side 40%, heating side 45%, cooling side 15% (Residential electricity demand is rising, increasing the weight of electricity side).

[0048] 07:30-07:45: Electricity demand 45%, heating demand 35%, cooling demand 20% (Electricity demand reaches its peak during the morning rush hour, while heating demand declines).

[0049] 07:45-08:00: Electric side 40%, hot side 30%, cold side 30% (As the ambient temperature rises, the demand for cold side begins to increase).

[0050] In one embodiment of the present invention, the component propagation characteristic parameters of each node are identified by combining characteristic parameters to generate a time series of node energy characteristics, including:

[0051] Calculate the normalized cross-correlation coefficient ρ jd,i (τ):

[0052] ;

[0053] Where C(t) represents the time series of upstream energy components, F jd,i (t+τ) represents the flow time series of the i-th node. This indicates that the time series of upstream energy components is within a preset time window T. ck The average value, This represents the traffic time series of the i-th node within a preset time window T. ck The average value, where t represents the sampling time and τ is the set time offset;

[0054] Take ρ jd,i (τ) The value of τ that reaches its maximum value is used as the delay parameter τ of the i-th node. jd,i ;

[0055] The first-order lag time constant T of the i-th node is obtained by fitting the discrete first-order lag equation. jd,i The discrete first-order hysteresis equation is as follows:

[0056] ; where y jd,i (t) represents the component time series of the i-th node, where Δt represents the sampling time interval;

[0057] The objective function for determining the time constant is based on the discrete first-order lag equation, as follows:

[0058] ;

[0059] By minimizing the objective function of the time constant, the first-order lag time constant T is obtained. jd,i Further, generate the component time series y of the i-th node. jd,i (t);

[0060] Generate node energy characteristic time series, which includes: node component time series y jd,i (t), Nodal volume heat value time series Q jd,i (t) and the node Wobbi index time series W jd,i (t);

[0061] Time series of nodal volume heat values

[0062] ;

[0063] Among them, Q base k represents the volumetric calorific value reference constant. r Indicates the sensitivity coefficient for volumetric calorific value;

[0064] Node Warby Index Time Series

[0065] ;

[0066] Among them, W base k represents the baseline constant of the Warby index. w This represents the sensitivity coefficient of the Warby index;

[0067] Among them, the node composition time series reflects the composition state of the energy at the i-th node, the node volumetric calorific value time series reflects the thermal characteristics of the energy per unit volume at the i-th node, and the node Wobbe index time series reflects the combustion stability related characteristics of the energy at the i-th node.

[0068] The normalized cross-correlation coefficient is used to measure the temporal correlation between changes in upstream energy composition and changes in flow at the i-th node. The calculation requires first obtaining the time series of upstream energy composition (recording changes in upstream energy components over time) and the time series of flow at the i-th node (recording changes in energy flow at that node over time), and then calculating the average value of these two series within a preset time window (to eliminate the interference of short-term data fluctuations on the correlation calculation).

[0069] The time offset is used to set different time offsets (i.e., the time difference between the propagation of changes in upstream components to the node), and the cross-correlation coefficient is calculated once for each offset.

[0070] The larger the cross-correlation coefficient, the stronger the correspondence between the upstream component change and the node flow change at that time offset. The time offset corresponding to the maximum value of the coefficient is taken as the time delay parameter of the i-th node, that is, this parameter reflects the time required for the upstream energy component change to propagate to the node.

[0071] Discrete first-order hysteresis equations are used to simulate the lag process in which energy components, after reaching a node, do not immediately stabilize but slowly approach a stable value over time (such as the lag in component changes caused by gas storage in pipeline networks). In the equations, the node component time series is the energy component data of the node to be simulated, and the sampling time interval is the time difference between two data acquisitions (e.g., 5 minutes / time).

[0072] The objective function is the sum of squared errors between the simulated node components and the actual node components, aiming to make the simulation results as close as possible to the actual situation.

[0073] By minimizing this objective function, the first-order lag time constant is obtained. This parameter reflects the rate of change of energy components at the node (the larger the constant, the longer it takes for the components to stabilize after changing); combined with this constant, a node component time series that reflects the changes of node components over time is generated.

[0074] Node volumetric calorific value time series: Based on the volumetric calorific value benchmark constant (i.e., the heat per unit volume of energy under standard composition), plus the product of the volumetric calorific value sensitivity coefficient and the node composition time series. Since changes in node composition alter the heat per unit volume of energy (e.g., hydrogen blending reduces the heat per unit volume of natural gas), the sensitivity coefficient quantifies the degree of influence of compositional changes on the volumetric calorific value. The node volumetric calorific value time series reflects the thermal characteristics per unit volume of energy at the i-th node.

[0075] The node-specific Worby index time series is based on the Worby index baseline constant (i.e., the reference value for the combustion stability of energy under standard composition), plus the product of the Worby index sensitivity coefficient and the node composition time series. The Worby index is directly related to the combustion stability of energy; changes in composition will alter this index. The sensitivity coefficient is used to quantify the degree of influence of composition changes on combustion stability. The node-specific Worby index time series reflects the combustion stability characteristics of energy at the i-th node, ensuring the stable operation of downstream energy-consuming equipment (such as gas turbines).

[0076] In one embodiment of the present invention, constructing a system-level deviation time series and an energy reconciliation residual time series includes:

[0077] Constructing the system-level deviation time series P xt (t):

[0078] ;

[0079] Among them, P xt (t) The value of the system-level deviation time series at time t, μ jd,i (t) represents the value of the system routing weight time series at the i-th node at time t, Q jd,i R(t) represents the volumetric calorific value of the i-th node at time t, and R(t) represents the value of the settlement reference value time series at time t.

[0080] Constructing the energy reconciliation residual time series L xt (t):

[0081] ;

[0082] Among them, L xt F(t) represents the value of the energy reconciliation residual time series at time t. jd,i (t) represents the value of the flow time series of the i-th node at time t.

[0083] System routing weight time series (value of the i-th node at time t): reflects the importance of the i-th node in the system's energy allocation (such as the proportion of electrical / heating / cooling load undertaken by the node). The larger the weight, the stronger the node's influence on the overall energy balance of the system.

[0084] Volumetric calorific value time series (value at time t of the i-th node): The actual calorific value per unit volume of energy at a node (such as the actual calorific value per unit volume of natural gas at a certain time), which is an indicator of the actual energy calorific value.

[0085] Settlement reference value time series (value at time t): The standard heat reference value (such as the average daily heat value of a region) used for settlement in the system's accounts is the core indicator of the accounting settlement standard.

[0086] System-level deviation time series: First, calculate the difference between the actual volumetric calorific value and the book settlement reference value for each node. Then, weight the difference using the system routing weight of that node. Finally, sum the weighted results of all nodes to obtain the system-level deviation value at time t. Over time, the deviation values ​​at all times constitute the system-level deviation time series. From the perspective of energy allocation weights, this reflects the degree of deviation between the actual energy calorific value and the book settlement standard of the entire system. That is, if a node with a high weight has a large deviation, it will significantly increase the system-level deviation, reflecting the impact of key nodes on the deviation of the settlement standard.

[0087] Flow time series (value at time t of node i): The volume of energy flowing through node i at time t (such as the actual flow of natural gas at a certain time node), which is an indicator of actual energy consumption.

[0088] Energy reconciliation residual time series: First, calculate the difference between the book settlement reference value and the actual volumetric calorific value at each node (the difference is in the opposite direction to the system-level deviation, focusing on the energy difference between the book and the actual values). Then, weight the difference using the actual flow rate of that node (Energy = Flow Rate × Calorific Value; flow rate weighting quantifies the energy difference corresponding to the actual consumption). Finally, sum the weighted results of all nodes to obtain the energy reconciliation residual value at time t. Over time, the residual values ​​at all times constitute the energy reconciliation residual time series. From the perspective of actual energy consumption, this directly quantifies the difference between the total energy settled in the system's books and the total energy actually consumed by the nodes. That is, the larger the residual, the more significant the deviation between the book record and the actual consumption during energy reconciliation.

[0089] In one embodiment of the present invention, the characteristic parameters of the system-level deviation time series and the energy reconciliation residual time series at characteristic frequencies are extracted to determine the corresponding hysteresis characteristic quantities, including:

[0090] The system-level deviation time series and the energy reconciliation residual time series are compared within a preset time window T. ck The internal projection is calculated as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] ;

[0095] Among them, S zp For P xt (t) at the preset characteristic frequency f tz The sinusoidal projection quantity S yp For P xt (t) at the preset characteristic frequency ftz The cosine projection of S zc For L xt (t) at the preset characteristic frequency f tz The sinusoidal projection quantity S yc For L xt (t) at the preset characteristic frequency f tz The cosine projection quantity under the current;

[0096] Calculate the dominant frequency amplitude, phase, and phase difference at the characteristic frequency based on the projected value:

[0097] ;

[0098] ;

[0099] ;

[0100] ;

[0101] ;

[0102] Among them, A pc For P xt (t) at the preset characteristic frequency f tz The main frequency amplitude, A cc For L xt (t) at the preset characteristic frequency f tz The main frequency amplitude, For P xt (t) at the preset characteristic frequency f tz The main frequency phase, For L xt (t) at the preset characteristic frequency f tz The main frequency phase, For the phase difference, N ck Indicates the preset time window T ck The number of sampling times within the time frame.

[0103] Preset time window: Define a fixed time range (e.g., 1 hour) to focus on the sequence data within this fixed time range, avoid interference from single random fluctuations, and ensure that the extracted features are representative.

[0104] Preset characteristic frequency: The dominant frequency of changes in energy composition and load fluctuations in the system (such as the main frequency of intraday hydrogen blending adjustment and changes in electricity load) is the benchmark for analyzing the fluctuation pattern of the sequence.

[0105] The sinusoidal projection of the system-level deviation sequence: Within a preset time window, the system-level deviation sequence is decomposed according to the sinusoidal law of the preset characteristic frequency to obtain the degree of fit between the sequence and the sinusoidal law, reflecting the magnitude of the component of the deviation sequence that fluctuates sinusoidally at that frequency.

[0106] Cosine projection of system-level deviation sequence: Within a preset time window, the system-level deviation sequence is decomposed according to the cosine law of a preset characteristic frequency, reflecting the magnitude of the component of the deviation sequence that fluctuates cosinely at that frequency.

[0107] Sine projection of energy reconciliation residual sequence: Within a preset time window, the energy reconciliation residual sequence is decomposed according to a sinusoidal law of a preset characteristic frequency, reflecting the magnitude of the component of the residual sequence that fluctuates sinusoidally at that frequency.

[0108] Cosine projection of the energy reconciliation residual sequence: Within a preset time window, the energy reconciliation residual sequence is decomposed according to the cosine law of a preset characteristic frequency, reflecting the magnitude of the cosine fluctuation component of the residual sequence at that frequency.

[0109] The dominant frequency amplitude of the system-level deviation sequence is calculated based on the sine and cosine projections of the system-level deviation sequence, combined with the number of sampling moments within a preset time window (reflecting data density). This quantifies the strength of the fluctuation of the sequence at a preset characteristic frequency; that is, the larger the amplitude, the more significant the fluctuation of the deviation sequence at that frequency.

[0110] The dominant frequency amplitude of the energy reconciliation residual sequence is calculated based on the sine and cosine projections of the residual sequence and the number of sampling times, quantifying the strength of the fluctuation of the residual sequence at a preset characteristic frequency.

[0111] The dominant frequency phase of the system-level deviation sequence is calculated by the arctangent of the sine and cosine projections of the system-level deviation sequence. It reflects the time position of the fluctuation of the sequence at the preset characteristic frequency (such as the time when the fluctuation peak occurs) and is used to locate the fluctuation rhythm of the deviation sequence.

[0112] The dominant frequency phase of the energy reconciliation residual sequence: by calculating the arctangent of the sine and cosine projections of the residual sequence, it reflects the temporal position of the fluctuation of the residual sequence at a preset characteristic frequency.

[0113] Phase difference: The phase difference is obtained by subtracting the dominant frequency phase of the system-level deviation sequence from the dominant frequency phase of the residual sequence. This quantifies the time difference of fluctuations between the two sequences at a preset characteristic frequency. The larger the phase difference, the more significant the misalignment of the fluctuation rhythms between the residual and deviation sequences.

[0114] In one embodiment of the present invention, extracting the characteristic parameters of the system-level deviation time series and the energy reconciliation residual time series at characteristic frequencies, and determining the corresponding hysteresis characteristic quantity, further includes:

[0115] Based on the dominant frequency amplitude and phase difference, calculate the hysteresis characteristic of the preset time window:

[0116] ;

[0117] ;

[0118] Among them, S zh K represents the hysteresis area. zh Main axis ratio;

[0119] The hysteresis characteristic is obtained by combining the hysteresis area and the principal axis ratio.

[0120] The hysteresis area is calculated based on the dominant frequency amplitude of the system-level deviation sequence, the dominant frequency amplitude of the energy reconciliation residual sequence, and the phase difference. The sine value of the phase difference is an influencing factor; that is, the phase difference reflects the time misalignment of the two types of sequence fluctuations. The magnitude of the sine value directly determines the increase or decrease of the area: the larger the phase difference (within a reasonable range), the larger the sine value, and the larger the hysteresis area. The hysteresis area quantifies the size of the closed region formed by the fluctuations of the system-level deviation and the energy reconciliation residual at the characteristic frequency. The larger the area, the more severe the energy mismatch caused by the propagation delay of pipeline components and the deviation in settlement methods.

[0121] Based on the dominant frequency amplitude and phase difference of the two types of sequences, incorporating the square relationship of the amplitude and the cosine value of the phase difference, the ratio of the major axis to the minor axis of the hysteresis curve is calculated. The relative magnitude of the amplitude and the cosine value of the phase difference jointly determine the principal axis ratio: if the amplitude of one sequence is significantly greater than that of the other, or if the cosine value of the phase difference approaches an extreme value, the principal axis ratio will deviate significantly from 1. The principal axis ratio reflects the flattened circular shape of the hysteresis curve: when the principal axis ratio is close to 1, the curve is close to a perfect circle, indicating that the balance of the fluctuation intensity and time misalignment of the two types of sequences is good; when the principal axis ratio is much greater than 1, the curve is narrow and elongated, indicating that the fluctuation of one sequence dominates the mismatch, or the time misalignment leads to an unbalanced fluctuation direction, indirectly reflecting the morphological characteristics of energy mismatch (such as a deviation in a certain link being the main cause of mismatch).

[0122] Since the hysteresis area only reflects the degree of mismatch and the principal axis ratio only reflects the form of mismatch, neither parameter alone can fully characterize the energy mismatch pattern. Therefore, the two are combined to form a hysteresis characteristic quantity that combines both degree and form.

[0123] In one embodiment of the present invention, establishing a correlation between hysteresis characteristics and system energy consumption based on historical data, and generating an energy consumption correlation value, includes:

[0124] Determine the energy consumption of the target system at each time step to form the system energy consumption time series H. xt (t);

[0125] Determine the system energy consumption time series H xt(t) within the preset time window T ck Total system energy consumption E all ;

[0126] Obtain M historical calculation windows within a historical time period, and extract the hysteresis feature corresponding to the m-th historical calculation window: hysteresis area. Compared to the spindle The historical calculation window size is equal to the preset time window.

[0127] The design matrix X is constructed based on the hysteresis characteristics of each historical calculation window. ls With the observation vector y ls :

[0128] , ;

[0129] The parameter vector is calculated using the least squares method. ;in, Indicates the transpose operation;

[0130] Extracting the parameter vector θ gl The first element yields the constant term parameter θ. cl ;

[0131] Extracting the parameter vector θ gl The second element yields the hysteresis area coefficient θ. mj ;

[0132] Extracting the parameter vector θ gl The third element yields the principal axis ratio coefficient θ. zz ;

[0133] Based on the hysteresis area and principal axis ratio, combined with the parameter vector θ gl Calculate energy consumption correlation values:

[0134] ;

[0135] Among them, V nhgl This represents the energy consumption correlation value for a preset time window.

[0136] System energy consumption time series: Through real-time monitoring or historical records, the actual energy consumption of the target system at each operating moment is collected (e.g., the total power consumption and gas consumption converted to total energy consumption are recorded every 5 minutes), and then arranged into a sequence in chronological order. The system energy consumption time series completely records the temporal variation pattern of system energy consumption and serves as the original data source for subsequent calculations of window-level total energy consumption, ensuring the continuity and timeliness of energy consumption data.

[0137] The total system energy consumption within the preset time window achieves the unification of energy consumption data granularity and hysteresis feature granularity. That is, since the hysteresis feature is calculated based on the preset time window, only by aggregating energy consumption into window-level total energy consumption can a one-to-one correspondence between the two be established, avoiding correlation deviations caused by granularity differences.

[0138] The design matrix is ​​a table formed by arranging the hysteresis features of M historical windows (each window contains 1 hysteresis area and 1 principal axis ratio) in a fixed format (each row corresponds to the feature of one historical window), which carries the input features; the observation vector is a column vector formed by arranging the total energy consumption of M historical windows in order, which carries the output results.

[0139] By using the least squares method, combined with the design matrix and observation vectors, a set of optimal parameter vectors is calculated. The parameter vectors are a quantitative representation of the relationship between hysteresis characteristics and energy consumption, with each element corresponding to the influence coefficient of a characteristic on energy consumption.

[0140] Constant parameters: These reflect the basic energy consumption level of the system when there is no significant mismatch (hysteresis area and principal axis ratio approach the ideal value), and are the benchmark parameters for energy consumption;

[0141] Hysteresis area coefficient: quantifies the impact of hysteresis area (mismatch degree) on energy consumption. That is, when the coefficient is positive, the larger the hysteresis area, the higher the energy consumption.

[0142] Main axis ratio coefficient: quantifies the influence of the main axis ratio (mismatch mode) on energy consumption, that is, the sign and magnitude of the coefficient reflect the changing trend of energy consumption under different mismatch modes.

[0143] The energy consumption correlation value transforms the current system mismatch state (hysteresis characteristic) into a quantitative energy consumption index that can be directly used for optimization. The higher the energy consumption correlation value, the higher the energy consumption caused by the current mismatch state.

[0144] In one embodiment of the present invention, using the system routing weight time series and the settlement reference value adjustment time series as decision variables, under system operation constraints, the system energy cost and energy consumption correlation value is minimized to obtain the target decision variables, including:

[0145] Define decision variables, including: routing weight time series μ jd,i (t) and the time series of settlement reference value adjustment ΔR(t);

[0146] Define the time series after adjusting the settlement reference value ;

[0147] Construct system-level bias-adjusted time series and energy reconciliation residual-adjusted time series for optimization calculations:

[0148] ;

[0149] ;

[0150] Among them, P xttz (t) represents the value of the system-level bias-adjusted time series at time t, L xttz (t) represents the value of the energy reconciliation residual adjusted time series at time t.

[0151] Within a preset time window and a preset characteristic frequency, the corresponding hysteresis area adjustment value S is calculated based on the system-level deviation adjusted time series and the energy reconciliation residual adjusted time series, respectively. tzzh and spindle ratio adjustment value K tzzh ;

[0152] Based on the hysteresis area adjustment value S tzzh and spindle ratio adjustment value K tzzh Calculate the corresponding adjusted energy consumption correlation value. ;

[0153] Define the system energy cost within a preset time window:

[0154] ;

[0155] Among them, C nl Indicates the system's energy cost;

[0156] Construct the objective function J for the decision variables:

[0157] ;

[0158] Where, λ nl The weighting coefficient λ represents the system energy cost. gl This represents the weighting coefficient of the adjusted energy consumption correlation value;

[0159] System operating constraints include:

[0160] ;

[0161] ;

[0162] ;

[0163] Among them, B min (t) represents the minimum value of the settlement reference value adjustment, B max (t) represents the maximum value of the settlement reference value adjustment;

[0164] Find the time series of routing weights that minimize the objective function J of the decision variable. Time series of settlement reference value adjustments As a target decision variable.

[0165] Routing weight time series: controls the allocation ratio of energy from each node to different loads such as electricity, heat, and cooling (e.g., whether natural gas is prioritized for power generation or heating at a certain node), directly affecting energy utilization efficiency: selecting it as a variable is to reduce misallocation energy consumption by optimizing energy destination.

[0166] Settlement reference value adjustment time series: Correcting the original fixed settlement standard (such as the average daily calorific value of the zone) to make it closer to the actual energy characteristics of the node: Selecting it as a variable is to reduce the deviation between the book standard and the actual characteristics, thereby reducing the additional costs corresponding to the reconciliation residuals.

[0167] Based on the adjusted settlement reference value, the system-level deviation (the sum of the deviations between the actual heat value of the node weighted by the routing weight and the new settlement value) and the energy reconciliation residual (the sum of the deviations between the new settlement value of the node weighted by the node flow and the actual heat value of the node) are recalculated to obtain the adjusted deviation and residual sequence, which directly reflects the system mismatch state after the decision variables are adjusted.

[0168] The correlation between the adjusted hysteresis characteristic and energy consumption is calculated, thereby transforming the mismatch state after the adjustment of decision variables into a quantifiable additional energy consumption indicator.

[0169] System energy cost: The total cost of the system's actual purchase of energy (such as natural gas and electricity) (e.g., gas purchase price × gas volume + electricity purchase price × electricity volume), is an indicator reflecting the system's economic efficiency.

[0170] The system energy cost and the adjusted energy consumption correlation value are multiplied by their respective weighting coefficients and then summed. These weighting coefficients balance economic efficiency and energy conservation (e.g., increasing the weight of the energy consumption correlation value to prioritize energy conservation, and increasing the weight of the cost to prioritize economic efficiency). This integrates the two objectives of reducing costs and minimizing mismatched energy consumption into a single, minimizeable objective function, avoiding the pitfalls of single-objective optimization (e.g., reducing costs while increasing mismatched energy consumption).

[0171] The total energy allocation ratio of all nodes is 100%, which conforms to the actual allocation rule of energy conservation (e.g., natural gas of a certain node can only be allocated to electricity, heat and cooling loads, and cannot be over-allocated or under-allocated).

[0172] The routing weight is between 0 and 1: the proportion of a single node allocated to a certain load cannot be negative (negative energy cannot be allocated) nor can it exceed 100% (over-allocation is not allowed), ensuring that the allocation rules are legal and compliant.

[0173] The adjustment amount of the settlement reference value shall be between the positive and negative maximum values: the adjustment amount shall not exceed the range allowed by policies, contracts or measurement specifications (e.g., the deviation of the settlement calorific value in a certain region shall not exceed ±5%), so as to avoid settlement disputes caused by illegal adjustments.

[0174] Under the premise of satisfying all operational constraints, a set of time series for routing weights and time series for settlement reference value adjustments is found through mathematical optimization algorithms (such as gradient descent and linear programming) to minimize the objective function (cost + energy consumption correlation value). The objective decision variable is the optimal solution that balances the lowest economic cost and the least mismatch energy consumption.

[0175] In one embodiment of the present invention, applying the target decision variable to the target system includes:

[0176] The target decision variables are issued to the target system as control commands.

[0177] Applying the target decision variables (i.e., the optimized system routing weight time series and settlement reference value adjustment time series) to the target system involves transforming these two types of system-level control parameters into standardized control commands adapted to the existing operational interfaces of the target system and issuing them.

[0178] On the one hand, the target decision variables need to be formatted and adapted to ensure that the instructions include a clear timestamp (consistent with the time granularity of the system operation, such as 5-15 minutes / segment), node identifier (corresponding to each energy node), specific parameter values ​​(such as the routing weight ratio of a certain node in a certain period of time, the adjustment range of the settlement reference value), and a check code (to avoid data transmission errors).

[0179] On the other hand, the routing weight time series command needs to be distributed according to parameter type. That is, the routing weight time series command is sent to the Energy Hub routing management module of the target system to adjust the distribution ratio of energy to different loads such as electricity, heat and cooling in real time (such as increasing the routing weight of natural gas to the heating side of a certain node to match the current heating demand and reduce energy mismatch). The settlement reference value adjustment time series command is sent to the system's settlement management module to update the calorific value reference standard of the book settlement within the scope permitted by policy (such as adjusting the settlement reference value according to the actual quality of the current node to reduce the deviation between the book and actual energy characteristics).

[0180] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A multi-source integrated energy system energy-saving optimization platform based on data mining, characterized in that: include: The data acquisition module collects characteristic parameters of the target system, including: upstream energy composition time series, flow time series of each node, settlement reference value time series, and system routing weight time series; The data processing module identifies the component propagation characteristic parameters of each node by combining characteristic parameters, generates a time series of node energy characteristics, and constructs a system-level deviation time series and an energy reconciliation residual time series, including: Constructing the system-level deviation time series P xt (t): ； Among them, P xt (t) represents the value of the system-level deviation time series at time t, μ jd,i (t) represents the value of the system routing weight time series at the i-th node at time t, Q jd,i R(t) represents the volumetric calorific value of the i-th node at time t, and R(t) represents the value of the settlement reference value time series at time t. Constructing the energy reconciliation residual time series L xt (t): ； Among them, L xt F(t) represents the value of the energy reconciliation residual time series at time t. jd,i (t) represents the value of the flow time series of the i-th node at time t; The correlation analysis module extracts the characteristic parameters of the system-level deviation time series and the energy reconciliation residual time series at characteristic frequencies, and determines the corresponding hysteresis characteristic quantities, including: The system-level deviation time series and the energy reconciliation residual time series are compared within a preset time window T. ck The internal projection is calculated as follows: ； ； ； ； Among them, S zp For P xt (t) at the preset characteristic frequency f tz The sinusoidal projection quantity S yp For P xt (t) at the preset characteristic frequency f tz The cosine projection of S zc For L xt (t) at the preset characteristic frequency f tz The sinusoidal projection quantity S yc For L xt (t) at the preset characteristic frequency f tz The cosine projection quantity under the current; Calculate the dominant frequency amplitude, phase, and phase difference at the characteristic frequency based on the projected value: ； ； ； ； ； Among them, A pc For P xt (t) at the preset characteristic frequency f tz The main frequency amplitude, A cc For L xt (t) at the preset characteristic frequency f tz The main frequency amplitude, For P xt (t) at the preset characteristic frequency f tz The main frequency phase, For L xt (t) at the preset characteristic frequency f tz The main frequency phase, For the phase difference, N ck Indicates the preset time window T ck Number of sampling times within; Based on the dominant frequency amplitude and phase difference, calculate the hysteresis characteristic of the preset time window: ； ； Among them, S zh K represents the hysteresis area. zh Main axis ratio; The hysteresis characteristic quantity is obtained by combining the hysteresis area and the principal axis ratio; Establish the correlation between hysteresis characteristic quantities and system energy consumption based on historical data, and generate energy consumption correlation values; The optimization decision module uses the system routing weight time series and the settlement reference value adjustment time series as decision variables. Under system operation constraints, it minimizes the correlation between system energy cost and energy consumption to obtain the target decision variable. The execution module applies the target decision variables to the target system.

2. The energy-saving optimization platform for multi-source integrated energy systems based on data mining according to claim 1, characterized in that, By combining characteristic parameters to identify the component propagation characteristic parameters of each node, a time series of node energy characteristics is generated, including: Calculate the normalized cross-correlation coefficient ρ jd,i (τ): ； Where C(t) represents the time series of upstream energy components, F jd,i (t+τ) represents the flow time series of the i-th node. This indicates that the time series of upstream energy components is within a preset time window T. ck The average value, This represents the average value of the traffic time series of the i-th node within a preset time window Tck, where t represents the sampling time and τ is the set time offset. Take ρ jd,i (τ) The value of τ that reaches its maximum value is used as the delay parameter τ of the i-th node. jd,i ; The first-order lag time constant T of the i-th node is obtained by fitting the discrete first-order lag equation. jd,i The discrete first-order hysteresis equation is as follows: ； Among them, y jd,i (t) represents the component time series of the i-th node, where Δt represents the sampling time interval; The objective function for determining the time constant is based on the discrete first-order lag equation, as follows: ； By minimizing the objective function of the time constant, the first-order lag time constant T is obtained. jd,i Further, generate the component time series y of the i-th node. jd,i (t); Generate node energy characteristic time series, which includes: node component time series y jd,i (t), Nodal volume heat value time series Q jd,i (t) and the node Wobbi index time series W jd,i (t); Time series of nodal volume heat values ； Among them, Q base k represents the volumetric calorific value reference constant. r Indicates the sensitivity coefficient for volumetric calorific value; Node Warby Index Time Series ； Among them, W base k represents the baseline constant of the Warby index. w This represents the sensitivity coefficient of the Warby index; Among them, the node composition time series reflects the composition state of the energy at the i-th node, the node volumetric calorific value time series reflects the thermal characteristics of the energy per unit volume at the i-th node, and the node Wobbe index time series reflects the combustion stability related characteristics of the energy at the i-th node.

3. The energy-saving optimization platform for multi-source integrated energy systems based on data mining according to claim 2, characterized in that, Based on historical data, establish the correlation between hysteresis characteristics and system energy consumption, and generate energy consumption correlation values, including: Determine the energy consumption of the target system at each time step to form the system energy consumption time series H. xt (t); Determine the system energy consumption time series H xt (t) within the preset time window T ck Total system energy consumption E all ; Obtain M historical calculation windows within a historical time period, and extract the hysteresis feature corresponding to the m-th historical calculation window: hysteresis area. Compared to the spindle The historical calculation window size is equal to the preset time window. The design matrix X is constructed based on the hysteresis characteristics of each historical calculation window. ls With the observation vector y ls : ， ； The parameter vector is calculated using the least squares method. ； in, Indicates the transpose operation; Extracting the parameter vector θ gl The first element yields the constant term parameter θ. cl ; Extracting the parameter vector θ gl The second element yields the hysteresis area coefficient θ. mj ; Extracting the parameter vector θ gl The third element yields the principal axis ratio coefficient θ. zz ; Based on the hysteresis area and principal axis ratio, combined with the parameter vector θ gl Calculate energy consumption correlation values: ； Among them, V nhgl This represents the energy consumption correlation value for a preset time window.

4. The energy-saving optimization platform for multi-source integrated energy systems based on data mining according to claim 3, characterized in that, Using the system routing weight time series and the settlement reference value adjustment time series as decision variables, and under system operation constraints, the goal is to minimize the correlation between system energy cost and energy consumption to obtain the target decision variables, including: Define decision variables, including: the value μ of the system routing weight time series at the i-th node at time t. jd,i The values ​​of ΔR(t) and the settlement reference value adjustment time series at time t; Define the time series after adjusting the settlement reference value ; Construct system-level bias-adjusted time series and energy reconciliation residual-adjusted time series for optimization calculations: ； ； Among them, P xttz (t) represents the value of the system-level bias-adjusted time series at time t, L xttz (t) represents the value of the energy reconciliation residual adjusted time series at time t. Within a preset time window and a preset characteristic frequency, the corresponding hysteresis area adjustment value S is calculated based on the system-level deviation adjusted time series and the energy reconciliation residual adjusted time series, respectively. tzzh and spindle ratio adjustment value K tzzh ; Based on the hysteresis area adjustment value S tzzh and spindle ratio adjustment value K tzzh Calculate the corresponding adjusted energy consumption correlation value. ； Define the system energy cost within a preset time window: ； Among them, C nl Indicates the system's energy cost; Construct the objective function J for the decision variables: ； Where, λ nl The weighting coefficient λ represents the system energy cost. gl This represents the weighting coefficient of the adjusted energy consumption correlation value; System operating constraints include: ； ； ； Among them, B min (t) represents the minimum value of the settlement reference value adjustment, B max (t) represents the maximum value of the settlement reference value adjustment; Find the time series of routing weights that minimize the objective function J of the decision variable. Time series of settlement reference value adjustments As a target decision variable.

5. The energy-saving optimization platform for multi-source integrated energy systems based on data mining according to claim 4, characterized in that, Applying the target decision variables to the target system includes: The target decision variables are issued to the target system as control commands.

Images