Controlling Resistance in Well System

Some implementations relate to operations for hydraulic fracturing in a well. More specifically, some implementations relate to controlling resistance in a well system.

The hydraulic fracture treatment process for a stage may be determined by a pre-job design that may attempt to optimize certain performance metrics. One such metric may be the designed stage resistance at the end of the hydraulic fracture treatment. During the treatment process, stage resistivity may not be trending correctly to achieve the desired pressure at the end of the treatment. This may lead to poor fracture treatment efficiency where the performance metrics are not achieved before running out of resources. Alternatively, the hydraulic fracture treatment may be successful, but not optimized with respect to time and cost.

The description that follows may include example systems, methods, techniques, and program flows that embody implementations of the disclosure. However, this disclosure may be practiced without these specific details. For clarity, some well-known instruction instances, protocols, structures, and techniques may not be shown in detail.

Efficiency of a hydraulic fracture treatment process may be improved by periodically monitoring the resistance during the hydraulic fracture treatment and adjusting treatment factors to optimally drive the resistance to the desired value. These factors may include breakdown sequence, proppant pumping scheme, rate/proppant cycling sequences, diverter drop information or any application of formation conditioning agents, remaining resources (e.g. fluid/proppant volume), and any other suitable aspect of the hydraulic fracture treatment process. In this disclosure, a factor vector refers to one or more factors. The factors are described in more detail below.

Some implementations may monitor (periodically, continuously, or otherwise) flow resistance of a wellbore or boundary (also referred to herein as “resistance”) throughout a hydraulic fracture treatment. During the hydraulic fracture treatment, some implementations may attempt to achieve or maintain a target resistance (i.e., a desired resistance for the treatment) in the well. To achieve or maintain the target resistance, some implementations may compute the current resistance from the amplitude and decay rate of water hammers generated from rapid drops in pressure and flow rate. Some implementations utilize the current resistance to select different factor vectors that have a likelihood of achieving (or maintaining) the target resistance over the course of the hydraulic fracture treatment. For example, a controller may select a factor vector for a first time interval of the hydraulic treatment. Depending on the current resistance, the controller may modify the factor vector (or select a completely different factor vector) for the next time interval of the hydraulic fracture treatment. Periodic updates to the factor vector may be repeated until the target resistance is achieved (or is likely). Although factor vectors may be updated during a hydraulic fracture treatment, some implementations also may operate across treatments by using a previous treatment's factor vector and resistance to determine a later treatment's factor vector. This approach of optimizing the factor vector in response to conditions (such as current resistance) during intervals of the hydraulic fracture treatment may increase production and reduce waste of fluid, proppant, and other material pumped for the each well.

Some implementations may train a learning machine (such as a neural network or other suitable machine learning model) to predict a factor vector that is likely to achieve or maintain a specified target resistance for the next time interval of the hydraulic fracture treatment. The learning machine may make the prediction based on the current factor vector, current resistance, and target resistance. Additional details about training the learning machine are described below.

Some implementations periodically compute resistance during the hydraulic fracture treatment from the amplitude and decay rate of a pressure water hammer measured at the well head.is a graph showing curves representing pressure and flow of a water hammer. In, the graphincludes a plot of pressure and a plot of flow rate for a water hammer during a hydraulic fracture treatment. During hydraulic fracture treatments, some implementations use the water hammer to compute the resistance of the boundary. The rising and falling edgesof the pressure plot may constitute a boundary condition with which some implementations compute resistance based on the water hammer. The pressure pulses of the water hammer may occur one or more times during one or more hydraulic fracture treatments.

The wellbore can be modeled using the following equations:

In Equations 1-3, C is the capacitance, H is the head, t is the time, Q is the flowrate, x is the spatial dimension (measured depth), I is the inductance, R is the resistance, g is gravity, f is the friction coefficient, D is diameter, and A is the area of cross section. In some implementations, values for one or more of the variables noted herein may be measured or derived from one or more sensors in the well or at the surface.

The resistance of the boundary can be modeled by

The parameters R, C, and I associated with the wellbore may be known, and all the boundary conditions may be known. Thus, the only unknown, R*, may be computed using an inversion process by using the measured pressure pulse signature (). In the example here,

may be computed via inversion process.

Many sets of sample data

may obtained, where FVmay represent the factor vector that was in-use when

was computed. FVmay represent the factor vector of the next interval of the treatment. Each factor vector may include all the treatment data up to the instance of measurement (pressure, rate, chemicals, proppant concentration, etc.), wellbore geometry detail, perforation design and fluid properties, etc.

Some implementations train a machine learning model to learn from data patterns and to make decisions without knowledge of the explicit relationships.is a diagram illustrating example operations for supervised training. Training samples from a training data set may be used to train the machine learning model. Each training sample may include the following features: current factor vector (FV), current computed resistance

resistance for the next interval

and the next factor vector (FV). Using the current factor vector (FV), current resistance

and the next resistance

the machine learning model can predict the next factor vector (predicted FV) that is likely to achieve the next resistance

The predicted next factor (predicted FV) is compared to the known next factor vector from the sample data (FV). If the predicted factor next factor vector does not match the known next factor vector, the machine learning model may be updated (such as by updating, weights, biases, activation functions, etc.). This process may repeat for all sample in the training data set and for multiple training data sets.

After the machine learning model is trained, it may be inverted to determine next factor vectors that may achieve a target resistance for a hydraulic fracture treatment. The target resistance is a desired resistance for the hydraulic fracture treatment. The target resistance may come from different sources—for example, operators or system components may desire to have the target resistance be a certain fraction of design perforation resistance. The design perforation resistance can be calculated using:

where rate Q, number of perforations N, discharge coefficient C, and hydraulic perforation diameter h=d√{square root over (C)}, where d is a perforation diameter. For example, operators or system components may desire to have target resistance to be 90% of. Other sources of target resistance may involve using correlation between resistance and cluster flow distribution metric, such as uniformity index. There may be any suitable source of target resistance.

In some implementations, the machine learning model may be implemented via a neural network. The neural network may be configured to learn a function that transforms input data into meaningful predictions or classifications about factor vectors of a hydraulic fracture treatment. The function may be defined by aspects of the neural network such as weights, biases, activation functions, and other functionality of the neural network.is diagram illustrating an example neural network that may be used in conjunction with some implementations. The neural networkmay include a plurality of neurons. The neural networkalso may include an input layer having any suitable number of neurons(supporting any suitable number of features). The input layer may intake information (sometimes referred to as features) indicating a current resistance, the designed target resistance for the hydraulic fracture treatment, and the current factor vector. The neural networkalso may include an output layer that predicts an optimal factor vector for the next time interval of the hydraulic fracture treatment based on the information fed into the input layer-that is, based on the current resistance, target resistance, and current factor vector. The output layer may include any suitable number of neurons. The neural network may be embodied in the factor vector unitvia computer-executable instructions, hardware, circuitry, and/or other logic for performing the functionality described herein.

In some implementations, the factor vector unitmay be integrated into a computer system.is a block diagram illustrating a computer system that may be utilized with some implementations. In, a computer systemmay include one or more processorsconnected to a system bus. The system busmay be connected to memoryand a network interface. The memorymay include any suitable memory random access memory (RAM), non-volatile memory (e.g., magnetic memory device), and/or any device for storing information and instructions executable by the processor(s). The network interfacemay provide connectivity to any suitable network, such as a wired network, wireless network, satellite network, etc.

The computer systemmay include additional peripheral devices. For example, the computer systemmay include multiple external multiple processors. In some implementations, any of the components can be integrated or subdivided.

The computer systemalso may include a factor vector unit. The factor vector unitmay implement the methods and operations described herein. The factor vector unitmay include the neural network(as described herein). The factor vector unitmay include any suitable instructions, media, circuitry, and/or logic for performing the operations described herein. In some implementations, the computer systemmay be included in the well system (such as the well system described with reference to) and may cooperate with other components and/or systems to perform the functionality described herein.

The computer systemalso may include a resistance unitconfigured to perform operations for determining resistance during a well treatment. The resistance unitmay determine the resistance (such as a current resistance) by utilizing one or more of the equations described herein. The resistance unitalso may perform computations for determining design resistance (see also Equation 5).

The computer systemalso may include a fracturing controllerconfigured to perform operations for controlling fracturing operations in a well. The fracturing controllermay respond to output from the factor vector unit. For example, the fracturing controllermay implement a new factor vector during a fracturing treatment, where the factor vector unitunit outputs the new factor vector.

Although the components are shown separately, any of the components of the computer systemmay be further combined or subdivided. For example, the factor vector unitand fracturing controllermay be combined into a single component or subdivided into three or more components. Any component of the computer systemcan be implemented as hardware, firmware, and/or machine-readable media including computer-executable instructions for performing the operations described herein. For example, some implementations include one or more non-transitory machine-readable media including computer-executable instructions including program code configured to perform functionality described herein. Machine-readable media includes any mechanism that provides (e.g., stores and/or transmits) information in a form readable by a machine (e.g., a computer system). For example, tangible machine-readable media includes read only memory (ROM), random access memory (RAM), magnetic disk storage media, optical storage media, flash memory machines, etc. Machine-readable media also includes any media suitable for transmitting software over a network.

As noted, some implementations determine a current resistance of a hydraulic fracture treatment. These implementations may utilize the current resistance to determine a next factor vector that is likely to achieve a target resistance during hydraulic fracture treatments.is a flow diagram illustrating operations for determining factor vectors during one or more hydraulic fracture treatments. In, the flowbegins at block, where the fracturing controllermay initialize a factor vector for a hydraulic fracture treatment and commence the hydraulic fracture treatment.

At block, the factor vector unitmay record one or more water hammer pressure pulses. In some implementations, the factor vector unitrecords at least two water hammer pressure pulses before moving to block.shows an example water hammer pressure pulse.

At block, the resistance unitmay determine, based on one or more water hammer pressure pulses, target and current resistances for the current factor vector. In some implementations, the target resistance may be determined earlier as part of the design process for the hydraulic fracture treatment. However, in some implementations, the resistance unitmay determine the target resistance based on well conditions, factors in the factor vector, and/or other data. As for the current resistance, the resistance unitmay utilize one or more the equations described herein to determine the current resistance in the well based on the first water hammer pressure pulse.

At block, the factor vector unitmay obtain the current factor vector, current resistance, and target resistance. At this point, the factor vector unitis already trained (as described above) and capable of predicting a new factor vector during the hydraulic fracture treatment, where the new factor vector is likely to achieve the target resistance.

At block, the factor vector unitmay predict a new factor vector and continue the hydraulic fracture treatment using the new factor vector.

At block, the resistance unitmay obtain one or more water hammer pressure pulses.

At block, the resistance unitmay determine a current resistance based on the water hammer pressure pulse(s).

At block, the resistance unitmay compare the current resistance with the target resistance. If the current resistance is above a resistance threshold (such as the current resistance is unacceptable) (see block), operations continue at block. By continuing at block, the fracturing controllermay find a new factor vector that achieves a current resistance closer to the target resistance. If the current resistance is below a resistance threshold (such as the current resistance is acceptable) (see block), operations continue at block.

At block, the fracturing controllermay determine that current resistance was achieved or at least was acceptable (such as below the resistance threshold).